Datasets:

ccdv
/

arxiv-summarization

Dataset card Data Studio Files Files and versions Community

Subset (2)

section · 216k rows

Split (3)

train · 203k rows

article stringlengths 0 745k	abstract stringlengths 0 92.7k
the current observed sample of galactic black - hole ( bh ) x - ray binaries ( xrbs ) provides us with a unique opportunity for understanding the formation of black holes in binaries . in recent years the sample has grown significantly , reaching 18 bh xrbs identified through dynamical mass measurements ( orosz 2004 , mcclintock & remillard 2004 ) . for these systems there exists a wealth of observational information about their current physical state : bh and donor masses , orbital period , donor s position on the h - r diagram and surface chemical composition , transient or persistent and roche - lobe overflow ( rlo ) or wind - driven character of the mass - transfer ( mt ) process , and distances . even more recently proper motions have been measured for a handful of these systems ( mirabel et al . 2001 , 2002 ; mirabel & rodrigues 2003a ) , complementing the earlier measurements of center - of - mass radial velocities and giving us information about the 3-dimensional kinematic properties of these binaries . similar information exists for some of the known xrbs with neutron star ( ns ) accretors ( ankay et al . 2001 ; rib et al . 2002 ; mirabel & rodrigues 2003b ; mirabel , rodrigues , & liu 2004 ) . understanding the events of compact object formation in these known systems requires that we investigate their evolutionary history from the present state back to the time just prior to the core collapse event . in what follows we describe how the currently available observational constraints for xrbs can be used to uncover this past history and provide us with a consistent picture for how the bhs and nss were formed . the ultimate goal of this project is to examine the systematics of the derived compact object progenitor masses and requirements for supernova ( sn ) kicks and their dependence on compact object masses and associated mass loss at the core collapse events . such a comprehensive investigation will not just reveal the origin of these galactic xrbs , but may also offer insight to the physics of supernovae and the possible association of natal kicks with bh formation . although this is not the first study that has considered the evolutionary history of known xrbs in the galaxy ( see , e.g. , de kool , van den heuvel , & pylyser 1987 ; brandt , podsiadlowski , & sigurdsson 1995 ; nelemans , tauris , & van den heuvel 1999 ; podsiadlowski , rappaport , & pfahl 2002a ; podsiadlowski , rappaport , & han 2003 ; gualandris et al . 2004 ) , it is the first one that incorporates such a large number of observational constraints for a whole sample of systems and considers the complete set of evolutionary processes from the present back to the binary stage right before compact object formation . in this first of a series of papers , we focus on a detailed description of our analysis methodology and the rationale for combining the current constraints to complete a self - consistent picture for the evolutionary history . we also focus on the computational methods used and the uncertainties involved , and apply our analysis to the well - studied soft x - ray transient groj1655 - 40 ( nova sco 1994 ) . in subsequent papers we will present our results for the rest of bh xrbs and for the ns xrbs with all the above constraints available . the plan of the paper is as follows . in 2 we briefly review the currently available constraints on the properties of the soft x - ray transient groj1655 - 40 . a general outline of the analysis used to reconstruct the system s evolutionary history is presented in 3 , while the individual steps of the analysis are discussed in more detail in 47 . in 8 , we combine these steps and derive constraints on the binary s pre - sn orbital separation , on the mass of the bh s helium star progenitor , and on the kick velocity that may have been imparted to the bh at birth . in 9 , hydrodynamic core - collapse simulations are presented and the nature of the bh progenitor is constrained even further . the final section is devoted to a discussion of our results and some concluding remarks . lccc distance & @xmath5 & @xmath6kpc & ( hjellming & rupen 1995 ) + galactic longitude & @xmath7 & @xmath8 & ( tingay et al . 1995 ) + galactic latitude & @xmath9 & @xmath10 & ( tingay et al . 1995 ) + velocity towards the galactic center & @xmath11 & @xmath12kms@xmath1 & ( mirabel et al . 2002 ) + velocity in the direction of the galactic rotation & @xmath13 & @xmath14kms@xmath1 & ( mirabel et al . 2002 ) + velocity towards the northern galactic pole & @xmath15 & @xmath16kms@xmath1 & ( mirabel et al . 2002 ) + & & greene et al . ( 2001 ) & beer & podsiadlowski ( 2002 ) + orbital period & @xmath17 & @xmath18days & @xmath19days + black hole mass & @xmath20 & @xmath21 & @xmath22 + donor mass & @xmath23 & @xmath24 & @xmath25 + donor luminosity & @xmath26 & @xmath27 & @xmath28 + donor effective temperature & @xmath29 & @xmath30k & @xmath31k the soft x - ray transient groj1655 - 40 was first detected on 27 july 1994 by the burst and transient source experiment ( batse ) on board the compton gamma - ray observatory ( zhang et al . the optical counterpart to the x - ray source was discovered by bailyn et al . ( 1995a ) . shortly thereafter , bailyn et al . ( 1995b ) found periodic optical eclipses of unequal depth which are most likely caused by alternating eclipses of the optical component and the bh s accretion disk . the same authors also derived the orbital period to be @xmath32days and used the semi - amplitude of the radial - velocity variations to derive a minimum mass for the compact object of @xmath33 , strongly supporting the bh nature of the compact star . since then , numerous investigations have been devoted to unraveling the properties of the bh and its companion . spectroscopic observations led orosz & bailyn ( 1997 ) to classify the optical component as an f3iv f6iv star ( see also bailyn et al . 1995b ) , while van der hooft et al . ( 1998 ) used photometric measurements to refine the orbital period measurement to @xmath19days . the case for the bh nature of the accreting object was strengthened considerably by soria et al . ( 1998 ) who derived a lower limit for the mass of the compact star of @xmath34 , and by phillips , shahbaz & podsiadlowski ( 1999 ) and shahbaz et al . ( 1999 ) who constrained the compact object mass to be @xmath35 and @xmath36 , respectively . evidence for the occurrence of a sn explosion was presented by israelian et al . ( 1999 ) who reported a large overabundance of oxygen , magnesium , silicon , and sulfur in the atmosphere of the optical component . since the latter is not massive enough to create these elements , israelian et al . ( 1999 ) attributed the overabundances to the accretion of sn material from the bh progenitor . the distance of groj1655 - 40 from the sun , finally , was determined by hjellming & rupen ( 1995 ) using a kinematic model for the jets of the system . they derived @xmath37kpc , which is in excellent agreement with previous distance estimates by mckay & kesteven ( 1994 ) , tingay et al . ( 1995 ) , and bailyn et al . ( 1995a ) . more recently greene , bailyn , & orosz ( 2001 ) and beer & podsiadlowski ( 2002 ) have undertaken detailed studies of the ellipsoidal light variations when the system is in quiescence and derived rather different constraints for the two component masses and the position of the donor in the h - r diagram . greene et al . ( 2001 ) ( hereafter gbo ) used spherical nextgen atmosphere models for a fixed effective temperature of 6336k ( corresponding to a spectral type of f6iii ) to model the observed b , v , i , j , k light curves without any contribution from an accretion disk surrounding the bh . the model depends on two free parameters : the mass ratio @xmath38 , where @xmath20 is the mass of the bh and @xmath23 the mass of the companion , and the orbital inclination @xmath39 . as additional constraints , gbo used the radial - velocity curve derived by shahbaz et al . ( 1999 ) and the projected rotational velocity @xmath40 derived by israelian et al . the authors found a bh mass @xmath41 . beer & podsiadlowski ( 2002 ) ( hereafter bp ) fitted the ellipsoidal light variations in the photometric b- , v- , r- , and i - bands using kurucz models for the atmosphere of the roche - lobe filling secondary and a bh accretion disk model with a flat temperature profile . their model is characterised by five free parameters : the mass ratio @xmath38 , the orbital inclination @xmath39 , the polar temperature @xmath42 of the donor star , the color excess @xmath43 , and the distance @xmath5 from the sun . they furthermore adopted the distance of @xmath44kpc derived by hjellming & rupen ( 1995 ) as an additional ad - hoc constraint to further tighten the limits on the derived model parameters . in their best - fitting model , bp found @xmath45 . at present it does not seem clear which of the two analyses ( gbo or bp ) yields the most reliable measurements ( see also shahbaz 2003 ) . in this study we therefore consider both sets of constraints and study them as two separate cases . an overview of all observationally inferred properties of groj1655 - 40 relevant to this investigation is presented in table [ 1655param ] . in particular , these properties are the distance @xmath5 from the sun , the galactic longitude @xmath7 and the galactic latitude @xmath9 , the center - of - mass velocity components @xmath11 , @xmath13 , and @xmath15 with respect to a galactic frame of reference , the orbital period @xmath17 , the bh mass @xmath20 , the donor mass @xmath23 , the donor luminosity @xmath26 , and the donor effective temperature @xmath29 . according to our current understanding , the formation of a bh xrb with a low- to intermediate - mass donor star requires a primordial binary with an extreme mass ratio and a primary ( the bh progenitor ) mass in excess of @xmath46 . if the period of the primordial binary is in the range from @xmath47 to @xmath48yr , the primary is expected to become larger than its critical roche lobe and , due to the extreme mass ratio , lose most of its hydrogen - rich envelope in a dynamically unstable common - envelope phase . provided that enough energy is available to completely expel the envelope and avoid a merger , the common - envelope phase results in the formation of a tight binary consisting of a helium star ( the core of the roche - lobe filling primary ) and a relatively unevolved low- to intermediate - mass main - sequence ( ms ) companion . a bh xrb is then formed when the helium star collapses into a bh and stellar evolution and/or orbital angular momentum losses ( or exchange through tides ) cause the secondary in its turn to fill its roche lobe and transfer mass to the bh . this formation channel is analogous to that usually considered for the formation of ns xrbs and has been considered previously by , e.g. , romani ( 1996 ) , portegies zwart , verbunt , & ergma ( 1997 ) , ergma & van den heuvel ( 1998 ) , kalogera ( 1999 ) , and podsiadlowski et al . ( 2003 ) . in this paper , we restrict ourselves to the formation of bh xrbs through the above `` standard '' evolutionary channel and do not consider alternative formation channels involving hierarchical triple star interactions ( eggleton & verbunt 1986 ) or tidal capture or exchange events in dense stellar environments ( clark 1975 , hills 1976 ) . hence , we assume the bh s immediate progenitor to be a helium star of mass @xmath49 orbiting a ms companion of mass @xmath23 . in view of the strong tidal forces operating on the binary during the common - envelope phase of the primary , we furthermore consider the pre - sn orbit to be circular . the masses of the bh and its companion are assumed to remain constant until the onset of the rlo and x - ray phase . our goal in this analysis is to trace back the evolutionary history of known xrbs ( in this paper of the soft x - ray transient groj1655 - 40 ) and to constrain the properties of the system s progenitor just before and right after the sn explosion that formed the bh . the method adopted to derive the pre- and post - sn constraints incorporates the following set of calculations . we first use a stellar evolution code and calculate a grid of evolutionary sequences for binaries in which a bh is accreting mass from a roche - lobe filling companion . to consider the full range of possibilities , we include sequences for both conservative ( for sub - eddington rates ) and fully non - conservative mt . for each sequence , we examine whether at any point in time the calculated binary properties , i.e. , bh and donor masses , donor effective temperature and luminosity , and orbital period , are in agreement with the observational measurements or derivations of these quantities within their associated uncertainties . among these quantities , the orbital period is measured with the highest accuracy and hence presents the most stringent constraint to be satisfied . we consider a large number of mt sequences and , although many of them can satisfy _ some _ of the constraints , the majority of sequences clearly fail to simultaneously satisfy _ all _ of the constraints at any given time . in the case of groj1655 - 40 the successful sequences furthermore have to satisfy the additional requirement that the system exhibits transient behavior . for the time interval during which all other constraints are satisfied , the long - term mt rate from the donor star must therefore be lower than than the critical rate separating transient from persistent behavior ( dubus et al . 1999 ; see also king , kolb , & burderi 1996 ; king , kolb , & szuszkiewicz 1997 ) . this last requirement restricts the successful sequences even further . with the remaining fully successful sequences , we derive the properties of the binary at the onset of the rlo phase : initial bh and donor masses , orbital period , and age of the donor star . the time at which the fully successful sequences satisfy all observational constraints furthermore provides an estimate for the donor s current age . next , we consider the kinematic evolutionary history of the xrb in the galactic potential . in particular , we use the current position and the measured 3d velocity with their associated uncertainties to trace the galactic motion back in time . combined with the tight constraints on the current age of the system given by the successful mt sequences this allows us to determine the position and velocity of the binary at the time of bh formation ( we denote these as the `` birth '' location and velocity ) . by subtracting the local galactic rotational velocity at this position from the system s total center - of - mass velocity , we then obtain an estimate for the _ peculiar _ velocity of the binary right after the formation of the bh . the `` birth '' or post - sn peculiar velocity holds information about the mass loss and possibly the natal kick associated with the bh s formation . in order to extract this information we must however also constrain the orbital period and orbital eccentricity right after the formation of the bh . we derive these constraints in the third step of our analysis : we consider pairs of post - sn orbital periods and eccentricities and integrate the equations governing the evolution of the orbit under the influence of tides and general relativity ( possibly important for the most highly eccentric orbits ) forward in time . by using the age of the donor star at the onset of rlo for each of the successful mt sequences as an estimate for the time expired since the formation of the bh , we are able to map the post - sn orbital parameters to those at the onset of rlo . comparison with the binary properties at the onset of rlo given by the sequences then allows us to select only those pairs of post - sn orbital period and eccentricity that can match these properties at the right time ( i.e. at the right age or evolutionary stage of the donor star ) . in the fourth step of our analysis we use the derived post - sn masses , orbital period , eccentricity , and peculiar velocity and examine the orbital dynamics of the compact object formation allowing for a natal kick . based on angular momentum and energy conservation we derive constraints on the pre - sn binary properties ( bh progenitor mass and orbital separation ) and the natal kick ( magnitude and direction ) that may have been imparted to the bh . in our final step , we use the initial bh masses from the successful mt sequences and the matching range of bh progenitor masses to study constraints on the core collapse mechanism . by modeling the collapse of bh progenitors and the resultant sn explosion , we can calculate the final range of sn explosion energies and nucleosynthetic yields for groj1655 - 40 . assuming the abundance enhancements in the bh companion are due to the sn explosion that formed the bh , we find strong evidence for the need for asymmetries in sn explosions associated with this bh formation event . in turn , with current stellar evolution and sn models , we can also further constrain the bh progenitor mass , and ultimately the pre - sn binary properties and the sn kick . these five steps are described in more detail in the following sections . we calculate a set of evolutionary sequences for mass - transferring binaries using an up - to - date stellar evolution code described in detail in podsiadlowski et al . ( 2002a ) , ivanova et al . ( 2003 ) , and kalogera et al . binary orbits are assumed to be circular and effects of stellar rotation are neglected at all times . mass - transfer rates are calculated self - consistently by imposing that the stellar radius remains close to the roche - lobe radius during the entire mt phase . in order to determine whether or not a system is transient , we use the criterion @xmath50 where @xmath51 denotes the mt rate , and @xmath52 the critical rate separating transient from persistent behavior ( dubus et al . 1999 , kalogera et al . mass accretion onto the bh is furthermore limited to the eddington rate at all times and any excess mass leaving the system is assumed to carry away the specific orbital angular momentum of the accretor . mass - loss via stellar winds is taken into account using the rates given in hurley , pols , & tout ( 2000 ) . for the mixing - length and convective - overshooting parameters , finally , we adopt values of 0.2 and 0.25 pressure scale heights , respectively . we furthermore note that all mt sequences considered have initial donor masses greater than @xmath53 so that no magnetic braking takes place in the donor star and the binary . for those sequences where mt causes the donor mass to decrease below this value by just one or two tenths of a solar mass ( relevant for bp s current donor mass estimates ) , magnetic braking is still not included , given the relatively wide current orbit of the system . our goal with these calculations of mt sequences is to map out the 3-dimensional parameter space of bh and donor masses , and binary orbital period _ at the onset of rlo _ , for which _ all six _ observational constraints ( _ current _ masses , orbital period , donor luminosity and effective temperature , and transient behavior ) are satisfied _ simultaneously_. since each mt calculation takes a significant amount of computational time ( from a few hours to a few days on one processor ) and often requires user intervention , it is currently not feasible to develop an automated and comprehensive scheme for the systematic and exhaustive search of the entire available 3-dimensional space . instead we therefore use a trial - and - error scheme guided strongly by the systematic behavior of the mt sequences in _ both _ the h - r diagram and the mass - period plane . we first examine the h - r tracks and adjust the rlo parameters so that tracks cross the h - r error box . next , we identify the time intervals for which the h - r constraints are satisfied and examine whether these parts of the tracks also cross the bh and donor mass errors bars _ and _ the orbital period measurement , the latter being the most stringent of all constraints . last we examine whether the sequence exhibits transient behavior for a significant fraction of the time interval over which the other constraints are satisfied . in this way we are able to outline the extent of the parameter space that leads to `` successful '' mt sequences . it turns out that this extent is small enough so that the variations among the successful sequences do not significantly affect our final conclusions for constraints on the bh xrb progenitors ( see section [ prog ] ) , as long as we stick to a given set of observational constraints ( gbo or bp ) . we therefore conclude that it is not necessary to strictly find _ all possible _ successful sequences , but instead to map the outline of their initial parameter space within about @xmath54m@xmath55 in the masses and about @xmath56d in the orbital period . in order to capture the uncertainties related to the accretion of matter by the bh , we furthermore consider both conservative and non - conservative mt sequences for each of the two sets of observational constraints available for groj1655 - 40 . we note that in the conservative case , mass accretion onto the bh is still limited to the eddington rate and any mass transferred in excess of this rate is assumed to be lost from the system carrying the specific orbital angular momentum of the accretor . before describing the sequences that are able to satisfy the observational constraints for groj1655 - 40 , we briefly discuss the systematic behavior of the sequences and their dependency on the initial component masses and orbital period . the systematic behavior is illustrated in fig.[tracks ] for some selected evolutionary tracks which were chosen mainly to provide a clear and instructive picture . the displayed sequences therefore do _ not _ represent our best possible matches to the observed properties of groj1655 - 40 . for the latter , we refer to tables [ gboseq ] and [ bpseq ] where we list the main properties of some selected mt sequences and where the successful sequences are indicated in boldface . since we did not find any successful sequences with post - ms donor stars , we furthermore restricted fig . [ tracks ] to systems with donors that are still on the ms . the initial component masses and orbital period of the sequences at the onset of mt are indicated in the top panel of the figure . sequences 15 illustrate the behavior for conservative mt , while sequence 6 illustrates the behavior for non - conservative mt . for convenience , the parts of the sequences where the donor star satisfies the gbo or bp constraints in the h - r diagram are emphasized by thick solid lines . lcccccccccccccccc + + gbo & 5.5 & 2.5 & 1.5 & 0.238 & 498 & & 6.1 & 1.8 & 1.39 & 3.78 & 0.054 & 643 & & - & - & - + gbo & 5.5 & 2.5 & 1.7 & 0.192 & 526 & & 6.0 & 2.0 & 1.47 & 3.80 & 0.093 & 593 & & - & - & - + * gbo & * 5.5 & * 2.5 & * 1.8 & * 0.172 & * 537 & & * 6.0 & * 2.0 & * 1.51 & * 3.81 & * 0.100 & * 582 & & * 53110 & * 5.59.5 & * 40150 + * gbo & * 5.5 & * 2.5 & * 1.9 & * 0.152 & * 547 & & * 5.9 & * 2.1 & * 1.55 & * 3.81 & * 0.100 & * 577 & & * 53110 & * 5.59.5 & * 40150 + * gbo & * 5.5 & * 2.5 & * 2.0 & * 0.134 & * 556 & & * 5.9 & * 2.1 & * 1.59 & * 3.82 & * 0.097 & * 574 & & * 53110 & * 5.59.5 & * 40150 + gbo & 5.5 & 2.5 & 2.1 & 0.118 & 564 & & 5.8 & 2.2 & 1.62 & 3.83 & 0.088 & 575 & & - & - & - + gbo & 5.5 & 3.0 & 1.2 & 0.358 & 252 & & 6.6 & 1.9 & 1.53 & 3.82 & 0.181 & 362 & & - & - & - + * gbo & * 5.5 & * 3.0 & * 1.3 & * 0.314 & * 303 & & * 6.5 & * 2.0 & * 1.59 & * 3.83 & * 0.179 & * 392 & & * 51108 & * 5.59.5 & * 40160 + gbo & 5.5 & 3.0 & 1.4 & 0.284 & 318 & & 6.4 & 2.1 & 1.64 & 3.84 & 0.174 & 385 & & - & - & - + gbo & 5.5 & 3.0 & 1.5 & 0.257 & 331 & & 6.3 & 2.2 & 1.69 & 3.85 & 0.169 & 380 & & - & - & - + gbo & 5.5 & 3.0 & 1.6 & 0.246 & 301 & & 6.3 & 2.2 & 1.73 & 3.85 & 0.165 & 333 & & - & - & - + gbo & 5.5 & 3.0 & 1.7 & 0.223 & 310 & & 6.2 & 2.3 & 1.78 & 3.86 & 0.157 & 333 & & - & - & - + gbo & 5.5 & 3.0 & 1.8 & 0.203 & 317 & & 6.1 & 2.4 & 1.83 & 3.87 & 0.147 & 334 & & - & - & - + gbo & 5.5 & 3.0 & 1.9 & 0.185 & 323 & & 5.9 & 2.5 & 1.88 & 3.88 & 0.135 & 336 & & - & - & - + gbo & 5.5 & 3.0 & 2.0 & 0.166 & 329 & & 5.8 & 2.5 & 1.92 & 3.89 & 0.124 & 339 & & - & - & - + gbo & 5.5 & 3.5 & 0.8 & 0.568 & 90 & & 7.2 & 1.8 & 1.49 & 3.81 & 0.182 & 396 & & - & - & - + gbo & 5.5 & 3.5 & 1.0 & 0.456 & 148 & & 7.0 & 2.0 & 1.64 & 3.84 & 0.206 & 302 & & - & - & - + gbo & 5.5 & 3.5 & 1.1 & 0.422 & 146 & & 6.9 & 2.1 & 1.71 & 3.85 & 0.209 & 246 & & - & - & - + gbo & 5.5 & 3.5 & 1.2 & 0.385 & 160 & & 6.8 & 2.2 & 1.77 & 3.86 & 0.207 & 234 & & - & - & - + gbo & 5.5 & 3.5 & 1.3 & 0.362 & 170 & & 6.7 & 2.3 & 1.82 & 3.88 & 0.203 & 225 & & - & - & - + gbo & 5.5 & 3.5 & 1.4 & 0.320 & 180 & & 6.6 & 2.3 & 1.88 & 3.89 & 0.198 & 221 & & - & - & - + gbo & 5.5 & 3.5 & 1.5 & 0.292 & 188 & & 6.4 & 2.4 & 1.94 & 3.90 & 0.190 & 218 & & - & - & - + gbo & 5.5 & 3.5 & 1.6 & 0.268 & 195 & & 6.2 & 2.6 & 2.00 & 3.91 & 0.180 & 217 & & - & - & - + gbo & 6.0 & 2.5 & 1.7 & 0.193 & 525 & & 6.5 & 2.0 & 1.48 & 3.80 & 0.098 & 588 & & - & - & - + * gbo & * 6.0 & * 2.5 & * 1.8 & * 0.243 & * 526 & & * 6.5 & * 2.0 & * 1.52 & * 3.81 & * 0.103 & * 569 & & * 52105 & * 6.010.0 & * 40150 + * gbo [ rtw ] & * 6.0 & * 2.5 & * 1.9 & * 0.153 & * 546 & & * 6.4 & * 2.1 & * 1.56 & * 3.82 & * 0.101 & * 576 & & * 52110 & * 6.010.5 & * 40150 + * gbo & * 6.0 & * 2.5 & * 2.0 & * 0.127 & * 555 & & * 6.3 & * 2.1 & * 1.60 & * 3.82 & * 0.092 & * 574 & & * 53110 & * 6.010.5 & * 40150 + gbo & 6.0 & 2.5 & 2.1 & 0.109 & 564 & & 6.3 & 2.2 & 1.63 & 3.83 & 0.085 & 576 & & - & - & - + gbo & 6.0 & 2.5 & 2.2 & 0.103 & 571 & & 6.2 & 2.3 & 1.67 & 3.84 & 0.076 & 580 & & - & - & - + gbo & 6.0 & 3.0 & 1.2 & 0.350 & 284 & & 7.0 & 2.0 & 1.56 & 3.82 & 0.175 & 407 & & - & - & - + gbo & 6.0 & 3.0 & 1.3 & 0.327 & 267 & & 7.0 & 2.0 & 1.61 & 3.83 & 0.180 & 345 & & - & - & - + gbo & 6.0 & 3.0 & 1.4 & 0.287 & 280 & & 6.9 & 2.1 & 1.66 & 3.84 & 0.173 & 339 & & - & - & - + gbo & 6.0 & 3.0 & 1.5 & 0.258 & 330 & & 6.8 & 2.2 & 1.71 & 3.85 & 0.165 & 381 & & - & - & - + gbo & 6.0 & 3.0 & 1.6 & 0.247 & 301 & & 6.7 & 2.3 & 1.75 & 3.86 & 0.161 & 335 & & - & - & - + gbo & 6.0 & 3.0 & 1.7 & 0.213 & 309 & & 6.7 & 2.3 & 1.80 & 3.87 & 0.152 & 334 & & - & - & - + gbo & 6.0 & 3.0 & 1.8 & 0.190 & 358 & & 6.6 & 2.4 & 1.84 & 3.88 & 0.143 & 379 & & - & - & - + gbo & 6.0 & 3.0 & 1.9 & 0.172 & 322 & & 6.5 & 2.5 & 1.88 & 3.88 & 0.134 & 336 & & - & - & - + gbo & 6.0 & 3.0 & 2.0 & 0.155 & 328 & & 6.4 & 2.6 & 1.93 & 3.89 & 0.123 & 338 & & - & - & - + + + + gbo & 6.3 & 2.2 & 1.4 & 0.244 & 712 & & 6.3 & 1.7 & 1.27 & 3.76 & 0.049 & 938 & & - & - & - + gbo & 6.3 & 2.2 & 1.5 & 0.257 & 727 & & 6.3 & 1.8 & 1.31 & 3.77 & 0.044 & 913 & & - & - & - + gbo & 6.3 & 2.2 & 1.6 & 0.193 & 756 & & 6.3 & 1.8 & 1.35 & 3.77 & 0.033 & 914 & & - & - & - + gbo & 6.3 & 2.2 & 1.7 & 0.171 & 774 & & 6.3 & 1.8 & 1.38 & 3.78 & 0.053 & 905 & & - & - & - + gbo & 6.3 & 2.2 & 1.8 & 0.201 & 781 & & 6.3 & 1.9 & 1.41 & 3.79 & 0.052 & 888 & & - & - & - + gbo & 6.3 & 2.2 & 1.9 & 0.131 & 803 & & 6.3 & 1.9 & 1.45 & 3.79 & 0.004 & 893 & & - & - & - + gbo & 6.3 & 2.2 & 2.0 & 0.117 & 803 & & 6.3 & 2.0 & 1.47 & 3.80 & 0.062 & 875 & & - & - & - + gbo & 6.3 & 2.2 & 2.1 & 0.142 & 818 & & 6.3 & 2.0 & 1.58 & 3.83 & 0.008 & 1201 & & - & - & - + gbo & 6.3 & 2.2 & 2.2 & 0.135 & 831 & & 6.3 & 2.0 & 1.64 & 3.84 & 0.008 & 1324 & & - & - & - + gbo & 6.3 & 2.6 & 1.0 & 0.455 & 319 & & 6.3 & 1.8 & 1.34 & 3.77 & 0.024 & 662 & & - & - & - + gbo & 6.3 & 2.6 & 1.2 & 0.336 & 386 & & 6.3 & 1.9 & 1.43 & 3.79 & 0.088 & 578 & & - & - & - + gbo & 6.3 & 2.6 & 1.3 & 0.302 & 409 & & 6.3 & 1.9 & 1.47 & 3.80 & 0.111 & 546 & & - & - & - + * gbo & * 6.3 & * 2.6 & * 1.4 & * 0.331 & * 417 & & * 6.3 & * 2.0 & * 1.51 & * 3.81 & * 0.121 & * 520 & & * 52114 & * 6.510.5 & * 40160 + * gbo & * 6.3 & * 2.6 & * 1.5 & * 0.641 & * 495 & & * 6.3 & * 2.1 & * 1.56 & * 3.82 & * 0.152 & * 583 & & * 53110 & * 6.510.5 & * 40150 + * gbo & * 6.3 & * 2.6 & * 1.6 & * 0.292 & * 446 & & * 6.3 & * 2.1 & * 1.59 & * 3.82 & * 0.124 & * 503 & & * 53111 & * 6.511.0 & * 40150 + gbo & 6.3 & 2.6 & 1.7 & 0.202 & 468 & & 6.3 & 2.2 & 1.62 & 3.83 & 0.114 & 5.13 & & - & - & - + gbo & 6.3 & 2.6 & 1.8 & 0.169 & 540 & & 6.3 & 2.3 & 1.68 & 3.84 & 0.095 & 5.82 & & - & - & - + gbo & 6.3 & 3.0 & 0.9 & 0.500 & 175 & & 6.3 & 1.9 & 1.51 & 3.81 & 0.166 & 4.06 & & - & - & - + * gbo & * 6.3 & * 3.0 & * 1.0 & * 0.433 & * 234 & & * 6.3 & * 2.0 & * 1.57 & * 3.82 & * 0.166 & * 429 & & * 53109 & * 6.510.5 & * 30160 + gbo & 6.3 & 3.0 & 1.1 & 0.406 & 231 & & 6.3 & 2.1 & 1.63 & 3.83 & 0.170 & 360 & & - & - & - + gbo & 6.3 & 3.0 & 1.2 & 0.350 & 284 & & 6.3 & 2.1 & 1.68 & 3.84 & 0.161 & 400 & & - & - & - + gbo & 6.3 & 3.0 & 1.3 & 0.603 & 301 & & 6.3 & 2.2 & 1.73 & 3.85 & 0.186 & 392 & & - & - & - + gbo & 6.3 & 3.0 & 1.4 & 0.306 & 279 & & 6.3 & 2.3 & 1.77 & 3.86 & 0.154 & 341 & & - & - & - + gbo & 6.3 & 3.0 & 1.5 & 0.271 & 291 & & 6.3 & 2.4 & 1.81 & 3.87 & 0.148 & 341 & & - & - & - + gbo & 6.3 & 3.3 & 0.7 & 0.637 & 52 & & 6.3 & 1.9 & 1.51 & 3.81 & 0.171 & 383 & & - & - & - + gbo & 6.3 & 3.3 & 0.8 & 0.569 & 96 & & 6.3 & 2.0 & 1.60 & 3.83 & 0.186 & 332 & & - & - & - + gbo & 6.3 & 3.3 & 0.9 & 0.501 & 152 & & 6.3 & 2.2 & 1.73 & 3.86 & 0.171 & 354 & & - & - & - + gbo & 6.3 & 3.3 & 1.0 & 0.459 & 154 & & 6.3 & 2.2 & 1.73 & 3.86 & 0.185 & 289 & & - & - & - + gbo & 6.3 & 3.3 & 1.1 & 0.416 & 173 & & 6.3 & 2.3 & 1.79 & 3.87 & 0.181 & 278 & & - & - & - + * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * lcccccccccccccccc + + bp & 3.5 & 3.0 & 0.6 & 0.682 & 18 & & 5.3 & 1.2 & 1.04 & 3.73 & 0.090 & 1078 & & - & - & - + bp & 3.5 & 3.0 & 0.7 & 0.599 & 94 & & 5.2 & 1.3 & 1.13 & 3.74 & 0.052 & 902 & & - & - & - + bp & 3.5 & 3.0 & 0.8 & 0.541 & 147 & & 5.1 & 1.4 & 1.19 & 3.75 & 0.036 & 782 & & - & - & - + bp & 3.5 & 3.0 & 0.9 & 0.485 & 185 & & 5.0 & 1.5 & 1.26 & 3.77 & 0.003 & 700 & & - & - & - + bp & 3.5 & 3.0 & 1.0 & 0.429 & 215 & & 4.9 & 1.5 & 1.26 & 3.76 & 0.031 & 619 & & - & - & - + bp & 3.5 & 3.5 & 1.1 & 0.409 & 151 & & 4.9 & 1.8 & 1.49 & 3.81 & 0.216 & 278 & & - & - & - + bp & 3.5 & 3.0 & 1.2 & 0.348 & 257 & & 4.7 & 1.7 & 1.36 & 3.78 & 0.130 & 446 & & - & - & - + bp & 3.5 & 3.5 & 0.6 & 0.698 & 1 & & 5.6 & 1.3 & 1.15 & 3.75 & 0.056 & 885 & & - & - & - + * bp & * 3.5 & * 3.5 & * 0.7 & * 0.627 & * 51 & & * 5.5 & * 1.4 & * 1.28 & * 3.77 & * 0.004 & * 730 & & * 47112 & * 4.06.0 & * 40200 + * bp & * 3.5 & * 3.5 & * 0.8 & * 0.569 & * 87 & & * 5.4 & * 1.5 & * 1.28 & * 3.77 & * 0.059 & * 580 & & * 52110 & * 3.56.0 & * 30200 + * bp & * 3.5 & * 3.5 & * 0.9 & * 0.511 & * 114 & & * 5.3 & * 1.6 & * 1.35 & * 3.78 & * 0.128 & * 436 & & * 49109 & * 3.56.0 & * 0210 + * bp & * 3.5 & * 3.5 & * 1.0 & * 0.463 & * 134 & & * 5.1 & * 1.7 & * 1.42 & * 3.80 & * 0.184 & * 335 & & * 52111 & * 3.56.5 & * 0210 + bp & 3.5 & 3.5 & 1.1 & 0.409 & 151 & & 4.9 & 1.8 & 1.49 & 3.81 & 0.216 & 278 & & - & - & - + bp & 3.5 & 3.5 & 1.2 & 0.371 & 164 & & 4.7 & 1.9 & 1.56 & 3.82 & 0.229 & 241 & & - & - & - + bp & 3.5 & 4.0 & 0.6 & 0.699 & 1 & & 5.8 & 1.4 & 1.27 & 3.76 & 0.052 & 764 & & - & - & - + bp & 3.5 & 4.0 & 0.7 & 0.639 & 28 & & 5.8 & 1.5 & 1.30 & 3.77 & 0.049 & 580 & & - & - & - + * bp & * 3.5 & * 4.0 & * 0.8 & * 0.583 & * 55 & & * 5.5 & * 1.6 & * 1.41 & * 3.80 & * 0.157 & * 386 & & * 55106 & * 4.06.0 & * 30210 + bp & 3.5 & 4.0 & 0.9 & 0.515 & 74 & & 5.4 & 1.7 & 1.49 & 3.81 & 0.199 & 290 & & - & - & - + bp & 3.5 & 4.0 & 1.0 & 0.461 & 91 & & 5.2 & 1.8 & 1.58 & 3.83 & 0.223 & 234 & & - & - & - + bp & 3.5 & 4.0 & 1.1 & 0.419 & 102 & & 4.9 & 1.9 & 1.66 & 3.85 & 0.236 & 196 & & - & - & - + bp & 3.5 & 4.5 & 0.7 & 0.654 & 16 & & 5.9 & 1.6 & 1.38 & 3.79 & 0.105 & 461 & & - & - & - + bp & 3.5 & 4.5 & 0.8 & 0.588 & 37 & & 5.7 & 1.7 & 1.50 & 3.82 & 0.171 & 317 & & - & - & - + bp & 3.5 & 4.5 & 0.9 & 0.529 & 53 & & 5.5 & 1.8 & 1.61 & 3.84 & 0.199 & 245 & & - & - & - + bp & 3.5 & 4.5 & 1.0 & 0.476 & 65 & & 5.2 & 1.9 & 1.70 & 3.86 & 0.218 & 196 & & - & - & - + bp & 4.0 & 2.5 & 0.6 & 0.661 & 64 & & 5.4 & 1.1 & 0.97 & 3.71 & 0.121 & 1.320 & & - & - & - + bp & 4.0 & 2.5 & 0.7 & 0.575 & 186 & & 5.3 & 1.2 & 1.02 & 3.72 & 0.100 & 1.133 & & - & - & - + bp & 4.0 & 2.5 & 0.8 & 0.504 & 268 & & 5.2 & 1.3 & 1.07 & 3.73 & 0.071 & 1.004 & & - & - & - + bp & 4.0 & 2.5 & 0.9 & 0.486 & 323 & & 5.1 & 1.4 & 1.12 & 3.74 & 0.048 & 908 & & - & - & - + bp & 4.0 & 2.5 & 1.0 & 0.441 & 367 & & 5.0 & 1.4 & 1.15 & 3.74 & 0.038 & 844 & & - & - & - + bp & 4.0 & 2.5 & 1.1 & 0.358 & 411 & & 5.0 & 1.5 & 1.19 & 3.75 & 0.000 & 806 & & - & - & - + bp & 4.0 & 2.5 & 1.2 & 0.317 & 437 & & 4.9 & 1.5 & 1.24 & 3.76 & 0.046 & 768 & & - & - & - + bp & 4.0 & 2.5 & 1.3 & 0.288 & 464 & & 4.9 & 1.6 & 1.32 & 3.77 & 0.004 & 739 & & - & - & - + bp & 4.0 & 3.0 & 0.6 & 0.688 & 12 & & 5.7 & 1.3 & 1.09 & 3.73 & 0.063 & 989 & & - & - & - + bp & 4.0 & 3.0 & 0.7 & 0.604 & 90 & & 5.6 & 1.4 & 1.17 & 3.75 & 0.049 & 836 & & - & - & - + * bp [ bpeg ] & * 4.0 & * 3.0 & * 0.8 & * 0.535 & * 144 & & * 5.5 & * 1.5 & * 1.27 & * 3.77 & * 0.001 & * 733 & & * 51112 & * 4.07.0 & * 30190 + * bp & * 4.0 & * 3.0 & * 0.9 & * 0.489 & * 182 & & * 5.5 & * 1.5 & * 1.26 & * 3.76 & * 0.016 & * 648 & & * 49115 & * 4.07.0 & * 0200 + * bp & * 4.0 & * 3.0 & * 1.0 & * 0.450 & * 244 & & * 5.4 & * 1.6 & * 1.31 & * 3.77 & * 0.091 & * 618 & & * 48115 & * 4.07.0 & * 0200 + * bp & * 4.0 & * 3.0 & * 1.1 & * 0.398 & * 235 & & * 5.3 & * 1.7 & * 1.36 & * 3.78 & * 0.117 & * 464 & & * 51110 & * 4.07.0 & * 20190 + * bp & * 4.0 & * 3.0 & * 1.2 & * 0.351 & * 255 & & * 5.2 & * 1.8 & * 1.42 & * 3.79 & * 0.156 & * 407 & & * 55113 & * 4.07.5 & * 30200 + bp & 4.0 & 3.0 & 1.3 & 0.309 & 270 & & 5.1 & 1.8 & 1.47 & 3.80 & 0.176 & 365 & & - & - & - + bp & 4.0 & 3.0 & 1.4 & 0.279 & 283 & & 5.0 & 1.9 & 1.53 & 3.81 & 0.185 & 343 & & - & - & - + bp & 4.0 & 3.0 & 1.5 & 0.251 & 333 & & 4.9 & 2.0 & 1.59 & 3.82 & 0.185 & 375 & & - & - & - + bp & 4.0 & 3.5 & 0.6 & 0.699 & 1 & & 6.0 & 1.4 & 1.24 & 3.75 & 0.058 & 791 & & - & - & - + bp & 4.0 & 3.5 & 0.7 & 0.626 & 47 & & 6.0 & 1.5 & 1.27 & 3.77 & 0.016 & 651 & & - & - & - + bp & 4.0 & 3.5 & 0.8 & 0.566 & 85 & & 5.9 & 1.6 & 1.34 & 3.78 & 0.111 & 480 & & - & - & - + bp & 4.0 & 3.5 & 0.9 & 0.499 & 111 & & 5.8 & 1.7 & 1.42 & 3.80 & 0.170 & 366 & & - & - & - + bp & 4.0 & 3.5 & 1.0 & 0.468 & 132 & & 5.6 & 1.8 & 1.49 & 3.81 & 0.208 & 298 & & - & - & - + bp & 4.0 & 3.5 & 1.1 & 0.401 & 149 & & 5.5 & 1.9 & 1.56 & 3.82 & 0.218 & 260 & & - & - & - + bp & 4.0 & 3.5 & 1.2 & 0.375 & 163 & & 5.3 & 2.0 & 1.63 & 3.84 & 0.223 & 238 & & - & - & - + bp & 4.0 & 3.5 & 1.3 & 0.341 & 174 & & 5.1 & 2.1 & 1.70 & 3.85 & 0.220 & 224 & & - & - & - + bp & 4.5 & 2.0 & 1.2 & 0.280 & 880 & & 5.2 & 1.3 & 1.03 & 3.72 & 0.094 & 1271 & & - & - & - + bp & 4.5 & 2.0 & 1.3 & 0.245 & 915 & & 5.1 & 1.4 & 1.05 & 3.72 & 0.097 & 1277 & & - & - & - + bp & 4.5 & 2.0 & 1.4 & 0.217 & 959 & & 5.1 & 1.4 & 1.06 & 3.72 & 0.091 & 1219 & & - & - & - + bp & 4.5 & 2.0 & 1.5 & 0.224 & 982 & & 5.0 & 1.5 & 1.08 & 3.72 & 0.088 & 1194 & & - & - & - + bp & 4.5 & 2.0 & 1.6 & 0.164 & 1016 & & 4.9 & 1.5 & 1.10 & 3.72 & 0.085 & 1188 & & - & - & - + bp & 4.5 & 2.0 & 1.7 & 0.138 & 1.032 & & 4.8 & 1.6 & 1.33 & 3.76 & 0.062 & 1193 & & - & - & - + bp & 4.5 & 2.0 & 1.8 & 0.114 & 1.054 & & 4.8 & 1.6 & 1.14 & 3.73 & 0.083 & 1752 & & - & - & - + bp & 4.5 & 2.0 & 1.9 & 0.091 & 1.075 & & 4.8 & 1.7 & 1.17 & 3.73 & 0.078 & 2376 & & - & - & - + bp & 4.5 & 2.5 & 1.1 & 0.361 & 409 & & 5.4 & 1.5 & 1.22 & 3.75 & 0.000 & 790 & & - & - & - + bp & 4.5 & 2.5 & 1.2 & 0.323 & 438 & & 5.4 & 1.6 & 1.32 & 3.77 & 0.001 & 755 & & - & - & - + * bp & * 4.5 & * 2.5 & * 1.3 & * 0.341 & * 456 & & * 5.3 & * 1.7 & * 1.32 & * 3.77 & * 0.003 & * 715 & & * 47115 & * 4.57.5 & * 40170 + * bp [ example1 ] & * 4.5 & * 2.5 & * 1.4 & * 0.315 & * 476 & & * 5.3 & * 1.7 & * 1.33 & * 3.77 & * 0.009 & * 689 & & * 52110 & * 4.58.0 & * 40170 + * bp & * 4.5 & * 2.5 & * 1.5 & * 0.295 & * 491 & & * 5.2 & * 1.8 & * 1.34 & * 3.77 & * 0.029 & * 657 & & * 50114 & * 4.58.0 & * 40170 + bp & 4.5 & 2.5 & 1.6 & 0.273 & 506 & & 5.1 & 1.8 & 1.39 & 3.78 & 0.053 & 623 & & - & - & - + bp & 4.5 & 3.0 & 0.8 & 0.544 & 141 & & 6.0 & 1.5 & 1.27 & 3.77 & 0.008 & 695 & & - & - & - + bp & 4.5 & 3.0 & 0.9 & 0.479 & 180 & & 5.9 & 1.6 & 1.29 & 3.77 & 0.046 & 593 & & - & - & - + bp & 4.5 & 3.0 & 1.0 & 0.436 & 211 & & 5.8 & 1.7 & 1.35 & 3.78 & 0.106 & 496 & & - & - & - + bp & 4.5 & 3.0 & 1.1 & 0.392 & 235 & & 5.7 & 1.8 & 1.41 & 3.79 & 0.145 & 429 & & - & - & - + bp & 4.5 & 3.0 & 1.2 & 0.354 & 254 & & 5.7 & 1.8 & 1.46 & 3.80 & 0.175 & 380 & & - & - & - + bp & 4.5 & 3.0 & 1.3 & 0.322 & 269 & & 5.6 & 1.9 & 1.52 & 3.81 & 0.184 & 356 & & - & - & - + bp & 4.5 & 3.0 & 1.4 & 0.292 & 282 & & 5.5 & 2.0 & 1.57 & 3.82 & 0.188 & 339 & & - & - & - + bp & 4.5 & 3.0 & 1.5 & 0.276 & 292 & & 5.4 & 2.1 & 1.63 & 3.83 & 0.182 & 331 & & - & - & - + + + + bp & 5.4 & 2.0 & 1.0 & 0.361 & 760 & & 5.4 & 1.4 & 1.03 & 3.72 & 0.099 & 1315 & & - & - & - + bp & 5.4 & 2.0 & 1.1 & 0.341 & 821 & & 5.4 & 1.4 & 1.05 & 3.72 & 0.095 & 1267 & & - & - & - + bp & 5.4 & 2.0 & 1.2 & 0.304 & 872 & & 5.4 & 1.5 & 1.07 & 3.72 & 0.092 & 1236 & & - & - & - + bp & 5.4 & 2.0 & 1.3 & 0.275 & 915 & & 5.4 & 1.5 & 1.09 & 3.72 & 0.086 & 1215 & & - & - & - + bp & 5.4 & 2.0 & 1.4 & 0.213 & 1079 & & 5.4 & 1.6 & 1.12 & 3.72 & 0.081 & 1357 & & - & - & - + bp & 5.4 & 2.0 & 1.5 & 0.208 & 1138 & & 5.4 & 1.6 & 1.17 & 3.73 & 0.062 & 1411 & & - & - & - + bp & 5.4 & 2.3 & 0.8 & 0.498 & 349 & & 5.4 & 1.4 & 1.10 & 3.73 & 0.066 & 1044 & & - & - & - + bp & 5.4 & 2.3 & 0.9 & 0.440 & 424 & & 5.4 & 1.5 & 1.13 & 3.73 & 0.054 & 978 & & - & - & - + bp & 5.4 & 2.3 & 1.0 & 0.383 & 547 & & 5.4 & 1.6 & 1.25 & 3.76 & 0.000 & 1035 & & - & - & - + * bp & * 5.4 & * 2.3 & * 1.1 & * 0.341 & * 596 & & * 5.4 & * 1.7 & * 1.29 & * 3.76 & * 0.000 & * 998 & & * 46111 & * 5.59.0 & * 40160 + bp & 5.4 & 2.3 & 1.2 & 0.348 & 557 & & 5.4 & 1.7 & 1.25 & 3.76 & 0.091 & 860 & & - & - & - + bp & 5.4 & 2.5 & 0.8 & 0.543 & 257 & & 5.4 & 1.5 & 1.20 & 3.75 & 0.042 & 854 & & - & - & - + * bp & * 5.4 & * 2.5 & * 0.9 & * 0.445 & * 389 & & * 5.4 & * 1.6 & * 1.33 & * 3.77 & * 0.000 & * 924 & & * 54115 & * 5.59.0 & * 40170 + * bp & * 5.4 & * 2.5 & * 1.0 & * 0.396 & * 445 & & * 5.4 & * 1.7 & * 1.40 & * 3.79 & * 0.000 & * 882 & & * 55108 & * 5.58.5 & * 50150 + * bp & * 5.4 & * 2.5 & * 1.1 & * 0.402 & * 399 & & * 5.4 & * 1.7 & * 1.35 & * 3.78 & * 0.004 & * 719 & & * 47115 & * 5.59.0 & * 40160 + * bp & * 5.4 & * 2.5 & * 1.2 & * 0.326 & * 436 & & * 5.4 & * 1.8 & * 1.36 & * 3.78 & * 0.022 & * 696 & & * 53110 & * 5.59.0 & * 40160 + bp & 5.4 & 2.5 & 1.3 & 0.293 & 461 & & 5.4 & 1.9 & 1.39 & 3.78 & 0.045 & 660 & & - & - & - + bp & 5.4 & 2.5 & 1.4 & 0.265 & 468 & & 5.4 & 1.9 & 1.43 & 3.79 & 0.067 & 618 & & - & - & - + bp & 5.4 & 3.0 & 0.6 & 0.699 & 5 & & 5.4 & 1.6 & 1.33 & 3.78 & 0.004 & 841 & & - & - & - + * bp & * 5.4 & * 3.0 & * 0.7 & * 0.612 & * 96 & & * 5.4 & * 1.7 & * 1.33 & * 3.77 & * 0.058 & * 679 & & * 51108 & * 5.58.5 & * 40170 + * bp & * 5.4 & * 3.0 & * 0.8 & * 0.548 & * 138 & & * 5.4 & * 1.8 & * 1.41 & * 3.79 & * 0.120 & * 485 & & * 52106 & * 5.58.5 & * 40160 + bp & 5.4 & 3.0 & 0.9 & 0.490 & 178 & & 5.4 & 1.9 & 1.48 & 3.80 & 0.158 & 417 & & - & - & - + bp & 5.4 & 3.3 & 0.6 & 0.700 & 1 & & 5.4 & 1.7 & 1.37 & 3.78 & 0.089 & 531 & & - & - & - + bp & 5.4 & 3.3 & 0.7 & 0.633 & 55 & & 5.4 & 1.8 & 1.47 & 3.80 & 0.163 & 400 & & - & - & - + bp & 5.4 & 3.3 & 0.8 & 0.566 & 98 & & 5.4 & 1.9 & 1.56 & 3.82 & 0.187 & 340 & & - & - & - + bp & 5.4 & 3.3 & 0.9 & 0.507 & 131 & & 5.4 & 2.0 & 1.63 & 3.84 & 0.190 & 309 & & - & - & - + * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * the shape of the evolutionary tracks in the h - r diagram is similar for all considered sequences : the donor star follows its normal ms evolution until the onset of mt ( indicated by the filled circles ) and then evolves towards lower luminosities and lower effective temperatures as it adjusts to its decreasing mass . the evolutionary stage of the donor at the onset of mt is determined by the the orbital period at the start of the rlo phase . for a given bh and donor mass , longer initial periods yield more evolved donor stars and higher initial mt rates . since tracks with longer initial periods are also associated with larger donor stars , they furthermore tend to be shifted towards higher luminosities ( cf . sequences 2 and 3 ) . a similar tendency occurs when increasing the donor mass for a fixed bh mass and initial orbital period ( cf . sequences 1 , 3 , and 4 ) . the requirement that the sequences must pass through the observationally inferred position of groj1655 - 40 in the h - r diagram therefore gives a rough _ lower and upper limit on the initial period and donor mass _ at the onset of rlo . for a given donor mass and a given initial period , the bh mass and the amount of mass lost from the system during the mt process affect the evolution in the h - r diagram through the rate at which the donor reduces its mass . higher bh masses and non - conservative mt yield lower mt rates and therefore a slower decrease of the donor s luminosity and effective temperature during the mt phase compared to that for lower bh masses and conservative mt ( cf . sequences 4 , 5 , and 6 ) . these effects are less severe though than changes in the initial donor mass and orbital period . the variations of the mt rate expressed in units of the critical rate for transient behavior depend on the relative changes between these two quantities . after the onset of the mt phase , the decrease of the mass ratio and the increase of the orbital period cause the mt rate to decrease and the critical rate to increase , so that overall the ratio @xmath57 decreases with increasing orbital period . this tendency continues until the donor reaches the end of core - hydrogen burning and the associated expansion of the stellar radius causes the mt rate to increase more rapidly than the critical rate for transient behavior . the ratio @xmath57 therefore increases as well . during the subsequent evolution through the hertzsprung gap , no nuclear burning takes place and the mt rate stays roughly constant . since @xmath52 still increases slowly , the ratio @xmath57 slowly decreases . examination of fig.[tracks ] indicates how we can effectively identify _ winner _ sequences . all six sequences shown as an example satisfy the gbo or bp constraints in the h - r diagram , but only two of them ( sequence 2 and 3 ) satisfy the bp mass constraints , and only one of those satisfies the transient constraint ( sequence 3 ) . the mt sequences which are able to satisfy the constraints derived by gbo and bp are highlighted in boldface in tables [ gboseq ] and [ bpseq ] , respectively . in the case of gbo constraints and conservative mt , successful sequences were found for initial bh masses ranging from @xmath58 to @xmath59 , initial donor masses ranging from @xmath60 to @xmath61 , and initial periods ranging from @xmath62 to @xmath63days . the lower limit on the initial bh mass arises from the requirement that , for conservative mt , lower mass bhs need a more massive donor star in order to accrete enough matter to reach the observed present - day bh mass . for orbital periods where rlo occurs when the donor is still close to the zams , the higher mt rates associated with more massive donor stars then cause the donor mass to decrease below the admissible range from 2.0 to2.8@xmath64 before the period reaches the present - day value of 2.62days . for longer orbital periods , on the other hand , the donor star has a somewhat more developed helium core at the onset of rlo , which causes the star to be too luminous when the system reaches the observed period of 2.62days . the upper limit on the initial bh mass is determined by the requisite that as the initial bh mass gets closer to the present value of @xmath21 , the donor star must start rlo closer to its present state as well . this eventually requires mt to start in the hertzsprung gap which generally leads to mt rates that are too high for the system to be transient ( see , however , kolb et al . 1997 and kolb 1998 for a discussion on soft x - ray transient in the hertzsprung gap ) . in the case of non - conservative mt , we kept the bh mass fixed to the mean value of @xmath65 . by keeping the bh mass constant we were able to to outline the initial donor masses at the onset of rlo with a somewhat higher resolution of 0.3 to @xmath66 without having to calculate an unrealistically large number of sequences . successful sequences were found for @xmath67@xmath61 and @xmath68@xmath69days . when the donor star crosses the allowed region in the h - r diagram , all successful mt sequences found for the gbo system parameters are furthermore only marginally consistent with the currently observed orbital period and donor mass . the mt rates at the period of 2.62days are always at least a factor of @xmath70 smaller than the critical mt rate for transient behavior . in the case of bp constraints and conservative mt , successful sequences were found for @xmath71@xmath72 , @xmath73@xmath74 , and @xmath75@xmath76days , while for non - conservative mt successful sequences were found for @xmath77 , @xmath78@xmath61 , and @xmath75@xmath79days . the lower limit on the initial bh mass for conservative mt is again related to the higher donor masses required by initially less massive bhs to reach the present - day mass . the difference with the gbo sequences is , however , that the bh masses at the start of rlo are now so low that this requirement implies that mt initially starts on the thermal time scale of the donor star . this leads to highly super - eddington mt rates and significant mass loss from the system . consequently , initially even more massive donor stars are required for the bh to accrete enough to reach the observed present - day mass . these higher donor masses increase the mt rates even further , rendering the system persistent rather transient . the upper limit on the initial bh mass arises for the same reason as for the gbo sequences . the mt rates obtained when the donor star satisfies the observational constraints in the h - r diagram and when the binary reaches the observed period of 2.62days span a wide range of values : for some sequences the mt rates are more than an order of magnitude smaller than the critical rate for transient behavior , while for others the system is only marginally transient . in general , however , it turned out to be easier to find mt sequences satisfying the bp constraints than sequences satisfying the gbo constraints . an overview of the initial donor star masses and the initial orbital periods of the mt sequences listed in tables [ gboseq][bpseq ] is given in fig . successful sequences are indicated by solid circles and unsuccessful ones by crosses . the figure clearly illustrates the extent of the parameter space covered by our calculations as well as the location of the successful sequences among the unsuccessful ones . the general trend in the initial parameters of the successful mt sequences is that lower initial donor masses require longer initial orbital periods in order to be successful . this is in agreement with the systematic behavior of the mt sequences illustrated in fig . [ tracks ] . we also note that mt sequences were calculated for bh masses outside the successful ranges quoted above , but for brevity these have been omitted from tables [ gboseq][bpseq ] and fig . [ init ] . [ init ] and tables [ gboseq][bpseq ] also show there is no single _ unique _ mt sequence that satisfies all the observational constraints . instead , a series of neighboring successful sequences is found for both gbo and bp constraints and for both conservative and non - conservative mt . in addition , the number of successful sequences found is limited only by our finite exploration of the available parameter space and the limitation of the calculations to the two extreme cases of fully conservative and fully non - conservative mt . adopting a higher resolution grid of initial component masses and orbital periods would increase the number of successful sequences even further as would the consideration of mt sequences for which the degree of mass conservation is in between the two extreme cases of fully conservative or non - conservative mt . fortunately , as we will see in section [ prog ] , mt sequences with closely spaced initial parameters yield only small variations in the properties of the pre - sn binary progenitor and the kick that may have been imparted to the bh at birth . the vital point in the calculation of the presented mt sequences is therefore that we managed to outline the initial parameters of the successful sequences within a pre - determined accuracy rather than finding all possible mt sequences able to satisfy the present - day observational constraints for groj1655 - 40 . given the present position and velocity of groj1655 - 40 in the galaxy , we derive the post - sn peculiar velocity of the binary s mass center by tracing its orbit in the galaxy back to the birth time of the bh . a specific birth time is derived for _ each _ successful mt sequence by setting it equal to the _ current _ age of the donor star . the latter is a good estimate of the bh birth time within an uncertainty equal to the lifetime of the massive progenitor of the bh , which is typically less than 10myr . this uncertainty is always much smaller than the actual time expired since the birth of the bh . we describe the motion of the system with respect to a right - handed cartesian frame of reference @xmath80 whose origin coincides with the galactic center and whose @xmath81-plane coincides with the mid - plane of the galactic disk . the direction from the projection of the sun s position onto the galactic plane to the galactic center is taken as the positive direction of the @xmath82-axis , the direction from the sun to the northern galactic pole as the positive direction of the @xmath83-axis , and the direction of the galactic rotational velocity at the position of the sun as the positive direction of the @xmath84-axis . in terms of these coordinates , the sun is located at @xmath85 with @xmath86kpc , @xmath87 , and @xmath88pc ( reid 1993 ; humphreys & larsen 1995 , and references therein ) . groj1655 - 40 is currently located at a distance from the sun of @xmath6kpc ( hjellming & rupen 1995 ) , a galactic longitude @xmath89 , and a galactic latitude @xmath90 ( tingay et al . 1995 ) ; corresponding to @xmath91kpc , @xmath92kpc , and @xmath93kpc . the velocity components of groj1655 - 40 with respect to the @xmath82- , @xmath84- , and @xmath83-axes are @xmath94 , @xmath95 , and @xmath96 ( mirabel et al . 2002 ) . to model the galaxy , we adopt the galactic potential of carlberg & innanen ( 1987 ) with updated model parameters of kuijken & gilmore ( 1989 ) . for each successful mt sequence ( marked in boldface in tables [ gboseq ] and [ bpseq ] ) , the equations governing the system s motion in the galaxy are integrated backward in time up to the time corresponding to the current age of the donor star , as given by the evolutionary sequence . in fig . [ orbit ] , we show the orbit of groj1655 - 40 projected on the @xmath81- and @xmath97-planes up to 1gyr in the past , which encompasses the entire range of donor ages from @xmath98 to @xmath99myr given by the successful mt sequences . the black line represents the orbit obtained for the mean values of the distance and velocity components , while the grey lines indicate the deviations from this orbit obtained from considering all possible combinations of the extreme values of the distance and velocity component measurements ( i.e. the endpoints of the error bars ) . in agreement with the findings of mirabel et al . ( 2002 ) , the system always remains within 3 to 7kpc from the galactic center and within 200pc from the galactic plane . the post - sn peculiar velocity @xmath100 of the binary is obtained by subtracting the local galactic rotational velocity from the total systemic velocity at the birth site of the bh . the variations of @xmath100 as a function of the time expired since the formation of the bh are displayed in fig . the hatched vertical bars indicate the bh birth times determined by the age of the donor star for each of the successful mt sequences listed in tables [ gboseq][bpseq ] . the width of the bars corresponds to an uncertainty in the birth times of 10myr . when accounting for the uncertainties in the distance and the velocity components ( cf . [ orbit ] ) , the post - sn peculiar velocity can span a wide range of values ranging from @xmath101 to @xmath102 . this wide range is a result of our determination of the possible uncertainties in the galactic orbit which likely overestimates the true error bars on @xmath103 . nevertheless , in order to be conservative , we consider the full range of possible post - sn peculiar velocities given by the grey shaded area in fig . after the formation of the bh , the orbital parameters of the binary are subjected to secular changes due to the perturbation of the external gravitational field from spherical symmetry by the ms star s tidal distortion and due to the loss of orbital angular momentum via gravitational radiation . since the tidal interactions depend on both the orbital and rotational properties of the ms star , the star s rotational angular velocity right after the sn explosion that formed the bh enters the problem as an additional unknown quantity . we here assume that the the rotational angular velocity of the ms star is unaffected by the sn explosion and that just before the explosion the star s rotation rate was synchronized with the orbital motion . the post - sn rotational angular velocity is then given by @xmath104 , where @xmath105 is the pre - sn orbital period . as a test , we also considered the more extreme cases @xmath106 and @xmath107 , but these were found to not affect significantly the outcome of the orbital evolution between the formation of the bh and the start of the rlo phase . we adopt the system of equations governing the tidal evolution of the semi - major axis @xmath108 , the orbital eccentricity @xmath109 , and the ms star s rotational angular velocity @xmath110 from hut ( 1981 ) : @xmath111,\end{aligned}\ ] ] @xmath112,\end{aligned}\ ] ] @xmath113,\end{aligned}\ ] ] where @xmath114 @xmath115 @xmath116 @xmath117 @xmath118 ( see also zahn 1977 , 1978 ) . in these equations , @xmath20 is the mass of the bh ; @xmath23 , @xmath119 , @xmath120 , and @xmath121 are the mass , radius , apsidal - motion constant , and moment of inertia of the companion star , @xmath122 is the mean orbital angular velocity , and @xmath123 is a characteristic time scale for the orbital evolution due to tides . for stars with radiative envelopes , as we are dealing with here , the dominant mechanism contributing to the dissipative tidal forces is radiative damping of dynamical tides ( zahn 1975 , 1977 ) . the factor @xmath124 can then be approximated as @xmath125 with @xmath126 ( hurley , tout , & pols 2002 ) . ( [ kt ] ) and ( [ e2 ] ) are strictly speaking only valid for dynamic tides with low forcing frequencies that are far away from any of the eigenfrequencies of the star s free modes of oscillation . witte & savonije ( 1999 ) have shown that in close binaries with eccentric orbits the system can easily become locked in a resonance between a dynamic tide and a free oscillation mode and that such resonance lockings can greatly accelerate the orbital evolution . in the expression for @xmath124 , we therefore incorporated the calibration factor @xmath127 to account for some of the still existing uncertainties in the strength of tidal dissipation in stars containing radiative envelopes . in our calculations , we adopt @xmath128 which has been shown to be appropriate for consistency with both the observations of circularization periods in open clusters and with orbital decay rates measured in high - mass x - ray binaries ( for more details see belczynski et al . 2004 ) . to follow the secular changes in the orbital parameters associated with the emission of gravitational waves , we adopt the system of equations given by junker & schfer ( 1992 ) : @xmath129 ^ 3 \nonumber \\ & & \times \left ( 96 + 292\,e^2 + 37\,e^4 \right ) \left ( 1 - e^2 \right ) \label{dagr}\end{aligned}\ ] ] @xmath130 ^ 4 \nonumber \\ & & \times { { e \left ( 304 + 121\,e^2 \right ) \left ( 1 - e^2 \right ) } \over { \left ( 1-e^2 \right)^{7/2}}}. \label{degr}\end{aligned}\ ] ] here , @xmath131 is the newtonian gravitational constant , @xmath132 is the speed of light , and @xmath133 . for brevity , we restricted eqs . ( [ dagr ] ) and ( [ degr ] ) to the lowest - order terms in the small parameter @xmath134 . in the numerical calculations however , we adopt the full 3.5 post - newtonian order equations derived by junker & schfer ( 1992 ) . the orbital evolution due to tides and gravitational radiation is illustrated in fig . [ orbev ] for the example case of a binary consisting of a @xmath72 bh and a @xmath135 zero - age ms star , and for four different combinations of initial orbital separations and eccentricities . the four combinations are chosen such that after approximately @xmath136myr of orbital evolution the total angular momentum of the binary is approximately equal to the total angular momentum of the binary represented by mt sequence bp[example1 ] ( see table [ bpseq ] ) at the onset of the mt phase . the choice to match the orbital angular momentum at the end of the orbital evolution calculations to the orbital angular momentum of the mt sequence at the onset of rlo is equivalent to assuming that tides circularize the binary instantaneously at @xmath137myr . ] . the two upper panels of fig . [ orbev ] show the relative rates of secular change of the semi - major axis @xmath108 and the orbital eccentricity @xmath109 ( for tides and gr separately ) as functions of the time expired since the formation of the bh . for all four combinations of initial orbital separations and eccentricities , the gr rates remain almost constant throughout the considered time interval , while the rates due to tides increase significantly as the star evolves on the ms . this behavior reflects the independence of gravitational radiation on the size and structure of the star and the strong dependence of tides on the ratio of the stellar radius to the orbital semi - major axis . the plots also illustrate that , for binaries consisting of a compact object and a young early - type secondary , gravitational radiation and tides , at some points in time , may affect the orbital evolution to comparable degrees . the actual evolution of the semi - major axis @xmath108 , the orbital eccentricity @xmath109 , and the periastron distance @xmath138 due to the combined effect of tides and gravitational radiation is displayed in the lower - left panels of fig.[orbev ] . as expected from the low rates of secular change displayed in the upper two panels , the orbital parameters remain almost constant during the first 300 - 400myr of evolution and start changing noticeably when the star gets closer to the end of the ms and tidal effects become dominant . finally , in the lower - right panel of fig . [ orbev ] we compare the evolution of the stellar radius with the evolution of the roche - lobe radius at the periastron of the ms star s relative orbit for each of the four considered combinations of initial orbital separations and eccentricities . the roche - lobe radii at periastron are determined by @xmath139 where @xmath140 ( eggleton 1983 ) . it follows that although all four pairs of initial orbital separations and eccentricities lead to orbital configurations that are consistent with the successful mt sequence bp[example1 ] at the onset of rlo , only one pair ( @xmath141 , @xmath142 ) leads to rlo at periastron at a time that is sufficiently close to the age of the donor star required to make the sequence successful in the first place . the other three pairs all yield rlo at periastron at times that are @xmath143myr too early . they would therefore only yield viable progenitors for groj1655 - 40 if rlo at periastron is assumed to conserve the binary s total angular momentum and if the total amount of mass transferred from the donor star is small enough to not significantly change its properties so that it would follow a different evolutionary path in the h - r diagram . if either of these conditions is violated , the considered pair of initial orbital separation and eccentricity is incompatible with the starting point of the successful rlo sequence under consideration . at the time of the sn explosion , the mass lost from the system and possibly the kick imparted to the bh during the explosion change the binary s orbital parameters . the pre- and post - sn component masses , orbital semi - major axis , and orbital eccentricity are related by the conservation laws of orbital energy and angular momentum , which in turn depend on the magnitude and the direction of the kick velocity that may be imparted to the bh . we describe these relations and the associated progenitor and kick velocity constraints in more detail in sections [ symmsn ] and [ asymmsn ] . in these sections , we adopt the same notations for the orbital elements as before , but add the subscripts `` presn '' and `` postsn '' to distinguish between their values before and after the sn explosion that formed the bh . we first note though that there are two additional constraints that need to be imposed on the bh progenitor besides those arising from the orbital dynamics of symmetric and asymmetric sn explosions . one is that , for each successful mt sequence , the progenitor mass must of course be more massive than the initial bh mass at the onset of rlo . the other is that the bh progenitor must fit within its pre - sn roche lobe . strictly speaking this is not a necessary condition ( dewi & van den heuvel 2004 ; willems & kalogera 2004 ; and willems et al . 2004 , for example , showed that the progenitor of psr j0737 - 3039b was filling its roche lobe at the time of its collapse ) . nevertheless , we impose this condition to avoid complications arising from accretion - induced changes in the stellar structure of the ms companion that will become the xrb donor star during the x - ray phase . we have tested this assumption by omitting the second constraint from our calculations and found no appreciable change in the derived progenitor and kick velocity constraints . the orbital dynamics of symmetric sn explosions have been discussed extensively in the past ( e.g. , blaauw 1961 ; boersma 1961 ; hills 1983 ; fryer & kalogera 1997 ) . we recall that the changes in the pre - sn orbital parameters are entirely determined by the amount of mass lost from the system during the explosion , yielding a post - sn semi - major axis and orbital eccentricity given by @xmath144 @xmath145 where @xmath49 is the mass of the bh s helium star progenitor . the mass loss from the system also imparts a kick to the binary s center of mass with a magnitude equal to @xmath146 where @xmath147^{1/2 } \label{sym4}\ ] ] is the relative orbital velocity of the helium star just before its instantaneous sn explosion . for asymmetric sn explosions , the newly formed bh receives a kick at birth , the magnitude and direction of which can either counteract or reinforce the effect of the mass lost from the system . the pre- and post - sn binary parameters are related by the equations @xmath148 @xmath149 } \nonumber \\ & = & g \left ( m_{\rm bh } + m_2 \right ) a_{\rm postsn } \left ( 1 - e_{\rm postsn}^2 \right ) \hspace{1.2 cm } \label{eq2}\end{aligned}\ ] ] ( e.g. , hills 1983 ; brandt & podsiadlowski 1995 ; kalogera 1996 ; fryer & kalogera 1997 ; kalogera & lorimer 2000 ) . here , @xmath150 is the magnitude of the kick velocity , @xmath151 is the polar angle between the kick velocity and the relative orbital velocity of the helium star just before the sn explosion , and @xmath152 is the corresponding azimuthal angle defined so that @xmath153 represents a plane perpendicular to the line connecting the centers of mass of the binary components ( see fig . 1 in kalogera 2000 for a graphical representation ) . the requirements that the binary must remain bound after the sn explosion and that the direction of the kick must be real ( @xmath154 and @xmath155 ) impose constraints on the pre- and post - sn parameters and on the magnitude of the kick velocity imparted to the bh at birth . in particular , the ratio of the pre- to post - sn semi - major axes must satisfy the inequalities @xmath156 @xmath157 @xmath158 the first of these inequalities expresses the condition that the post - sn orbit must pass through the position of the two stars at the time of the sn explosion ( flannery & van den heuvel 1975 ) . the last two inequalities correspond to lower and upper limits on the amount of orbital contraction or expansion that can take place for a given amount of mass loss and a given magnitude of the kick velocity ( see , e.g. , kalogera & lorimer 2000 ) . the magnitude of the kick velocity imparted to the bh at birth is restricted to the range determined by ( brandt & podsiadlowski 1995 ; kalogera & lorimer 2000 ) @xmath159 @xmath160 the first inequality expresses the requirement that the binary must remain bound after the sn explosion , while the second inequality yields the minimum kick velocity required to keep the system bound if more than half of the total system mass is lost in the explosion . last , an upper limit on the mass of the bh progenitor can be derived from the condition that the azimuthal direction of the kick is real , i.e. , @xmath161 ( fryer & kalogera 1997 ) : @xmath162^{1/2 } \bigg\}^{-1 } , \nonumber \end{aligned}\ ] ] where @xmath163.\ ] ] from eqs . ( 3 ) and ( 34 ) in kalogera ( 1996 ) , it also follows that the mass loss and kick experienced during the sn explosion yield a post - sn peculiar velocity for the binary s center of mass determined by @xmath164 where @xmath165 is the square of the relative orbital velocity of the bh _ immediately after _ the sn explosion . the coefficients @xmath166 , @xmath167 , and @xmath168 are defined as @xmath169 @xmath170 @xmath171 for each successful mt sequence , we use the post - sn orbital parameters derived by following the galactic motion and the orbital evolution due to tides and gravitation radiation backward in time ( see sections [ galmotion ] and [ tidaleq ] ) and explore the parameter space made up by @xmath172 , @xmath49 , @xmath150 , @xmath151 , and @xmath152 to look for pre - sn binary configurations and kick parameters satisfying eqs . ( [ eq1])([nu3 ] ) , or , in the case of @xmath173 , eqs . ( [ sym1])([sym4 ] ) . as we will see in the next section , this allows us to identify well defined regions of progenitor and kick properties that are fully consistent with the complete set of observational constraints described in section [ gro1655 ] . the elements presented in the previous sections can now be combined to establish a complete picture of the evolutionary history of groj1655 - 40 , the pre- and post - sn binary properties , and the dynamics involved in the core - collapse event that formed the bh . for _ each _ of the mt sequences that reproduce all of the system s currently observed properties , we first trace both the motion of the system in the galaxy and the orbital evolution due to tides and gravitational radiation back in time to the formation of the bh . the bh formation time is different for each mt sequence and is estimated by the age of the donor star . this procedure gives us the binary s peculiar velocity and orbital parameters right after the bh formation . we then use the conservation laws of orbital energy and orbital angular momentum together with all the associated orbital dynamics constraints at core collapse to link the pre - sn binary properties and the possible bh kick to the post - sn binary properties . in the following paragraphs , we first , as an example , discuss the progenitor constraints derived following this procedure for a single successful mt sequence , and next summarize the conclusions from all successful mt sequences . to illustrate the derivation of the constraints , we consider the conservative mt sequence bp[example1 ] ( see table [ bpseq ] ) with initial bh mass @xmath174 , initial donor mass @xmath175 , and initial orbital period @xmath176days ( note that this is the same sequence as used in fig . [ orbev ] ) . the constraints obtained for the mass @xmath49 of the bh s helium star progenitor , the magnitude @xmath150 of the kick velocity imparted to the bh at birth , the pre - sn orbital separation @xmath172 , and the post - sn orbital separation @xmath177 and eccentricity @xmath178 are presented in fig . the tight correlation between @xmath177 and @xmath179 ( top tight panel ) arises from matching the post - sn orbital elements to those of the mt sequence at the onset of rlo . as discussed in section [ tidaleq ] , pairs of @xmath177 and @xmath179 are considered to be acceptable if , after following the orbital evolution for a time equal to the age of the donor at rlo onset , the total angular momentum is equal to that given by the mt sequence within the uncertainties of the numerical calculations . this condition , however , does not guarantee that the eccentric orbit can accommodate the bh companion throughout the time interval between bh formation to rlo . in particular , for significantly eccentric orbits , rlo at periastron may occur at a time when the donor star has not yet reached the right evolutionary stage ( see the bottom right panel in fig . [ orbev ] and the associated discussion ) . these pairs of @xmath180 will therefore be valid post - sn orbital parameters only if rlo at periastron does not significantly affect the star and the orbit . in view of our limited knowledge of mt in eccentric binaries and its effects on the orbital elements , we include these solutions in the presentation of the progenitor constraints , although we consider them to be less compatible with the initial rlo conditions imposed by the mt sequence . in order to distinguish between solutions that lead to mt at periastron at times significantly close to rlo onset in the mt sequence and those that do not , we separate them into three groups in fig . [ eg ] : solutions for which rlo at periastron occurs much too early ( i.e. more than 200myr ) are represented by light grey circles ( green in the electronic edition ) , solutions for which rlo at periastron occurs between 50 and 200myr too early are represented by dark grey circles ( blue in the electronic edition ) , and solutions for which rlo at periastron occurs within 50myr of the proper time given by the mt sequence are represented by black circles ( red in the electronic edition ) . we consider the progenitor constraints associated with the latter solutions to be the most compatible with the mt sequence under consideration and therefore refer to them as our `` preferred set '' of solutions . the constraints on @xmath172 , @xmath177 , and @xmath178 are very tight when restricted to progenitors for which rlo at periastron occurs within @xmath181myr of the donor age imposed by the mt sequence : @xmath182 , @xmath183 , and @xmath184 . these ranges broaden somewhat when solutions for which rlo at periastron occurs between @xmath181 and @xmath185myr too early are included , but overall the differences with our preferred set of solutions is not too large . when rlo at periastron occurs more than @xmath186myr too early , additional highly eccentric post - sn orbits with large semi - major axes become available . correspondingly , the constraints on the pre - sn orbital separation also extend to very large values . a similar behavior is observed for the constraints on @xmath49 and @xmath150 . when rlo at periastron occurs within @xmath187myr of the required donor age , @xmath188 and @xmath189 . for our preferred set of solutions , the formation of the bh must therefore be accompanied by a natal kick . when solutions for which rlo at periastron occurs within @xmath190myr of the required donor age are included , the ranges slightly broaden to @xmath191 and @xmath192 . hence , the derived constraints are fairly robust even when we allow rlo at periastron to occur up to @xmath186myr too early . somewhat larger variations are found when rlo at periastron occurs more than @xmath186myr too early , but mt for these systems is likely to significantly affect both the star and the orbit so that the final parameters after @xmath136myr ( the required age of the donor at the start of rlo ) of orbital evolution are probably incompatible with the initial conditions of the considered mt sequence and thus these solutions are not favored . since highly eccentric post - sn orbital configurations always fall under the latter category , we do not consider post - sn orbital eccentricities larger than 0.9 in the derivation of any of the constraints . to understand the core - collapse event leading to the formation of the bh , we are mainly interested in the constraints derived for the mass of the bh s helium star progenitor and the kick velocity that may have been imparted to the bh at birth . for the remainder of this section , we therefore restrict ourselves to presenting the constraints derived for these two quantities . given that there is no way to distinguish which of the various successful mt sequences corresponds to the true progenitor of groj1655 - 40 , we derive the constraints for each successful sequence and examine them collectively . the constraints resulting from our preferred set of solutions for each individual sequence are summarized in the last two columns of tables [ gboseq][bpseq ] . in fig . [ overview ] , we first show the constraints on @xmath49 and @xmath150 for a selection of mt sequences with initial parameters on the edges of the regions leading to successful sequences ( see fig . [ init ] ) for conservative and non - conservative mt and for gbo and bp constraints separately . for clarity , only our preferred sets of solutions are plotted . the variations in the binary properties at the onset of rlo are seen to yield only small variations in the constraints for @xmath49 and @xmath150 . we can therefore be confident that the derived constraints are not too much affected by the finite number of successful mt sequences found , as long as we bare in mind that we outlined the ranges of initial component masses and orbital periods at the onset of rlo leading to successful sequences to within a predetermined uncertainty of @xmath193 in the masses and 0.1days in the orbital period . since we expect the true mt sequence describing the evolution of groj1655 - 40 from the start of the x - ray phase to its present configuration to be encompassed by the successful mt sequences , we overlay the @xmath49 and @xmath150 constraints for all the successful conservative and non - conservative mt sequences found for gbo and bp system parameters in fig . [ overview2 ] . it follows that even the combination of the extreme cases of fully conservative and fully non - conservative mt does not significantly relax the constraints on the progenitor masses and kick velocities with respect those obtained for individual sequences . for our preferred set of solutions , the helium star mass and kick velocity are constrained to @xmath194 and @xmath195 in the case of gbo parameters , and to @xmath196 and @xmath197 in the case of bp parameters to the binary s center of mass . if all solutions associated with @xmath198 are considered , the mass loss imparts a kick of @xmath199 to the binary s center of mass , corresponding to the entire range of admissible post - sn peculiar velocities found from the galactic motion calculations presented in section [ galmotion ] . ] . on average , higher progenitor masses furthermore tend to be associated with lower kick velocities since more mass loss requires less of kick to achieve the same result . a similar correlation was found by willems et al . ( 2004 ) for the progenitor of psrj0737 - 3039b . overlaying our preferred set of solutions for the other binary parameters yields @xmath200 , @xmath201 , and @xmath202 for gbo system parameters , and @xmath203 , @xmath204 , and @xmath205 for bp system parameters . as mentioned in section [ galmotion ] , the errors on the post - sn peculiar velocity , which imposes stringent constraints on @xmath49 and @xmath150 , were estimated by considering the galactic orbits of groj1655 - 40 for all possible combinations of the lower and upper limits on the observed distance and velocity components . since this likely overestimates the true errors on the post - sn peculiar velocity , we reconsidered the derivation of the bh progenitor mass and kick velocity constraints for a fixed error of @xmath206 on @xmath100 ( represented by the dark grey area in fig . [ vcm ] ) . the resulting constraints are shown in fig . [ overview3 ] . for a given helium star mass , the range of possible kick velocities generally narrows with respect to the range found in fig . [ overview2 ] , although the overall ranges in @xmath207 and @xmath150 do not differ significantly . are results are therefore remarkably robust with respect to the uncertainties in the post - sn peculiar velocity . we conclude that in the case of gbo parameters , it is likely that the bh in groj1655 - 40 received a kick at birth . depending on the errors in the post - sn peculiar velocity , the minimum kick velocity is @xmath208@xmath209 . in the case of bp parameters , a kick is much less of a requirement unless very small errors on the post - sn peculiar velocity are assumed . the main differences between the constraints derived for gbo and bp parameters are related to the higher initial bh masses and the correspondingly higher helium star masses which reduce the center - of - mass velocity imparted to the binary by a symmetric sn explosion in the case of gbo parameters . to model the collapse of the possible bh progenitors of gro j1655 - 40 , we use the pre - collapse stellar models of rauscher et al . ( 2003 ) . since the binary formed in the galactic disk , we focus on the solar metallicity models . we assume the mt phase that revealed the helium - star occurred late enough in the star s life ( @xmath210 case c mt ) that its structure was only slightly modified due to the loss of the hydrogen envelope , allowing us to simply remove the hydrogen envelopes from these single - star models to incorporate the effects of close binary evolution . one of the major uncertainties in our calculations is the mass loss . the high mass - loss rates in current stellar models are often too high to produce bhs in close binaries . this excessive mass - loss is especially extreme in the wolf - rayet phase , effectively making it impossible to make a bh from a star that loses its hydrogen envelope before the end of helium burning ( wellstein & langer 1999 ) . with current stellar models , we can not make bhs from systems that lose their hydrogen envelopes in a case b mt phase . even limiting ourselves to case c mt , it is difficult to produce helium cores above 11 - 12@xmath211 . this is the first constraint that stellar and core - collapse models can place on the helium star progenitor : if wind mass - loss rates are correct , @xmath212@xmath211 . this may be a real constraint , but could also be due to over - estimates of the mass - loss rates from winds by the stellar evolution community . to mimic helium - star bh progenitors of high mass up to 15@xmath211 , we use the @xmath213 solar metallicity progenitors from rauscher et al.(2003 ) . since the metallicity principally effects the mass loss alone ( heger et al . 2003 ) , these models can be used to represent the fate of solar - metallicity stars with a reduced mass loss . these stellar models / profiles are mapped into a 1-dimensional lagrangian hydrodynamics code ( herant et al . 1994 ; fryer et al.1999 ) . the equation of state used in the simulations includes a nuclear equation of state at high densities ( lattimer & swesty 1991 ) , and below @xmath214 an equation of state for nuclei obeying boltzmann statistics ( blinnikov , dunina - barkovskaya , & nadyozhin 1996 , timmes & arnett 1999 ) . both in the high density equation regime and in the low - density regime above temperatures of @xmath215k , abundances are determined assuming nuclear statistical equilibrium . between @xmath216 and @xmath215 , the abundances are evolved using a 14-element nuclear network ( benz , hills , & thielemann 1989 ) which yields reasonable approximations for the nuclear energy generation rate ( timmes , hoffman & woosley 2000 ) . lcccccc 6.2 & 0.9 & 6.2 & 0.0 & 0 . & 0 . & 0 . + 6.2 & 1 & 4.59 & 0.9 & 0.068 & 1.0 & 0.08 + 6.2 & 1.1 & 2.39 & 1.4 & 0.19 & 1.4 & 0.2 + 8.2 & 1.0 & 6.33 & 0.8 & 0.083 & 1.0 & 0.09 + 8.2 & 1.1 & 5.42 & 1.0 & 0.17 & 1.3 & 0.18 + 8.2 & 1.2 & 3.16 & 1.6 & 0.32 & 1.6 & 0.33 + 10.0 & 1.0 & 7.14 & 0.8 & 0.16 & 1.3 & 0.17 + 10.0 & 1.06 & 4.16 & 1.6 & 0.64 & 2.0 & 0.65 + 10.0 & 1.12 & 2.26 & 2.0 & 1.3 & 2.5 & 1.3 + 14.9 & 0.92 & 14.9 & 0 . & 0 . & 0 . & 0 . + 14.9 & 1.00 & 5.34 & 7.1 & 2.26 & 3.0 & 2.2 + 14.9 & 1.04 & 2.23 & 12.0 & 7.7 & 4.4 & 7.1 + neutrino transport is mediated using an explicit flux - limited algorithm ( herant et al . 1994 ) . above a radius determined by the neutrino mean free path , the neutrinos are transported in the free streaming limit , escaping the star instantly . for most models ( e.g. , fryer et al . 1999 ) , this radius is chosen such that less than 3 - 5% of the neutrinos in the free streaming limit interact with the matter as it leaves the star . in the models presented here , the neutrinos in this free streaming limit are increased by a factor @xmath217 and the limit is set to force roughly 10% of the neutrinos in the free - streaming limit to interact with matter . this effectively raises the neutrino energy in the `` gain '' region where neutrino heating deposits energy into the stellar material . by increasing @xmath217 , we increase the neutrino heating and can drive a sn explosion ( table [ ccmod ] lists the suite of these simulations ) . the high sensitivity of remnant mass on our fudge factor @xmath217 demonstrates just how delicate the balance between success and failure is for the neutrino - driven mechanism . this delicate balance is one reason why core - collapse theorists have struggled to get consistent explosions in their simulations of these explosions ( getting all of the necessary physics accurate to the 10% level is a daunting task ) . although a robust mechanism may actually exist where 10% effects are not important , it is important to understand that the actual progenitors of core - collapse ( massive stars with masses above 89@xmath211 ) are not too dissimilar , yet nature produces a range of remnant masses . it appears that nature too is sensitive to small changes in the details . a massive star collapses until it reaches nuclear densities . a proto - ns is formed and the explosion occurs if the neutrinos from the cooling ns can provide enough energy to drive off the rest of the star . as the explosion is launched , matter expands off the ns surface , and the cells are split to retain resolution with our lagrangian code near the ns surface . after @xmath2101s the explosion shock is well - developed ( @xmath218 km ) . at this point , the ns is cut out of the simulation and modelled as a gravitational source with a hard boundary . the explosion is then followed until the shock reaches the edge of the surface of the helium star . at this point , the energy in the exploding star is primarily in kinetic energy and we can determine the rough sn explosion energy of these objects ( see table [ ccmod ] ) . we also determine the bh mass by assuming all material with velocities less than the escape velocity will ultimately fall back onto the ns ( see table [ ccmod ] ) . for all cases , we can produce a bh with a mass lying somewhere between @xmath219@xmath220 , but the resulting energies for these bhs lies between @xmath221erg . our understanding of core - collapse theory can also be used to constrain the bh progenitors . fryer ( 1999 ) found that , using a multi - dimensional core - collapse code , massive stars above @xmath21020@xmath211 achieved explosion energy ( when the shock was near 1000 - 3000 km ) of roughly @xmath222erg . there are a number of uncertainties in these `` first - principle '' collapse calculations and this answer can change considerably with more accurate codes . but if we take these simulations at face value and assume that the energy is , within a factor of 2 , accurate , we can also rule out some progenitors . from table [ ccmod ] , we see that only progenitors below 10@xmath211 can obtain the appropriate bh mass with less than about @xmath223erg of explosion energy when the shock is launched . this constraint ( the second from stellar and core - collapse models ) is strongest for small bh masses . we can also use the nucleosynthetic yields to constrain the models . as noted above , the 14 isotope nuclear network used in the hydrodynamics simulations is used only to determine energetics . to obtain more detailed abundance information , we post - processed the thermodynamic trajectories of our lagrangian particles with a modern 500 isotope network ( e.g. , timmes 1999 ) . for simplicity we assumed an initial electron fraction of 0.5 . the iron , titanium , sulfur and oxygen yields as a function of enclosed mass for three representative models are presented in fig . [ fig : abun ] . these three models : @xmath224 , @xmath225 , @xmath226 , @xmath227 , @xmath228 , @xmath229 , each produce bhs with masses above 4@xmath211 . oxygen and even some sulfur is formed well out in the star , but titanium and iron are formed in the most intense burning at the base of the explosion . however , note that the more massive stars produce iron and titanium further out than the low - mass stars . this occurs because the more massive stars require stronger explosions to produce bhs of comparable mass to the lower mass stars . the stronger explosions produce intensive nuclear burning further out in the star , ultimately producing iron further out in the star as well . do such abundance profiles provide additional , useful information ? we often use the nucleosynthetic signature from supernovae to constrain sn progenitors . however , the yields from these stars must mix out extensively in the explosion just to get ejected . these bh forming stars will not dominate the nuclear yields of the galaxy . in addition , due to their weak explosions , the nucleosynthetic yields from these explosions are not likely to mix extensively into the interstellar medium . the timescale ( @xmath230 ) for the sn shock to once again become part of the interstellar medium ( when the shock velocity drops below the sound speed of the ambient medium ) is roughly ( mckee & hollenbach 1980 ) : @xmath231 where @xmath232 is the explosion energy in units of @xmath233erg ; and @xmath234 in units of @xmath235cms@xmath1 and @xmath236 are , respectively , the sound speed and particle density of the ambient medium . the corresponding maximum extent ( @xmath237 ) of the sn shock is : @xmath238 the timescales and effective volumes ( as a fraction of a normal @xmath233 erg explosion ) of the remnants of these bhs are given in table [ ccmod ] . with such short - lived , small remnants , it is unlikely that we will be able to constrain these stars by galactic or remnant abundances . however , israelian et al . ( 1999 ) observed the companion of gro j1655 - 40 during a quiescent x - ray phase using the high - resolution spectrograph at keck . at visible wavelengths , they found that the abundances of o , mg , s , si , and ti were 6 - 10 times larger than solar . they also found that [ fe / h ] = 0.1 @xmath239 0.2 , i.e. almost solar . the most straightforward and logical way of explaining these overabundances , and the one proposed by israelian et al . , is that these elements were produced during explosive oxygen and silicon burning during the bh formation and have enriched the companion . but to cause this enrichment , these elements must make it to the companion star . for the iron and titanium , this can be quite a difficult task . at the time of explosion the companion was roughly @xmath240 from the center of the collapsing star . but our 1-dimensional explosion models find that the iron and titanium does not make it beyond 0.05@xmath241 before falling back onto the compact remnant . even in 3-dimensional simulations , the amount of mixing found in symmetric models does not place the iron or titanium beyond 0.5@xmath241 before these elements fall back onto the compact remnant . the extensive mixing required to eject these elements requires large asymmetries in the explosion ( hungerford et al . 2003 ; hungerford et al . 2004 ) . hence , if the enrichment of the companion of gro j1655 - 40 is arising from the explosion forming the bh , its sn must have been asymmetric ! the amount of asymmetry reaches a minimum with the lower mass progenitors ( especially with massive bhs ) . for example , in the extreme case of a bh mass of 6.2@xmath211 , the 6.2@xmath211 progenitor will not , no matter what the asymmetry , be able to enrich the companion . even with smaller bh masses , it is likely that the helium progenitor for this system is more massive ( @xmath242 ) . this is our third , and final , constraint on the helium star mass . in principal , the exact abundances of the companion star could be used to determine the best - fitting bh progenitor ( see podsiadlowski et al . however , the uncertainties in the level of asymmetry and the uncertainties of the effects of asymmetry on the actual nuclear yield make it difficult to produce any accurate quantitative results . these uncertainties must then be folded into uncertainties in the accretion onto the companion . the study of asymmetry - induced mixing is still in its infancy . when it is better understood , the actual composition of the companion star can be used to constrain the progenitors as well . although there exist a number of uncertainties in stellar evolution and core - collapse calculations , these models can place additional constraints on the bh progenitor for groj1655 - 40 : * * i : @xmath243 - stellar winds . * with current mass - loss rates for winds , helium cores above this mass are simply not produced . the limitation of this constraint lies in the uncertainties in stellar winds . given that bh masses for some other systems have been estimated to be as high as @xmath244 ( mcclintock & remillard 2004 , and references therein ) , we do not consider this constraint to be very stringent . * * ii : @xmath245 - supernova energetics . * if explosion energies for these massive stars do decrease with increasing stellar mass , it is impossible to get low - mass bhs with more massive helium stars . this constraint is strongest for the low - end of the bh mass range allowed for gro j1655 - 40 . this constraint is limited by uncertainties in the core - collapse mechanism . * * iii : @xmath246 - nucleosynthetic yields . * if the high abundances of the companion to gro j1655 - 40 are due to ejecta enrichment , large asymmetries are required to mix out the iron and titanium so that it is ejected . prohibitively large asymmetries are required for low - mass progenitors . this constraint is strongest for bh masses at the high - end of the allowed range for gro j1655 - 40 . the dominant limitation in this constraint arises from the uncertainty of the true enrichment in the companion of gro j1655 - 40 . in addition to constraints on the progenitor mass , our calculations also show how observations of gro j1655 - 40 can constrain sn explosions . if the anomalous abundances in the companion to gro j1655 - 40 truly arise from enrichment from the bh - forming sn , then this object provides some of the best evidence not only for the occurrence of a sn explosion , but also for asymmetries in the explosions ( israelian et al . 1999 , podsiadlowski et al . 2002b ) . a better constraint on the progenitor ( as well as a better understanding of structure of the asymmetries ) could significantly constrain the magnitude of the asymmetry in this explosion . in this paper , we initiated a comprehensive study on the formation and evolution of galactic xrbs . based on the current binary properties and the current position and motion of observed xrbs , the kinematic and evolutionary history is traced back in time and constraints are derived for the pre - sn binary progenitor and for the kick velocity that may have been imparted to the compact object at birth . the analysis consists of the following elements : ( i ) mt calculations to model the ongoing x - ray phase , ( ii ) orbital evolution due to tides and gravitational radiation between the compact object formation time and the onset of rlo , ( iii ) motion of the system in the galaxy after the formation of the compact object , ( iv ) binary orbital dynamics at the time of core collapse , and ( v ) hydrodynamic modeling of the actual core collapse event . as a first application , we constrained the progenitor properties and the formation of the bh in the soft x - ray transient groj1655 - 40 . for this purpose , we calculated a large number of mt sequences to map the currently observed binary properties ( component masses , position of the donor in the h - r diagram , orbital period , and transient behavior ) to those at the onset of the mt phase . uncertainties related to the accretion of matter by the bh are incorporated by considering both conservative ( for sub - eddington rates ) and non - conservative mt . for the current system configuration , we considered both the parameters derived by gbo and those derived by bp ( see table [ 1655param ] for details ) . in the case of gbo parameters , successful mt sequences able to reproduce all currently observed system properties were found for initial bh masses @xmath247@xmath65 , initial donor star masses @xmath248@xmath61 , and initial orbital periods @xmath68@xmath63days . in the case of bp parameters , successful sequences were found for initial bh masses @xmath71@xmath249 , initial donor star masses @xmath250@xmath74 , and initial orbital periods @xmath251@xmath76days . for both cases , all successful sequences have a ms donor star at the onset of rlo . a graphical representation of the initial donor masses and orbital periods leading to successful mt sequences is presented in fig . [ init ] . properties of the sequences are given in tables [ gboseq][bpseq ] . the results of the mt calculations are in good agreement with previous investigations . regs , tout , & wickramasinghe ( 1998 ) showed that the present - day properties of groj1655 - 40 ( as determined by orosz & bailyn 1997 ) could be reproduced by a binary initially consisting of a @xmath252 ms star orbiting a @xmath253 bh with a period of 1.935days . these initial parameters are fairly close to those of the successful mt sequence gbo[rtw ] in table [ gboseq ] , except for the higher initial bh mass adopted by regs et al . this small difference can be attributed to the higher current bh mass of @xmath254 estimated by orosz & bailyn ( 1997 ) as well as to some different input physics in the stellar evolution codes , such as , e.g. , the amount of convective overshooting . kolb et al . ( 1997 ) and kolb ( 1998 ) , on the other hand , suggested that groj1655 - 40 may be in a very short - lived evolutionary phase in the hertzsprung gap during which the expansion of the donor s radius temporarily halts or even reverses . as a consequence , the mt rate drops below the high rates typically expected for hertzsprung - gap donor stars , allowing for a short phase of transient behavior . in view of the small size of the relevant region in the hertzsprung gap ( see , e.g. , fig . 2 in kolb et al . 1997 ) and the associated short evolutionary time scales , progenitors which start rlo while the donor is still on the ms seem statistically more favorable . we therefore did not attempt to reproduce the small transient strip in the hertzsprung gap found by kolb et al . ( 1997 ) and kolb ( 1998 ) . in their investigation , kolb et al . ( 1997 ) also adopted the bh mass derived by orosz & bailyn ( 1997 ) , which is lower than that derived by both gbo ( @xmath41 ) and bp ( @xmath255 ) . since , for a given donor mass and orbital period , the mt rate tends to increase with decreasing bh mass , the small transient strip in the hertzsprung gap may disappear when the more recent bh mass estimates are adopted . bp , finally , found that , using their estimates for the system parameters , both case a and early case b mt sequences are able to reproduce the observed system properties , but that case a sequences generally provided more satisfactory fits . their example sequence with initial parameters @xmath256 , @xmath175 , and @xmath257days is close to our sequence bp[bpeg ] ( see table [ bpseq ] ) , although the latter has a somewhat higher initial donor mass , possibly due to differences in the adopted convective overshooting . a crucial element emerging from the mt calculations is the age of the donor star when all observational constraints are satisfied ( i.e. the donor s _ current _ age ) . by using this age as an estimate for the time expired since the bh formation , we determined the possible birth sites of the bh by following the motion of the system in the galaxy backwards in time . it follows that groj1655 - 40 was born between 3 and 7kpc from the galactic center , and within 200pc from the galactic plane . the post - sn peculiar velocity at the birth site can be anywhere between @xmath101 and @xmath258 . we use this post - sn peculiar velocity to constrain the amount of mass lost in the sn explosion and the kick velocity that may have been imparted to the bh . this is a major difference with previous investigations on the kinematics of groj1655 - 40 which used the _ present - day _ radial velocity ( after correcting for the motion of the sun and differential galactic rotation , @xmath259 ; brandt et al . 1995 ) as a lower limit to the binary s center - of - mass velocity ( see brandt et al . 1995 , nelemans et al . 1999 , fryer & kalogera 2001 ) . the mt sequences also yield the age of the donor at the onset of rlo . knowledge of this age allows us to link the post - sn orbital parameters to those at the onset of rlo by numerically integrating the system of differential equations governing the orbital evolution due to tides and gravitational radiation . post - sn orbital semi - major axes and eccentricities compatible with successful mt sequences range from @xmath260 to @xmath261 and from @xmath262 to @xmath263 . the solution of the equations describing the conservation of orbital energy and angular momentum during compact object formation allows us to derive constraints on the pre - sn orbital separation , on the mass of the bh s helium star progenitor , and on the magnitude of the kick velocity that may have been imparted to the bh at birth . despite the large number of successful mt sequences found , the progenitor constraints turn out to be fairly robust to the initial parameters of the sequences at the onset of rlo ( see fig . [ overview ] ) . combining the constraints obtained for all successful mt sequences for both conservative and non - conservative mt , yields @xmath264@xmath265 and @xmath266@xmath267 for gbo system parameters and @xmath268@xmath269 and @xmath270@xmath271 for bp system parameters ( see fig . [ overview2 ] ) . it is clear that symmetric sn explosions lie at the edge of our solution ranges and that bh kicks of at least a few tens of kms@xmath1 are favored by the majority of the successful mt sequences ( see tables [ gboseq][bpseq ] ) . hence , although some of the constraints associated with individual mt sequences require a kick to be imparted to the bh at birth ( see figs . [ eg][overview3 ] and tables [ gboseq][bpseq ] ) , without any additional information , the still remaining uncertainties in the present - day system parameters at this stage do not allow us to exclude that the bh in groj1655 - 40 may have been formed through a symmetric sn explosion . in the final step of our investigation , we therefore model the core collapse using a 1d lagrangian hydrodynamics code . from the energetics of the sn explosion , it follows that the mass of the bh s helium star progenitor must be larger than @xmath272 . in addition , if the observed overabundances of heavy elements in the bh s companion are to be attributed to pollution from sn ejecta , the nucleosynthesis yields imply a maximum helium - star mass of @xmath273 . combining this with the constraints derived from the mt sequences , orbital dynamics , and motion in the galaxy gives @xmath274 in the case of gbo parameters and @xmath275 in the case of bp parameters . our final set of constraints on the progenitor are much different than those obtained by podsiadlowski et al . ( 2002b ) who predict the likely progenitor to lie between @xmath276 . but these differences are not due to different estimates of a given constraint , but the application of different constraints . podsiadlowski et al . ( 2002b ) focus their attention on the enrichment of the companion . as with our enrichment constraint ( constraint iii from 9 ) , they find that more massive stars are better suited to explain the nucleosynthetic enrichment in the companion . our constraint is slightly weaker , based on the simulations of hungerford et al . ( 2003 , 2004 ) which suggest that asymmetries can lead to very extended mixing . we have an additional limit based on both our mass - loss and supernova explosion model constraints ( i and ii from 9 ) that push for lower - mass progenitors . since podsiadlowski et al . ( 2002b ) do nt consider these constraints , they do not have this upper limit in their paper . the question of whether or not a natal kick was imparted to the bh in groj1655 - 40 was addressed previously by brandt et al . ( 1995 ) , nelemans et al . ( 1999 ) , and fryer & kalogera ( 2001 ) . all three of these investigations used a lower limit on the present - day center - of - mass velocity given by the _ current _ radial velocity ( @xmath277 ) instead of the actual post - sn peculiar velocity to constrain the mass loss and kick magnitude associated with the bh s formation . as illustrated in fig . [ vcm ] , this neglects changes in the system s center - of - mass velocity resulting from its acceleration in the galactic potential . we here find that , even though following the galactic motion of the system backward in time yields post - sn peculiar velocities all the way down to @xmath278 , the additional constraints on the binary properties make the possibility of a symmetric sn explosion only marginally acceptable solutions . jonker & nelemans ( 2004 ) recently reconsidered the vertical distribution of low - mass xrbs with respect to the galactic plane and found no significant difference between ns and bh systems , suggesting that bhs may be subjected to natal kicks as well . the authors furthermore noted that the distances to galactic soft x - ray transients may be systematically underestimated . if this is truly so , revised distance estimates may have important consequences for the determination of quiescent x - ray luminosities , peak outburst luminosities , bh masses , and galactic distribution of bh xrbs . clearly , if future measurements significantly revise any of the observational constraints , the analysis presented in this paper should be repeated to account for the most up - to - date knowledge of groj1655 - 40 s current system properties . in subsequent analyses , we intend to apply the procedure outlined above to the soft x - ray transient xte1118 + 480 , the high - mass xrbs cyg x-1 , ls5039 , lsi@xmath279 , velax-1 , 4u1700 - 37 , and the ns low - mass xrb scox-1 . by examining both ns and bh systems and both rlo and stellar wind induced mt , we hope to unravel the systematic dependencies between the masses of newly formed compact objects and their immediate pre - sn progenitors , the mass lost at core collapse , and the possible kick velocity magnitude imparted to the compact object at birth . we expect that such constraints will help us understand better the physical origin of asymmetries in the collapse of massive stars . we are indebted to laura blecha for sharing the code used to follow the motion of groj1655 - 40 in the galactic potential , and to jon miller , jerome orosz , jeffrey mcclintock , klaus schenker , and gijs nelemans for useful and stimulating discussions . this work is supported by a david and lucile packard foundation fellowship in science and engineering grant and made extensive use of nasa s astrophysics data system bibliographic services . ankay , a. , kaper , l. , de bruijne , j.h.j . , dewi , j. , hoogerwerf , r. , & savonije , g.j . 2001 , a&a , 370 , 170 bailyn , c.d . , et al . 1995a , nature , 374 , 701 bailyn , c.d . , orosz , j.a . , mcclintock , j.e . , & remillard , r.a . 1995b , nature , 378 , 157 beer , m. e. & podsiadlowski , ph . 2002 , mnras 331 , 351 ( bp ) belczynski , k. , kalogera , v. , taam , r.e . , rasio , f.a . , zezas , a. , bulik , t. , & maccarone , t. 2004 , in preparation benz , w. , hills , j.g . , & thielemann , f .- k . 1989 , apj , 342 , 986 blaauw , a. 1961 , bulletin of the astronomical institute of the netherlands , 15 , 265 blinnikov , s.i . , dunina - barkovskaya , n.v . , & nadyozhin , d.k . 1996 , apjs , 106,171 boersma , j. 1961 , bulletin of the astronomical institute of the netherlands , 15 , 291 brandt , n. & podsiadlowski , ph . 1995 , mnras , 274 , 461 brandt , w.n . , podsiadlowski , p. , & sigurdsson , s. 1995 , mnras , 277 , l35 carlberg , r.g . & innanen , k.a . 1987 , aj , 94 , 666 clark , g.w . 1975 , apj , 199 , l143 de kool , m. , van den heuvel , e.p.j . , & pylyser , e. 1987 , a&a , 183 , 47 dewi , j.d.m . & van den heuvel , e.p.j . 2004 , mnras , 349 , 169 dubus , g. , lasota , j. , hameury , j. , & charles , p. 1999 , mnras , 303 , 139 eggleton , p.p . 1983 , apj , 268 , 368 eggleton , p.p . & verbunt , f. 1986 , mnras , 220 , 13p ergma , e. & van den heuvel , e.p.j . 1998 , a&a , 331 , l29 flannery , b.p . & van den heuvel , e.p.j . 1975 , a&a , 39 , 61 fryer , c. 1999 , apj , 522 , 413 fryer , c. & kalogera , v. 1997 , apj , 489 , 244 fryer , c. & kalogera , v. 2001 , apj , 554 , 548 fryer , c.l . , benz , w. , herant , m. , & colgate , s.a . 1999 , apj , 516 , 892 greene , j. , bailyn , c.d . , & orosz , j.a . 2001 , apj , 554 , 1290 ( gbo ) gualandris , a. , colpi , m. , portegies zwart , s. , & possenti , a. 2005 , apj , 618 , 845 heger , a. , fryer , c.l . , woosley , s.e . , langer , n. , & hartmann , d.h . 2003 , apj , 591 , 288 herant , m. , benz , w. , hix , w.r . , fryer , c.l . , & colgate , s.a . 1994 , apj , 435 , 339 hills , j.g . 1976 , mnras , 175 , 1p hills , j.g . 1983 , apj , 267 , 322 hjellming , r.m . & rupen , m.p . 1995 , nature , 375 , 464 humphreys , r.m . & larsen , j.a . 1995 , aj , 110 , 2183 hungerford , a.l . , fryer , c.l . , & warren , m. s. 2003 , apj , 594 , 390 hungerford , a.l . , fryer , c.l . , rockefeller , g. , & timmes , f. x. 2004 , in preparation hurley , j.r . , pols , o.r . , & tout , c.a . 2000 , mnras , 315 , 543 hurley , j.r . , tout , c.a . , & pols , o.r . 2002 , mnras , 329 , 897 hut , p. 1981 , a&a , 99 , 126 israelian , g. , rebolo , r. , basri , g. , casares , j. , & martn , e.l . 1999 , nature , 401 , 142 ivanova , n. , belczynski , k. , kalogera , v. , rasio , f.a . , & taam , r.e . 2003 , apj , 592 , 475 jonker , p.g . & nelemans , g. 2004 , mnras , 354 , 355 junker , w. & schfer , g. 1992 , mnras , 254 , 146 kalogera , v. 1996 , apj , 471 , 352 kalogera , v. 1999 , apj , 521 , 723 kalogera , v. 2000 , apj , 541 , 319 kalogera , v. & lorimer , d.r . 2000 , apj , 530 , 890 kalogera , v. , henninger , m. , ivanova , n. , & king , a.r . 2004 , apj , 603 , l41 king , a.r . , kolb , u. , & burderi , l. 1996 , apj , 464 , l127 king , a.r . , kolb , u. , & szuszkiewicz , e. 1997 , apj , 488 , 89 kolb , u. , king , a.r . , ritter , h. , & frank , j. 1997 , apj , 485 , l33 kolb , u. 1998 , mnras , 297 , 419 kuijken , k. & gilmore , g. 1989 , mnras , 239 , 571 lattimer , j.m . , & swesty , f.d . 1991 , nucl . a. , 535 , 331 mcclintock , j.e . & remillard , r.a . 2005 , in compact stellar x - ray sources , eds . lewin & m. van der klis , ( cambridge : cambridge university press ) , astro - ph/0306213 mckay , d. & kesteven , m. 1994 , iau circ . , 6062 , 2 mckee , c.f . & hollenbach , d.j . 1980 , ara&a , 18 , 219 mirabel , i.f . , dhawan , v. , mignani , r.p . , rodrigues , i. , & guglielmetti , f. 2001 , nature , 413 , 139 mirabel , i.f . , mignani , r. , rodrigues , i. , combi , j.a . , rodrguez , l.f . , & guglielmetti , f. 2002 , a&a , 395 , 595 mirabel , i.f . & rodrigues , i. 2003a , science , 300 , 1119 mirabel , i.f . & rodrigues , i. 2003b , a&a , 398 , l25 mirabel , i.f . , rodrigues , i. , & liu , q.z . 2004 , a&a , 422 , l29 nelemans , g. , tauris , t.m . , & van den heuvel , e.p.j . 1999 , a&a , 352 , l87 orosz , j.a . 2004 , orosz , j.a . & bailyn , c.d . 1997 , apj , 477 , 876 phillips , s.n . , shahbaz , t. , & podsiadlowski , p. 1999 , mnras , 304 , 839 podsiadlowski , p. , rappaport , s. , & pfahl , e.d . 2002a , apj , 565 , 1107 podsiadlowski , p. , nomoto , k. , maeda , k. , nakamura , t. , mazzali , p. , & schmidt , b. 2002b , apj , 567 , 491 podsiadlowski , p. , rappaport , s. , & han , z. 2003 , mnras , 341 , 385 portegies zwart , s.f . , verbunt , f. , & ergma , e. 1997 , a&a , 321 , 207 rauscher , t. , heger , a. , hoffman , r.d . , & woosley , s.e . 2003 , nucl . phys . a , 718 , 463 regs , e. , tout , c.a . , & wickramasinghe , d. 1998 , apj , 509 , 362 reid , m.j . 1993 , ara&a , 31 , 345 rib , m. , paredes , j.m . , romero , g.e . , benaglia , p. , mart , j. , fors , o. , & garca - snchez , j. 2002 , a&a , 384 , 954 romani , r.w . 1996 , iau symp . 165 : compact stars in binaries , 165 , 93 shahbaz , t. 2003 , mnras , 339 , 1031 shahbaz , t. , van der hooft , f. , casares , j. , charles , p.a . , & van paradijs , j. 1999 , mnras , 306 , 89 soria , r. , wickramasinghe , d.t . , hunstead , r.w . , & wu , k. 1998 , apj , 495 , l95 timmes , f. x. 1999 , apjs , 124 , 241 tingay , s.j . , et al . 1995 , nature , 374 , 141 van der hooft , f. , heemskerk , m.h.m . , alberts , f. , & van paradijs , j. 1998 , a&a , 329 , 538 wellstein , s. , & langer , n. 1999 , a&a , 350 , 148 willems , b. & kalogera , v. 2004 , apj , 603 , l101 willems , b. , kalogera , v. , & henninger , m. 2004 , apj , in press witte , m.g . & savonije , g.j . 1999 , a&a , 350 , 129 zahn , j .- 1975 , a&a , 41 , 329 zahn , j .- 1977 , a&a , 57 , 383 zahn , j .- 1978 , a&a , 67 , 162 zhang , s.n . , wilson , c.a . , harmon , b.a . , fishman , g.j . , wilson , r.b . , paciesas , w.s . , scott , m. , & rubin , b.c . 1994 , iau circ .	in recent years proper motion measurements have been added to the set of observational constraints on the current properties of galactic x - ray binaries . we develop an analysis that allows us to consider all this available information and reconstruct the full evolutionary history of x - ray binaries back to the time of core collapse and compact object formation . this analysis accounts for five evolutionary phases : mass transfer through the ongoing x - ray phase , tidal circularization before the onset of roche - lobe overflow , motion through the galactic potential after the formation of the compact object , binary orbital dynamics at the time of core collapse , and hydrodynamic modeling of the core collapse that connects the compact object to its progenitor and any nucleosynthetic constraints available . in this first paper , we present this analysis in a comprehensive manner and we apply it to the soft x - ray transient groj165540 . this is the first analysis that incorporates all observational constraints on the current system properties and uses the full 3d peculiar velocity constraints right after core collapse instead of lower limits on the current space velocity given by the present - day radial velocity . we find that the system has remained within 200pc from the galactic plane throughout its entire life time and that the mass loss and a kick possibly associated with the black hole formation imparted a kick velocity of @xmath0115kms@xmath1 to the binary s center of mass . right after black hole formation , the system consists of a @xmath2 black hole and a @xmath3 main - sequence star . at the onset of the x - ray phase the donor is still on the main sequence . we find that a symmetric black hole formation event can not be formally excluded , but that the associated system parameters are only marginally consistent with the currently observed binary properties . black hole formation mechanisms involving an asymmetric supernova explosion with associated black hole kick velocities of a few tens of kms@xmath1 , on the other hand , satisfy the constraints much more comfortably . we also derive an upper limit on the black hole kick magnitude of @xmath4 .
central to many unsolved problems in particle physics and cosmology , ranging from the flatness of the slow - roll inflaton potential to the hierarchy of scales in grand unified theories , is the need to understand the origin of `` unnaturally '' large or small dimensionless parameters . in this letter we demonstrate how oscillons localized , oscillatory solutions to nonlinear field equations that either are stable or decay only after many cycles can provide an explicit example in which such unnatural behavior emerges dynamically . we show that in the presence of an expanding background metric , oscillons in a simple toy model have lifetimes that are exponentially large compared to the natural scales of the system , even though all the inputs to the theory are of order one in these units . although the situation is reminiscent of tunneling behavior in quantum systems , this decay is entirely classical . a wide variety of nonlinear field theories have been found to support oscillon solutions ( also known as breathers ) @xcite . in some special cases , such as the sine - gordon breather @xcite and @xmath9-ball @xcite , conserved charges guarantee the existence of exact , periodic solutions . in @xmath10 theory in three dimensions , in contrast , @xcite , oscillons are observed to decay suddenly after lifetimes of order @xmath11 to @xmath12 in natural units . in other three - dimensional models @xcite , no sign of decay is visible in current numerical simulations . in a broad class of one - dimensional models , perturbative analyses @xcite and numerical simulations @xcite suggest that the oscillon lifetime is infinite , but analytic arguments @xcite point to the existence of non - perturbative , exponentially suppressed decay modes . such models then have a curious property : in a simple , classical theory in which all parameters are of order one in natural units , classical dynamics generate a quantity the oscillon lifetime that is exponentially large compared to the natural scales of the system . in this letter , we investigate explicitly a simple example of such a system . we consider a toy model consisting of a classical real scalar field @xmath13 of mass @xmath1 in one space dimension , governed by the lagrangian density @xmath14 \ , . \label{lagden}\ ] ] in this classical theory , the dynamics are unchanged when the lagrangian density is multiplied by a constant ( equivalent to rescaling @xmath15 ) . we have used this freedom , combined with a rescaling of the field @xmath16 , to fix the magnitude of the coefficient of the @xmath10 term while maintaining the conventional form of the free term . the arbitrary choice of the ratio of magnitudes of the @xmath10 and @xmath17 terms is chosen for convenience and is not essential to our results . in the natural units of the mass @xmath1 , then , all the coefficients of the potential are of order @xmath18 . the only important choices are the signs : the sign of the @xmath19 term gives the @xmath16 field a conventional ( non - tachyonic ) mass ; the sign of the @xmath10 term is necessary for oscillons to exist ; and the sign of the @xmath17 term is necessary for the field to be stable against runaway growth . we note that we have observed the same behavior in @xmath10 theory with the standard symmetry - breaking potential . here we have chosen the @xmath17 model just to avoid any distractions that might be caused by the existence of static `` kink '' solutions in the @xmath10 theory ; our theory contains no static solutions . like the other one - dimensional models discussed above , in a static background the model of eq . ( [ lagden ] ) supports oscillons that appear to live indefinitely in numerical simulations . they are spatially localized , with size of order @xmath20 and fundamental frequencies of oscillation comparable to but always below @xmath1 . we would like to include an additional element : coupling to an expanding background metric , inspired by the consideration of oscillons in the early universe ( work in progress @xcite ; see also studies of nontopological solitons and oscillons in hybrid inflaton @xcite and axion - oscillons @xcite , and for a broader review of lattice field theory simulations in the early universe see @xcite ) . in comoving coordinates , then , the lagrangian takes the form @xmath21 a(t ) dx \label{lag}\ ] ] leading to the equation of motion @xmath22 where the physical coordinate is related to the comoving coordinate @xmath23 by the scale factor @xmath24 and we hold the hubble constant @xmath25 fixed , giving an exponential expansion rate . here @xmath26 is the derivative of @xmath16 with respect to time and @xmath27 is the derivative of @xmath16 with respect to the comoving coordinate @xmath23 . this expansion destabilizes oscillons whose width is of order of the horizon size @xmath28 , but in numerical simulations we observe that oscillons of smaller size remain stable , maintaining a fixed size in physical units . a technical advantage of this model is we can study oscillon behavior efficiently for extremely long times , because any regions at distances significantly greater than the horizon length @xmath28 from an oscillon can not influence its evolution . thus we do not need to expand the box proportionally to the runtime @xcite or introduce absorbing boundaries @xcite in order to prevent unwanted reflections from disturbing the oscillon solution under study . starting from thermal initial conditions with @xmath29 , we see that oscillons emerge copiously as the expansion cools our universe . these initial conditions are generated using a canonical ensemble of quantized modes of the free scalar field , and thus contain no fine - tuning . appears explicitly and we can no longer scale out the overall normalization of the coupling constants through our choice of units , as we have in the classical calculation above . the equal - time commutation relation @xmath30 = i\hbar\delta(x - y)$ ] implies that @xmath16 has units of @xmath31 . therefore the lagrangian density takes the form @xmath32 a(t ) dx\ ] ] where @xmath33 is a constant with units of @xmath34 . in the classical theory , we may choose our units so this quantity is equal to @xmath18 . in the thermal calculation , however , we encounter the quantity @xmath35 in the boltzmann factor for modes of energy @xmath36 . choosing units so that @xmath37 in this expression forces us to include @xmath33 explicitly in the classical lagrangian density . an oscillon solution with @xmath38 will continue to be a solution for general @xmath33 , provided that we scale up its amplitude by a factor of @xmath39 . for our classical treatment of the time evolution to be valid , the oscillon must have an energy that is large compared to the energy of a typical quantum fluctuation @xmath40 , where @xmath41 is its fundamental frequency of oscillation . since the oscillon energy goes like @xmath42 , this requirement can be implemented by choosing @xmath43 to be small the classical approximation is valid at small coupling . we have verified that our results are not affected by the value of this parameter ; as long as @xmath44 , oscillons are generated copiously . they then evolve according to the same classical dynamics with the rescaled field . ] to simplify the classical analysis , we neglect zero - point fluctuations . oscillons have been similarly found to emerge spontaneously during phase transitions @xcite . we use a straightforward numerical simulation in which we discretize @xmath23 at the level of the lagrangian in eq . ( [ lag ] ) , working in natural units where @xmath45 . for the space derivatives we use ordinary first - order differences , @xmath46 where @xmath47 refers to the value of @xmath16 at lattice point @xmath48 . we work on a regular lattice with spacing @xmath49 , and impose periodic boundary conditions . varying this lagrangian yields lattice equations of motion with second - order space derivatives , @xmath50 finally , we use second - order differences @xmath51 with @xmath52 to compute the field at @xmath53 based on the values at @xmath54 and @xmath55 . we have verified that our results are not sensitive to the particular choice of lattice spacing and time step . ( by the courant condition we must choose @xmath56 or the algorithm will be unstable . ) in the absence of expansion , our system conserves energy exactly in the limit of infinitesimal time step @xmath57 ( regardless of the spatial lattice size ) . in an expanding background , the configuration loses energy at a rate given by the pressure times the rate of change in volume @xmath58 where the pressure density is @xmath59 \,.\ ] ] we use this result to check the accuracy of our calculation , and find that it is maintained to better than @xmath18 part in @xmath12 , as we d expect for a second - order method . as the universe expands , our lattice expands with it , but the oscillons do not . therefore , whenever our lattice has expanded by a factor of @xmath60 , we refine the lattice , doubling the total number of lattice points and bringing the lattice spacing back to its original value in physical units . we assign values to the field for the new intermediate lattice points by polynomial interpolation . this numerical algorithm is highly stable , maintaining precision without any sign of degradation even after extremely long runs . we begin our simulation using values of @xmath61 and @xmath62 drawn from a random thermal configuration with @xmath63 ( setting the quantum scaling parameter @xmath33 to one for simplicity ) . as the expanding universe redshifts away the ordinary perturbative fluctuations , it becomes simple to pick out the oscillon peaks . at this point , we excise a particular oscillon , keeping only a window of size @xmath64 in physical units around it . ( we identify the starting and ending points of the oscillon profile as the places where the derivative of the energy density is @xmath65 in natural units . ) now , each time we insert new lattice points , we also truncate the lattice back down to this size , so that the total computational cost of the run scales only linearly with time . because of the expansion , any noise introduced by this truncation can never affect the oscillon dynamics . we have verified that changing the box size used for this truncation does not affect the oscillon lifetimes we observe . although the oscillon profiles we obtain by this method vary , they all exhibit common behavior . we use a small - amplitude asymptotic analysis in a static background , following @xcite , to gain some insight into the general properties of oscillon solutions . at large @xmath66 , the magnitude of @xmath16 is small and we can ignore the nonlinear terms in the equations of motion , so our solution must be of the form @xmath67 , where @xmath68 . ( we have absorbed an arbitrary phase by our choice of the zero of @xmath54 . ) therefore it makes sense to work in terms of rescaled variables @xmath69 and @xmath70 , giving the equation of motion @xmath71 + \phi ^3 - \phi^5 \,.\ ] ] next we expand @xmath16 in powers of @xmath72 using these variables , @xmath73 and consider the equations of motion order by order in @xmath72 . at @xmath74 , we have simply @xmath75 and thus @xmath76 , where we have again absorbed a phase into the definition of @xmath77 , and the profile @xmath78 remains to be determined . at @xmath79 , we have @xmath80 the key point is that for all @xmath81 , the function of @xmath77 on the left - hand side of this equation is orthogonal to @xmath82 , so the right - hand side must be as well . as a result , setting the fourier coefficient of @xmath82 on the right - hand side equal to zero gives @xmath83 whose solution , consistent with the boundary conditions at infinity , is @xmath84 . thus we have a family of approximate solutions @xmath85 parametrized by the small amplitude @xmath72 . although this analysis is modified by higher orders in @xmath72 and the inclusion of the expanding background , we find numerically that its general features remain . in particular , we can have oscillons of large width as long as they have correspondingly small amplitudes ; the oscillon s total energy , which is proportional to its width times its amplitude squared , scales linearly with its amplitude and inversely with its width . each oscillon decays by a gradual process of expansion , through which its width increases and its amplitude and energy decrease . as the width approaches the horizon size @xmath28 , this decay becomes very rapid and the oscillon decays suddenly . this process is shown in fig . [ decay ] . we define the decay of the oscillon as the time when the derivative of its energy is everywhere below @xmath65 in natural units . since the decay process is so sudden , choosing any other reasonable criterion would make a negligible change in the lifetime we measure . and the time is measured in units of @xmath20 . ] the dramatic behavior shown in fig . [ decay ] is visible only at the end of the oscillon s life , however . for slightly higher energies , this curve becomes extremely close to horizontal ; oscillons with only marginally higher energy lie far to the left . as a result , the oscillon lifetime scales exponentially with its energy , as shown in fig . [ lifetimevsenergy ] . here we have used the energy to select oscillons that we predict will decay in a reasonable amount of time . however , we have also verified that more energetic oscillons live for the exponentially long times that this analysis would suggest . in particular , we have followed an individual oscillon for times exceeding @xmath86 in natural units . this analysis also verifies that the decay is not the result of accumulated numerical error . . the vertical axis is the natural logarithm of the lifetime of the oscillon and the horizontal axis is the oscillon energy , both in units where @xmath45 . we see that the oscillon lifetime scales like @xmath87 , where @xmath3 is the oscillon energy . the rate of exponential growth @xmath4 is given in as a function of @xmath88 in the table at right . the values of @xmath4 are roughly linear in @xmath88 , with a slope of approximately @xmath89 . ] c\|c @xmath90 & @xmath4 0.010 & 14.70.013 & 12.90.015 & 10.20.017 & 8.350.020 & 4.97 to track the flow of energy during the oscillon s decay , we consider a fiducial box with a radius of one hubble length @xmath28 . the rate of change in the energy within the box is then @xmath91 where the first term represents the flow of of any outgoing waves through the boundary and the second represents the energy lost to gravitational expansion . for our oscillon configurations , both terms oscillate . on average , however , we find the surprising result that the first term is both positive and much smaller in magnitude than the second term , meaning that the oscillon decays due to energy lost through the expansion rather than through the emission of scalar field waves . finally , we note that although the expansion of the universe explicitly introduces an exponentially large scale the size of the universe it is irrelevant to the oscillon lifetime . the oscillon only sees its local region of space , which is expanding at a rate of order one in natural units , with greater expansion of the universe corresponding to _ shorter _ oscillon lifetimes . we have found that oscillon lifetimes in an expanding background scale exponentially with the oscillon s energy in a simple toy model . the oscillon configurations form generically from an uncorrelated thermal background . as a result , although all the parameters of our theory are of order one in natural units , the dynamics of the theory generates an exponentially large time scale through the oscillon s decay . although our analysis is limited to one dimension , we have also conducted preliminary experiments showing similar behavior when the three - dimensional oscillons of refs . @xcite are coupled to an expanding background in an ansatz with spherical symmetry . work is in progress @xcite to explore the possible consequences of these ideas in the early universe . we thank n. alidoust , e. farhi , a. h. guth , a. scardicchio , r. rosales , and r. stowell for assistance . we also thank j. dunham and m. gleiser for discussions . n. g. and n. s. were supported in part by research corporation through a cottrell college science award . n. g. was also supported in part by the national science foundation through the vermont experimental program to stimulate competitive research ( vt - epscor ) and the middlebury faculty professional development fund . n. s. was also supported in part by the middlebury undergraduate collaborative research fund . m. gleiser , hep - ph/9308279 , phys . * d49 * ( 1994 ) 2978 ; e. j. copeland , m. gleiser and h. r. muller , hep - ph/9503217 , phys . rev . * d52 * ( 1995 ) 1920 ; a. adib , m. gleiser , and c. almeida , hep - th/0203072 , phys . rev . * d66 * ( 2002 ) 085011 . a. m. kosevich and a. s. kovalev , zh . eksp . * 67 * ( 1975 ) 1793 [ sov . jetp * 40 * ( 1975 ) 891 ] ; r. r. rosales , `` weakly nonlinear expansions for breathers , '' unpublished research notes ; j. n. hormuzdiar and s. d. hsu , phys . * c59 * ( 1999 ) 889 .	we consider a classical toy model of a massive scalar field in @xmath0 dimensions with a constant exponential expansion rate of space . the nonlinear theory under consideration supports approximate oscillon solutions , but they eventually decay due to their coupling to the expanding background . although all the parameters of the theory and the oscillon energies are of order one in units of the scalar field mass @xmath1 , the oscillon lifetime is exponentially large in these natural units . for typical values of the parameters , we see oscillon lifetimes scaling approximately as @xmath2 where @xmath3 is the oscillon energy and the constant @xmath4 is on the order of @xmath5 to @xmath6 for expansion rates between @xmath7 and @xmath8 .
the young galactic supernova remnant ( snr ) cassiopeia a ( cas a ) is widely believed to be the result of a core collapse supernova ( sn ) observed by flamsteed in 1680 ( @xcite ) , probably due to the core collapse of an early type wolf - rayet star ( @xcite ) . cas a is classified as an oxygen rich remnant since optical spectroscopic observations ( @xcite ) show the supernova ejecta ( in the form of fast moving knots ) to contain mostly oxygen and oxygen burning products such as sulphur , argon and calcium . a wolf - rayet star progenitor is supported by x - ray data which , when taking into account that most bremsstrahlung is caused by electrons interacting with almost completely ionized oxygen , gives a rather low ejecta mass of @xmath5m@xmath2(@xcite ) . at all wavelengths cas a has the appearance of a broken shell with a radius varying between 1.6 to 2.5 . in this overview of spectral studies of cas a we highlight the the salient features of the xmm - newton x - ray observatory for snr research : ( i ) an outstanding spectral grasp brought about by the large effective collecting area , bandwidth and spectral resolving power of the x - ray telescopes , ( ii ) an unprecedented high energy response which allows high quality x - ray imaging up to 15 kev , ( iii ) an unique capability for high resolution ( r@xmath3100 ) spectroscopy of moderately extended sources . the medium resolution spatially resolved spectroscopy of cas a with the solid state ( _ epic - mos _ and _ epic - pn _ ) x - ray camera s of xmm - newton ( @xcite ) provides a powerful tool to study in detail the dynamics , the distribution and the physical properties of the ( reverse ) shock heated plasma , while the bandwidth ( up to 15 kev ) offers an unique opportunity to assess the origin of the hard x - ray continuum ( @xcite ) . we also report on a recent deep exposure of cas a with the bepposax observatory . this observation gives a reliable estimate of the nuclear de - excitiation lines of @xmath0sc at 67.9 kev and 78.4 kev , produced in the decay chain @xmath0ti@xmath6@xmath0sc@xmath6@xmath0ca . the detection of these lines and the abundance patterns obtained from the _ epic - mos _ analysis provides independant measures of the zero - age - main - sequence ( zams ) progenitor mass . one of the outstanding problems in the study of cas a is the origin of a recently discovered hard x - ray continuum out to 100 kev in the spectrum of this remnant ( e.g. @xcite ; @xcite ; @xcite ) . from a theoretical point of view both synchrotron emission from shock accelerated electrons and non - thermal bremsstrahlung from electrons accelerated into a tail of the maxwellian velocity distribution could be possible explanations for the hard continuum . previous hard x - ray imaging observations , for example using bepposax ( @xcite ) , indicated that the hard continuum radiation originated predominantly in the west region of the remnant , however this result was based on deconvolved images from a typically 1-arcminute resolution telescope with moderate effective area . the combination of large collecting area in the energy band 4.0 to 15.0 kev and the angular resolution of a few arcseconds of xmm - newton provides a unique opportunity to search for the distribution and origin of this hard `` tail '' . xmm - newton clearly detects the hard x - ray tail from the remnant but the hard x - ray image ( 8 - 15 kev ) displays a remarkably similar brightness distribution to the softer thermal components , which presumably dominate the 4 - 6 kev continuum emission as shown in figure [ fig1 ] ( centre and left respectively ) . spectral fits to the cas a data ( @xcite ) indicate that the ratio between a powerlaw and thermal bremsstrahlung components runs from about 1 in the 8 - 10 kev interval to about 3 in the 10 - 15 kev band . a hardness ratio map ( 10 - 15)/(8 - 10 ) is shown in figure [ fig1 ] ( right ) . although this map shows significant changes in hardness over the remnant , all the bright features within the green contour have a remarkably similar spectrum above 8 kev . therefore the hard x - ray image and the hardness ratio indicate clearly that this flux does not predominate in a few localised regions , as would be expected from straightforward synchrotron emission models , e.g. a limb brightened shell commensurate with the radio morphology . apparently the hard x - ray emission pervades the whole remnant with a distribution similar to the thermal continuum emission . in conclusion , while the spectral form of the non - thermal high - energy `` tail '' is not inconsistent with a simple model of synchrotron emission from snrs ( @xcite ) , the morphology of the hard x - ray image does not support this interpretation in any way . an alternative explanation for the observation of hard x - ray tails in the spectra of supernova remnants is the presence of non - thermal bremsstrahlung generated by a population of suprathermal electrons ( @xcite ; @xcite ) . observing the nuclear decay emission associated with @xmath0ti ( see figure [ fig2 ] ) is probably the most direct way of observing the abundance of an element for a young supernova remnant . it does not depend on plasma properties , and temperature , except in the very unlikely case that @xmath0ti is almost completely ionized , which requires too high a temperature ( @xcite ) . if we know the distance to cas a and the decay time of @xmath0ti , which has recently been measured accurately to be @xmath7 yr ( @xcite ) , calculation of the initial @xmath0ti mass is straightforward . compared to other abundantly produced radio - active elements it has a relatively long decay time , which makes this the primary energy source for late time light curves of supernovae ; in sn1987a this occurred after @xmath82000 days , when most of the @xmath9co had decayed ( @xcite ) . its main importance is , however , the diagnostics it provides for type ii supernova explosions . is thought to be synthesized during explosive si burning , as the expanding plasma cools in the presence of sufficiently abundant , free @xmath10 particles , which is the case at relatively low densities ( @xcite;@xcite ) . this process is referred to as @xmath10-rich freeze out . as a consequence , its synthesis is very sensitive to the kinetic energy and asymetries of the explosion ( @xcite ) . moreover , the amount of @xmath0ti ejected depends on the position of the mass - cut , the border that divides the material falling back on the neutron star and that being ejected . also the mass - cut depends both on the energy of the explosion , and the mass of the progenitor ; the expansion velocity is larger for less massive progenitors , which allows more core material to escape the gravitational potential of the neutron star or black hole . the detection of 1157 kev line emission by cgro - comptel ( @xcite ) , was the first time that the presence of @xmath0ti in supernova remnants was actually observed . however , subsequent searches with hard x - ray detectors for the @xmath0sc line emission at 68 kev and 78 kev did not confirm the presence of @xmath0ti at the expected flux level . a deep ( 500 ks ) observation of cas a with bepposax , combined with archival data , has provided us with a net phoswhich detector system ( pds ) spectrum of 311 ks ( the pds efficiency is 50% , as always one of the two rocking collimators is measuring a 210 off source background ) . as reported by @xcite the deep observation resulted in a larger than 3.4@xmath11 detection of the @xmath0sc lines , each with a flux of @xmath12 @xmath13(90% error ) . this compares well with a more recent cgro - comptel measurement of @xmath14 @xmath13 ( @xcite ) . combining the two measurements suggest an initial @xmath0ti mass of @xmath15 m@xmath2 , with a 90% uncertainty range of @xmath16 m@xmath2 . the major uncertainty in measuring the @xmath0sc flux is the contribution of the hard x - ray continuum to the spectrum around 70 kev , due to the uncertain nature of the hard x - ray emission already addressed in the previous paragraph . @xcite employed various models , including a hard x - ray synchrotron model . here we expand on that measurement by using a model spectrum that includes a non - thermal bremsstrahlung continuum caused by electrons accelerated by lower hybrid waves , a process that is likely to be most efficient in the presence of heavy nuclei , which suggest that the hard x - ray emission originates from material heated by the reverse shock ( @xcite ) . note that this model is different from that proposed by @xcite , who suggest that the non - thermal electrons are the non - relativistic part of the electron cosmic ray population . the bremsstrahlung model was calculated using fully relativistic gaunt factors and , apart from the non - thermal continuum , a 3.5 kev thermal component was included and ta k@xmath10 and k@xmath17 line emission , originating in the pds collimator , which consists of tantalum . the best fit non - thermal model has @xmath18 and suggests a maximum electron acceleration of @xmath895 kev ( figure [ fig3 ] ) . as the continuum drops near the @xmath0sc emission the inferred line contribution is larger ( ( @xmath19 @xmath13 ) . this is very close to the latest comptel results , and implies an initial @xmath0ti mass of @xmath20 m@xmath2 . however , the model does not fit the data above 100 kev well . this may not be too much of a problem , since the statistics above 100 kev is poor and the signal may still be contaminated by residual background . however , assuming that most of the hard x - ray emission is indeed caused by electrons accelerated by lower hybrid waves , additional components , such as the one proposed by @xcite , or inverse compton radiation ( see e.g. @xcite ) , are likely to be present at some level . what is the implication of the observed amount of @xmath0ti ? comparing it to the predicted @xmath0ti mass ( @xcite ) , it seems that the actual @xmath0ti mass is higher than expected , except perhaps models s30b , s35c and s40c of @xcite , which model explosions with more than @xmath21 ergs of kinetic energy . apart from energetics , the @xmath0ti results for cas a may be explained by pre - supernova mass loss , as is expected to be the case for type ib sne ( @xcite ) . moreover , an asymmetric explosion would also contribute to a larger @xmath0ti abundance . in fact , from the following sections it will become clear that the cas a explosion indeed displays a high degree of asymmetry . less certain is whether cas a may also have been a much more energetic explosion . the brightness of cas a coupled to a long exposure of xmm - newton s solid state cameras makes it possible to do a detailed x - ray spectral analysis with an angular resolution of 20 arcseconds over the the full extent of the remnant . the energy resolution , gain stability and gain uniformity of the _ epic - mos_-cameras allows significant detection of emission line energy shifts of order 1 ev or greater for prominent emission lines like si - k , s - k and fe - k . proper modeling of these line blends with the aid of broad band spectral fitting , taking into account the non - equilibrium ionisation ( nei ) balance , allows an assessment of doppler shifts and abundance variations of the x - ray emitting material across the face of the remnant with adequate spatial resolving power . two nei components for the thermal emission were used as a minimum for representative spectral modeling . in addition the model accounts for two separate redshifts ( one for each plasma component ) and the amount of foreground absorption as free parameters . figure [ fig4 ] shows a typical spectral fit . the temperature distribution of the 2 nei components are similar ( but not the same ) . the temperature for the cool component ranges between 0.3 - 1 kev , displayed in figure [ fig5 ] , and those for the hot component between 2 - 6 kev . these values are similar to values obtained by @xcite for a coarse subdivision of the remnant in five regions . a map of the ionisation age of the cool component is also given in figure [ fig5 ] and shows a large spread of @xmath22@xmath23s . for ionisation ages larger than @xmath24 @xmath23s the plasma is almost in ionisation equilibrium and can not be distinguished from equilibrium spectra . there is also a region of very low ionisation age ( less than @xmath25 @xmath23s ) stretching from east to west just above the centre of the remnant . this region also has a low emissivity and can be understood as a low density wake just behind and inside of the shocked ejecta . the hot component has a more homogeneous ionisation age distribution centered around @xmath26@xmath23s ( not shown ) . figure [ fig6 ] is a montage of abundance maps of ne , mg , si , s , ar , ca , fe - l , fe - k and ni , where the abundance values are with respect to solar . again we see considerable variations over the remnant . the fe - l distribution comes from the cool component while the fe - k and ni are derived exclusively from the hot component . the distributions of si , s , ar and ca , which are all oxygen burning products , are similar and distinct from carbon burning products , ne and mg , and fe - l . figure [ fig7 ] shows the variation in the ratios s / si , ar / si and ca / si with respect to the abundance of si . on the one hand these ratios clearly vary over the remnant but on the other hand , for a si abundance range spanning more than two orders of magnitude , these ratios remain remarkably constant . the thick vertical bars indicate the mean and rms scatter of the ratio values . the models for nucleosynthesis yield from massive stars predict that the mass or abundance ratio @xmath27 of ejected mass of any element x with respect to silicon varies significantly as a function of the progenitor mass @xmath28 . we show the observed mean values of @xmath27 as well as its rms variation in table [ tab1 ] , together with the predictions for core collapse sne models with a progenitor mass of 12 , 30 m@xmath2 ( zams ) and for a type ib model with a progenitor mass of 60 m@xmath2 ( zams ) . the model for the 60 m@xmath2 star included mass loss , in order to mimic a type ib supernova . abundance values for the 12 and 30 m@xmath2 progenitors are taken from @xcite , while those for the 60 m@xmath2 progenitor are taken from @xcite . .mean measured abundance ratios and rms scatter compared with theoretical predictions for progenitor masses of 12 , 30 and 60 @xmath29 ( including windloss ) . [ cols="<,<,<,<,<,<",options="header " , ] [ tab1 ] the observed abundance ratios for x equal to ne , mg , s , ar , ca and fe - l ( related to the iron emissivity of the cool component ) are consistent with a low progenitor mass of type ii 12 m@xmath2 model , but the observed o ratio is an order of magnitude larger than predicted . the type ii 30 m@xmath2 model predicts the correct o ratio , but does not produce the observed ratios observed in the other elements . the best overall agreement of observed abundance ratios are provided by the type ib 60 m@xmath2 model . in all these models , most of the si , s , ar and ca comes from the zones where complete explosive o - burning and incomplete explosive si - burning occurs , and indeed as figure [ fig7 ] shows these elements track each other remarkably well . the correlation between si and s is remarkably sharp but not perfect , the scatter in figure [ fig7 ] is real . these remaining residuals can be attributed to small inhomogeneities . table [ tab1 ] also indicate that the rms scatter of the abundances with respect to si get larger as z increases , s to ar to ca and indeed through to fe and ni . this is to be expected since elements close together in z are produced in the same layers within the shock collapse structure while those of very different z are produced in different layers and at different temperatures . the fe which arises from complete and incomplete explosive si burning , together with existing fe , should give rise to iron line emission . for both the fe - l and fe - k lines we see that iron abundance varies over the remnant but does not show any straightforward correlation with either the other elements ( there is a very large scatter in @xmath30 ) or relative to each other . this is to be expected if most of the iron arises from complete si burning . we return to the different morphologies of si and fe later in the discussion . ne and mg are mostly produced in shells where ne / c burning occurs , and the relative scatter in terms of @xmath27 is indeed much larger than for s , ar and ca ( table [ tab1 ] ) . furthermore the abundance maps of ne and mg in figure [ fig6 ] are mutually similar but very different from the si , s , ar and ca group . it should be noted that the derivation of o / si ratio is difficult and less reliable than the other ratio s due to the strong galactic absorption and the poor spectral resolution of the epic camera s at low energies . the high resolution xmm rgs maps resolve the o distribution much more clearly and show that this differs from the other elements ( @xcite ) . we discuss this briefly in section 5 . it is possible to determine the doppler shifts of si - k , s - k and fe - k since these lines are strong and well resolved . doppler shifts of these lines have been calculated by taking advantage of the ionisation age information supplied by the model fits . after fits were made to the full spectrum we froze all the fit parameters . we selected the si - k ( 1.72 - 1.96 kev ) , s - k ( 2.29 - 2.58 kev ) and fe - k ( 6.20 - 6.92 kev ) bands for determining their respective doppler velocities while ignoring all other line emission . we then do a fit to each line separately by starting from the full fit model parameters as a template and subsequently allowing only the redshift and the abundance of the relevant element to vary . this method provides a fine tuning of the redshift which in turn gives the doppler velocity of the element under scrutiny . figure [ fig8 ] shows the resulting line flux images colour coded with the doppler velocity . the bottom left image is the colour coding used . the doppler shifts seen in different areas of the remnant are very similar in the three lines . the knots in the south east are blue shifted and the knots in the north are red shifted . this is consistent with previous measurements ( @xcite , @xcite , @xcite ) . moving from large radii towards the centre the shift generally gets larger as expected in projection . this is particularly pronounced in the north . at the outer edges the knots are stationary or slightly blue shifted . moving south a region of red shift is reached indicating these inner knots are on the far side of the remnant moving away from us . the distributions of flux as a function of doppler velocity are shown in figure [ fig9 ] . the distributions for si - k and s - k are very similar . the fe - k clearly has a slightly broader distribution for the red shifted ( + ve ) velocities . this plot was quite sensitive to small systematic changes in temperature , ionisation age or abundances in the spectral fitting since these can potentially have a profound effect on the derived doppler velocities . this emphasises the point that detailed modeling is an essential prerequisite to arrive at proper conclusions . the x - ray knots of cas a form a ring because the emitting plasma is confined to an irregular shell . @xcite calculated a best fit centre to this ring looking for the position that gave the most strongly peaked radial brightness distribution ( minimum rms scatter of flux about the mean radius ) . the best centre for the combined si - k , s - k and fe - k line image was 13 arc second west and 11 arc seconds north of the image centre ( the central chandra point source ) . using this centre the peak flux occurred at a radius of 102 arc secs , the mean radius was 97 arc secs and the rms scatter about the mean radius was 24 arc seconds . the left - hand panel of figure [ fig10 ] is a composite image of the remnant seen in the si - k , s - k and fe - k emission lines . the solid circle indicates the @xmath31 arc seconds and the dashed circle is the mean radius of the si - k and s - k flux @xmath32 arc seconds . the x - ray image of the remnant provides coordinates x - y in the plane of the sky . using the derived radial velocity field within the remnant we can use the measured doppler velocities @xmath33 to give us an estimate of the z coordinate position of the emitting material along the line of sight thus giving us an x - y - z coordinate for the emission line flux in each pixel . using these coordinates we can reproject the flux into any plane we choose . the right - hand panel of figure [ fig10 ] shows such a projection in a plane containing the line - of - sight , north upwards , observer to the right . arc seconds and the inner circle is the mean radius of the si - k and s - k flux . ] in this reprojection the line emission from si - k , s - k and fe - k are reasonably well aligned for the main ring of knots . the reprojection is not perfect because the _ epic - mos _ cameras are unable to resolve components which overlap along the line - of - sight and this produces some ghosting just north of the centre of the remnant . in the plane of the sky fe - k emission ( blue ) is clearly visible to the east between the mean radius of the si+s flux and the shock radius . similarly in the reprojection fe - k emission is seen outside the main ring in the north away from the observer . the si+s knot in the south away from the observer in the reprojection is formed from low surface brightens emission in the south west quadrant of the sky image . the x - ray emitting material is very clumpy within the spherical volume and is indeed surprisingly well characterised by the doughnut shape suggested by @xcite . however the distribution is distinctly different to that obtained in similar 3-d studies of the optical knots , lawrence et al . ( 1995 ) . the plane of the sky image in figure [ fig10 ] shows that the fe - k emission to the east is at a radius of 140 arc seconds , near the primary shock , with an implied shock velocity of @xmath34 in the range 3500 - 4500 km s@xmath35 . the fe - k emission at large radii is highly reminiscent of snr shrapnel discovered by @xcite around the vela snr . these are almost certainly bullets of material which were ejected from the progenitor during the collapse and subsequent explosion . they would initially be expected to have a radial velocity less than the blast wave but as the remnant develops , and the shock wave is slowed by interaction with the surrounding medium , the bullets would overtake the blast wave and appear outside the visible shock front as is the case in vela . our analysis of the abundances clearly indicates that the matter responsible for the x - ray emission is ejecta and this must have been ablated from the bullets rather than swept up by the shock . figure [ fig11 ] displays spectra from three extraction regions over the remnant situated in the n , ne and se of the remnant . the data from rgs1 and rgs2 have been combined in order to increase the statistical weight of the observation . several lines of highly ionized species of si , mg , ne , fe l and o are detected in the spectrum . due to interstellar absorption no features can be measured long - ward of @xmath820 . the analysis and interpretation of these spectra will be the subject of a forthcoming paper . as a first result @xcite extracted images of the ly-@xmath10 and ly-@xmath17 lines to probe small scale variations in absorption effects over this part of the remnant and to investigate the potential presence of resonance scattering in the limb brightened shells viewed edge - on . the temperature range relevant for cas a has no influence on the ly-@xmath10/ly-@xmath17 ratio . since rgs is a slitless spectrometer it is possible to extract dispersed monochromatic images of cas a. these images were converted from wavelength to spatial coordinates using the equation @xmath36 , where @xmath37 is the offset along the dispersion direction , @xmath38 the wavelength shift , @xmath10 the angle of incidence on the gratings and @xmath39 the line distance of the gratings . however , any doppler broadening is also convolved along the dispersion direction and , depending on its magnitude , doppler broadening could distort the rgs images . the ly-@xmath10 and ly-@xmath17 maps are presented in figure [ fig12 ] . the oxygen emission is seen to originate from three blobs on the e and n rim of the remnant . the emission is faintest in the se blob where there also seems to be a ly-@xmath10 deficit compared to ly-@xmath17 . such a decrease in the ly-@xmath10/ly-@xmath17 ratio could possibly be introduced by the presence of resonance scattering of the ly-@xmath10 photons in the x - ray bright rims if viewed edge - on ( @xcite ) . the hard x - ray continuum maps from xmm - newton indicate that the [email protected] kev flux , predominantly due to the previously reported high - energy spectral tail , does not originate from a few localized regions such as x - ray bright knots and filaments , nor does it originate from a limb brightened ( fractionary ) shell close to the shock front generated by the primary blast wave . therefore the hard x - ray image does not support in any way the interpretation in terms of simple synchrotron emission models . non - thermal bremsstrahlung , produced by a population of suprathermal electrons , would seem to be a plausible alternative . the detection of the @xmath0sc(@xmath0ti ) emission in the bepposax data can be reconciled with a progenitor star having a range 30 - 40 @xmath41 , assuming a very energetic explosion . it is worth noting , however that the relatively high @xmath0ti abundance can also be explained by one of the few available sn type ib models of a star with a zams mass of 60 m@xmath2 ( @xcite ) . moreover the elemental abundance ratios predicted by this particular model are also roughly consistent with the abundance ratios obtained from the xmm - newton analysis . the tight correlation between the variation in abundance of si , s , ar , ca over an absolute abundance range of two orders of magnitude is strong evidence for the nucleosynthesis of these ejecta elements by explosive o - burning and incomplete explosive si - burning due to the shock heating of these layers in the core collapse supernova . however the fe emission , both in the fe - k and the fe - l lines , does not show this correlation in any sense . a significant fraction of the fe - k emission is seen at larger radii than si - k and s - k as convincingly demonstrated in our doppler derived 3-d reprojection . moreover the fe - k emission is patchy , reminiscent of large clumps of ejecta material , rather than shock heated swept up circumstellar material . the x - ray emission would then arise from ablation of the fe - ejecta bullets . the 3-d reprojection also shows a largely bi - polar distribution of the fe - k emission which may indicate that the original explosion was aspherical , possibly with axial symmetry .	in this paper we present recent image and spectral data of cas - a obtained with xmm - newton and bepposax , which were used to further constrain the physical characteristics of the source . we analysed the hard x - ray continuum images between 8 and 15 kev . the data indicate that the hard x - ray tail , observed previously , does not originate in localised regions as demonstrated by the rather flat hardness ratio map of the 10 - 15 and 8 - 10 kev energy bands . this result does not support an interpretation of the hard x - radiation as synchrotron emission produced in the primary shock , non - thermal bremsstrahlung is a plausible alternative . moreover , a recent 500 kilosecond deep observation of the hard x - ray continuum by bepposax has revealed the positive detection of the @xmath0sc nuclear decay lines at 67.9 and 78.4 kev . appropriate modelling of the hard x - ray continuum leads to an estimate of the initial @xmath0ti mass of @xmath1 m@xmath2 . this implies , by employing available nucleosynthesis models for type ii sne , a progenitor mass of @xmath330 m@xmath2 . we also present a , spatially resolved , spectral analysis of the thermal emission of cas - a using detailed spectral fitting on a 20@xmath420pixel scale . this yields maps of the ionisation age , temperature , abundances for ne , mg , si , s , ar , ca , fe and ni . the observed elemental abundance ratio patterns can be reconciled best with a type ib sne , arising from a 60 m@xmath2 progenitor ( zams ) with high - mass - loss , which may also explain the relatively high @xmath0ti yield . the accurate modeling of the image - resolved emission line spectra has enabled us to derive reliable doppler velocities for the bright si - k , s - k , and fe - k line complexes . the combination of radial positions in the plane of the sky and the line of sight velocities have been used to assess the dynamics of the x - ray emitting plasma . this indicates a rather asymmetric explosion geometry , in which the hot fe - k emitting gas possibly originates from shock - heated ablated ejecta material , rather than from circumstellar plasma heated by the primary blast wave .
we consider the _ equation _ @xmath4 on the circle @xmath5 with @xmath1 real - valued . as is well known , the equation can be written as an infinite - dimensional hamiltonian system @xmath6 where we choose the standard sobolev space @xmath7 , @xmath3 , for the phase space , with the @xmath8-inner product @xmath9 and endow this space with the poisson structure proposed by gardner @xmath10 the mean value @xmath11 $ ] is a casimir for this bracket and hence is preserved by the flow . on the invariant subspace @xmath12 = 0},\ ] ] the poisson structure is nondegenerate and the evolution is given by @xmath13 as an infinite - dimensional hamiltonian system , the equation is well known to be completely integrable . according to @xcite , this is true in the strongest possible sense meaning that admits _ global birkhoff coordinates _ @xmath14 . to give a precise statement of our results , we define for @xmath15 in @xmath16 the sobolev norm @xmath17 by @xmath18 further , we introduce the model space @xmath19 with elements @xmath20 where the norm @xmath21 is defined by @xmath22 this space is endowed with the canonical poisson structure @xmath23 while all other brackets vanish . the _ birkhoff map _ @xmath24 is a bi - analytic , canonical diffeomorphism @xmath25 , whose restriction @xmath26 is again bi - analytic for any integer @xmath27 . on @xmath28 the transformed hamiltonian @xmath29 is a real - analytic function of the actions @xmath30 alone and the equations of motion take the particular simple form @xmath31 the mapping @xmath32 may thus be viewed as a nonlinear fourier transform for the equation . furthermore , the derivative @xmath33 of @xmath32 at the origin _ is _ a weighted fourier transform , and on @xmath34 @xmath35 which we refer to as parseval s identity cf . e.g. @xcite . our main result says that for higher order sobolev norms the following version of parseval s identity holds for the nonlinear map @xmath32 . [ b - est ] for any integer @xmath27 there exist absolute constants @xmath36 , @xmath37 such that the restriction of @xmath32 to @xmath16 satisfies the two - sided estimates @xmath38 and @xmath39 the estimates ( i ) and ( ii ) are reminiscent of the 1-smoothing property of the birkhoff map @xmath32 established in @xcite as they are linear in the highest sobolev norm @xmath17 and the highest weighted @xmath40-norm @xmath41 , respectively . [ s - est ] suppose @xmath42 is a solution of the equation with initial value @xmath43 , @xmath44 . then there exists an absolute constant @xmath45 such that for all time @xmath46 , @xmath47 the proof of theorem [ b - est ] relies on estimates for the action variables @xmath48 where @xmath49 , @xmath50 . the decay properties of the actions are closely related to the regularity properties of @xmath1 cf . we quantify this relationship by providing two - sided estimates of the sobolev norms of @xmath1 in terms of weighted @xmath51-norms of @xmath52 . for that purpose introduce for @xmath53 real the weighted sequence space @xmath54 whose norm is defined by @xmath55 [ act - sob - est ] for any integer @xmath27 there exist absolute constants @xmath56 , @xmath37 , such that @xmath57 as well as @xmath58 for all @xmath59 . note that in the estimates ( i ) and ( ii ) of theorem [ act - sob - est ] the corresponding exponents are of the same order since the action variables have to be viewed as quadratic quantities . for the direct problem , that is estimates ( i ) of theorems [ b - est ] & [ act - sob - est ] , we obtain estimates which hold for a larger family of spaces referred to as _ weighted sobolev spaces _ see @xcite for an introduction . a _ normalized , symmetric , submultiplicative _ , and _ monotone weight _ is a function @xmath60 with @xmath61 for all @xmath62 . the class of all such weights is denoted by @xmath63 and @xmath64 is the space of @xmath34 functions @xmath1 with finite @xmath65-norm @xmath66 further , @xmath67 denotes the elements @xmath68 with @xmath69 , where @xmath70 for any @xmath71 , the _ sobolev weight _ @xmath72 gives rise to the usual sobolev space @xmath73 . for @xmath74 and @xmath75 , the _ abel weight _ @xmath76 gives rise to the space @xmath77 of @xmath78-functions , which can be analytically extended to the open strip @xmath79 of the complex plane with traces in @xmath80 on the boundary lines . in between are , among others , the _ gevrey weights _ @xmath81 which give rise to the gevrey spaces @xmath82 , as well as weights of the form @xmath83 that are lighter than abel weights but heavier than gevrey weights . we assume all weights @xmath84 to be piecewise linearly extended to functions on the real line @xmath85 , @xmath86 . [ act - west ] for any weight @xmath87 with @xmath88 and @xmath89 there exists an absolute constant @xmath90 such that the restriction of the birkhoff map @xmath32 to @xmath91 takes values in @xmath92 and satisfies @xmath93 moreover , we find for the action variables on @xmath91 @xmath94 the bounds given in ( i ) and ( ii ) are valid for all submultiplicative weights including those growing exponentially fast . they reflect the nature of the weight : if , for example , @xmath65 grows polynomially , then the bounds ( i ) and ( ii ) are polynomial in @xmath95 , whereas if @xmath65 grows exponentially , then the bounds are exponential in @xmath95 . note that the bounds improve as soon as the weight @xmath65 incorporates the factor @xmath96 with @xmath97 . the following version of theorem [ act - west ] for sobolev spaces of real exponent complements the results of theorems [ b - est]-[act - sob - est ] . for any real @xmath74 there exist an absolute constant @xmath98 such that the restriction of the birkhoff map @xmath32 to @xmath73 takes values in @xmath99 and satisfies @xmath100 moreover , we find for the action variables on @xmath101 @xmath102 _ method of proof . _ the main ingredient into the proof of theorem [ b - est ] is the estimate of the action variables @xmath103 , @xmath50 , of theorem [ act - sob - est ] . to establish the latter , we use the action variables @xmath104 , @xmath50 , on level @xmath3 introduced by mckean & vaninsky @xcite in context of the nonlinear schrdinger ( nls ) equation . they are defined in terms of spectral data of the corresponding hill operator @xmath105 arising in the lax - pair formulation of see section [ s : setup ] . for @xmath106 , the actions @xmath107 coincide with @xmath103 , whereas for @xmath27 they have the asymptotic behavior @xmath108 furthermore , they satisfy the trace formula @xmath109 where @xmath110 denotes the @xmath111th hamiltonian in the _ hierarchy_. the first two hamiltonians of this hierarchy are given by @xmath112 more generally , these hamiltonians have the form @xmath113 with @xmath114 being a canonically - determined polynomial . taking the asymptotic behavior of the actions @xmath104 , the corresponding trace formulae , and the representation of the hamiltonians together we get @xmath115 which yields the bound stated in theorem [ act - sob - est ] ( i ) . to obtain item ( ii ) of theorem [ act - sob - est ] we rewrite the @xmath116th hamiltonian as @xmath117 using the trace formula we derive a bound of @xmath118 in terms of @xmath119 . since the @xmath120-term depends only on derivatives of @xmath1 up to order @xmath121 , an inductive argument then gives part ( ii ) of theorem [ act - sob - est ] . we remark that the asymptotics of @xmath104 are derived from sufficiently accurate asymptotics of the periodic eigenvalues of the corresponding hill operator , and that our method of proof does not involve any auxiliary spectral quantities such as the spectral heights . to prove theorem [ act - west ] , we take a slightly different approach by estimating the action variables in terms of the spacing of the periodic eigenvalues of the associated hill operator . for the latter , estimates in any weighted norm are available see e.g. @xcite allowing us to obtain theorem [ act - west ] . _ related results . _ first results on global bounds of the actions in terms of sobolev norms were obtained by korotyaev @xcite using conformal mapping theory . more precisely , he shows ( * ? ? ? * theorem 2.4 & 2.6 ) for any @xmath3 that @xmath122 is bounded by @xmath123{1+\n{\partial_{x}^{m}u}_{0}^{\frac{2m}{m+2}}(1+\n{\partial_{x}^{m}u}_{0})^{\frac{m(4m+2)}{3(m+2)}}}.\end{aligned}\ ] ] by augmenting @xmath36 but without increasing the degree of the right hand side , this bound may be simplified to @xmath124 in comparison , our estimate theorem [ act - sob - est ] ( i ) is linear in @xmath125 with the remainder @xmath126 involving only sobolev norms of order @xmath121 . for the inverse problem korotyaev obtains ( * ? ? ? * theorem 2.4 & 2.6 ) @xmath127 where @xmath128 note that @xmath129 grows factorially with @xmath116 . our estimate theorem [ act - sob - est ] ( ii ) considerably improves the latter one , since the bound is linear in @xmath130 , the exponent of the remainder @xmath131 is linear in @xmath116 , and the remainder only involves @xmath51-norms of order @xmath132 . a priori estimates of solutions obtained directly from the hamiltonians in the hierarchy are well known see e.g. @xcite . the application of theorem [ b - est ] to the initial value problem of stated in corollary [ s - est ] improves on these results since the bound of @xmath133 is linear in @xmath134 and the exponent of the remainder @xmath135 grows only quadratically in @xmath116 . first results on bounds of the birkhoff map for on the weighted sobolev spaces @xmath91 can be found in kappeler & pschel @xcite . the authors proved @xmath136 where the @xmath137 part holds true only for weights with subexponential growth . the estimates of @xmath138 presented in theorem [ act - west ] quantify this relationship . the methods developed in this paper were introduced by the author @xcite in the context of the nonlinear fourier transform for the periodic defocusing equation @xmath139 to show results corresponding to theorems [ b - est]-[act - west ] . we note that the result of theorem [ b - est ] improves on the corresponding result obtained for the defocusing nls equation since the exponents in the remainders of estimate ( i ) and estimate ( ii ) are equal . from a broader perspective we may view the weighted actions @xmath140 as a perturbation of the squared modulus @xmath141 of the @xmath142th fourier coefficient of @xmath1 . our method of comparing the norms @xmath122 with the hamiltonians of the hierarchy consists in a separate analysis of fourier modes of low and high frequencies . this idea has a long history in the analysis of nonlinear pdes . most recently , it lead colliander , keel , staffilani , takaoka & tao @xcite to invent the i - method , which allows to obtain global well - posedness of subcritical equations in low - regularity regimes where the hamiltonian ( or other integrals ) of the equation cease to be well defined . the idea is to damp all sufficiently high fourier modes of a local solution such that the hamiltonian can be controlled by weaker norms while still being an > > almost conserved < < quantity . the difficulty here is to choose the damping carefully enough such that the nonlinearity of the equation does not create a significant interaction of low and and high frequencies . our aim is so to say opposite to that of the i - method : as we look for quantitative global estimates , the most delicate part of our analysis is to get control of the modes of low frequencies . it is achieved by an appropriate localization of the periodic eigenvalues of the hill operator see proposition [ ev - as ] and the references mentioned in appendix a. the results on the analyticity and asymptotic expansion of the integral @xmath143 introduced in section [ s : setup ] have also been applied in @xcite concerning convexity properties of the hamiltonian . they were put together in a joint effort . _ organization of the paper . _ in section [ s : setup ] the action variables on integer levels @xmath3 are defined and the trace formulae relating them to the hierarchy of hamiltonians are proven . in sections [ s : act - est ] and [ s : sob - est ] theorem [ act - sob - est ] ( i ) and ( ii ) , respectively , are obtained by use of the localization of the hill spectrum , which for the convenience of the reader is proved in appendix [ s : appendix ] . finally , in section [ s : act - west ] we prove theorem [ act - west ] by obtaining a uniform estimate of the actions in terms of the spacing of the periodic eigenvalues . _ acknowledgement . _ the author is very grateful to professor thomas kappeler for frequent discussions and valuable comments that greatly improved the manuscript . in this section we briefly recall the definition of the action variables as well as the main properties of the spectral quantities used to define them . we follow the exposition @xcite see also @xcite . consider _ hill s operator _ @xmath144 on the interval @xmath145 $ ] , endowed with periodic boundary conditions and @xmath146 being a _ complex potential _ in @xmath147 . the spectrum of @xmath148 , called the _ periodic spectrum of @xmath146 _ , is pure point and complex in general as the operator is not self - adjoint . we may order the eigenvalues lexicographically first by their real part and second by their imaginary part such that @xmath149 their asymptotic behavior is @xmath150 and we define for any @xmath50 the _ gap length _ @xmath151 here @xmath152 stands for an @xmath153-sequence . for convenience we set @xmath154 . to obtain a suitable characterization of the periodic spectrum of @xmath146 , we denote by @xmath155 and @xmath156 the standard fundamental solutions of @xmath157 , and by @xmath158 the _ discriminant _ @xmath159 to simplify matters , we may drop some or all of its arguments from the notation whenever there is no danger of confusion . the periodic spectrum of @xmath146 is precisely the zero set of the entire function @xmath160 , and we have the product representation @xmath161 hence , the discriminant is uniquely determined by the periodic spectrum . we also need the @xmath162-derivative @xmath163 whose zeros are denoted by @xmath164 and satisfy @xmath165 . this derivative has the product representation @xmath166 for each real potential @xmath146 there exists an open neighborhood @xmath167 within @xmath168 such that for every @xmath169 the closed intervals @xmath170,\quad n\ge 1,\ ] ] are disjoint from each other . even more , there exist mutually disjoint neighborhoods @xmath171 , @xmath172 , called _ isolating neighborhoods _ , which satisfy : 1 . @xmath173 and @xmath164 are contained in the interior of @xmath174 for every @xmath169 , 2 . there exists a constant @xmath175 such that for @xmath176 , @xmath177 3 . @xmath178 for @xmath142 sufficiently large . throughout this text @xmath167 denotes a neighborhood of @xmath146 such that a common set of isolating neighborhoods for all @xmath169 exists . the union of all @xmath167 for @xmath146 real defines an open and connected neighborhood of @xmath179 within @xmath168 and is denoted by @xmath180 . following the approach of flaschka & mclaughlin @xcite , one can define action variables for the equation by arnold s formula @xmath181 here @xmath182 denotes a cycle around @xmath183 on the spectral curve @xmath184 on which the square root @xmath185 is defined . this curve is another spectral invariant associated with @xmath146 , and an open riemann surface of infinite genus if and only if the periodic spectrum of @xmath146 is simple . to avoid the technicalities involved with this curve , we fix proper branches of the square root which allows us to reduce the definition of the actions to standard contour integrals in the complex plane see also @xcite . denote by @xmath186{\phantom{\lm}}$ ] the _ principal branch _ of the square root on the complex plane minus the ray @xmath187 $ ] . furthermore , for @xmath188 the _ standard root _ @xmath189{(\lm_{n}^{+}-\lm)(\lm_{n}^{-}-\lm ) } , \qquad \lm\notin g_{n},\qquad n\ge 1,\ ] ] is defined by the condition @xmath190{1 - \gm_{n}^{2}/4(\tau_{n}-\lm)^{2 } } , \qquad \tau_{n } = ( \lm_{n}^{-}+\lm_{n}^{+})/2.\end{aligned}\ ] ] the standard root is analytic in @xmath162 on @xmath191 and in @xmath192 on @xmath193 . finally , we define the _ canonical root _ @xmath194{\dl^{2}(\lm)-4 } \defl -2\ii\sqrt[+]{\lm-\lm_{0}^{+}}\prod_{k\ge 1 } \frac{\vs_{k}(\lm)}{k^{2}\pi^{2}}.\ ] ] this root is analytic in @xmath162 on @xmath195 and in @xmath192 on @xmath196 . @xmath142th action variable _ of @xmath197 is then given by @xmath198{\dl^2(\lm)-4}}\,\dlm,\ ] ] where @xmath199 denotes any sufficiently close circuit around @xmath173 . more generally , we define the _ @xmath142th action variable on level @xmath3 _ by @xmath200{\dl^{2}(\lm)-4}}\,\dlm.\ ] ] note that the action on level zero @xmath107 equals the action @xmath103 . we proceed with the analysis of the analytical properties of the action integrand . to this end , we define for any @xmath197 on @xmath201 the complex 1-form @xmath202 by @xmath203{\dl^{2}(\lm)-4}}\,\dlm = \frac{1}{2\ii}\frac{1}{\sqrt[+]{\lm-\lm_{0}^{+ } } } \prod_{k\ge 1 } \frac{\lm_{k}^{{\text{\tiny\ensuremath{\bullet}}}}-\lm}{\vs_{k}(\lm)}\,\dlm.\end{aligned}\ ] ] a path in the complex plane is said to be _ admissible _ for @xmath146 if , except possibly at its endpoints , it does not intersect any non collapsed gap @xmath204 . [ w - closed ] for each @xmath197 , the 1-form @xmath205 has the following properties : @xmath205 is analytic on @xmath206 , @xmath207 is analytic in @xmath162 on @xmath208 . ] , and for @xmath50 and any admissible path from @xmath209 to @xmath210 in @xmath174 , @xmath211 in particular , for any closed circuit @xmath199 in @xmath174 around @xmath173 , @xmath212 \(i ) @xmath213 is analytic on @xmath214 and the canonical root is analytic on @xmath206 and does not vanish there . therefore , @xmath205 is analytic on @xmath206 . \(ii ) similarly , for @xmath146 fixed , @xmath207 is analytic in @xmath162 on @xmath215 . suppose @xmath216 for some @xmath50 , then @xmath217 is a double root of @xmath218 , hence @xmath219 and the term @xmath220 in the product representation of @xmath205 equals @xmath221 , hence @xmath207 is analytic on @xmath174 . \(iii ) we first consider the case of @xmath146 being real - valued . clearly , the functional @xmath222 , @xmath223 is analytic . further , for any @xmath169 real - valued , one has @xmath224 on @xmath173 so , after deforming the contour of integration to @xmath173 , according to the definition of the canonical root ( see @xcite for a sign table ) @xmath225{\dl^{2}(\lm)-4}}\,\dlm = -2\operatorname{arccosh}\frac{(-1)^{n}\dl(\lm)}{2}\bigg\|_{\lm_{n}^{-}}^{\lm_{n}^{+ } } = 0.\ ] ] thus @xmath226 vanishes on @xmath227 and hence on all of @xmath167 by lemma [ ana - vanish ] . in view of @xmath228 we conclude that @xmath229 for any closed circuit @xmath199 in @xmath174 around @xmath173 and any @xmath197 . now fix @xmath197 arbitrary . the identity @xmath230 clearly holds in the case @xmath217 . if @xmath231 , then we may define the left hand side @xmath232 and the right hand side @xmath233 of @xmath173 by @xmath234 the canonical root admits opposite signs on @xmath235 that is @xmath194{\dl^{2}(\lm)-4}\bigg\|_{g_{n}^{- } } = -\sqrt[c]{\dl^{2}(\lm)-4}\bigg\|_{g_{n}^{+}}.\ ] ] defining the contour @xmath199 by going from @xmath209 to @xmath210 along @xmath233 and then going back from @xmath210 to @xmath209 along @xmath232 gives @xmath236{\dl^{2}(\lm)-4}\big\|_{g_{n}^{-}}}\,\dlm = -2\int_{\lm_{n}^{-}}^{\lm_{n}^{+ } } \frac{{\dl^{{\text{\tiny\ensuremath{\bullet}}}}}(\lm)}{\sqrt[c]{\dl^{2}(\lm)-4}\big\|_{g_{n}^{+}}}\,\dlm.\ ] ] by contour deformation it then follows that @xmath230 along any admissible path in @xmath174 . writing the action variables as @xmath237 makes the claimed analyticity on @xmath180 evident . to proceed , we define for any @xmath197 and any @xmath172 on @xmath238 the primitive @xmath239 by @xmath240 clearly , this improper integral exists , as @xmath205 has the integrable singularity @xmath241 near @xmath210 . by lemma [ w - closed ] the integral is also independent of the chosen admissible path . [ f - prop ] for every @xmath197 and every @xmath172 , we have that @xmath242 is analytic in @xmath192 on @xmath243 and @xmath244 is analytic in @xmath162 on @xmath208 . @xmath245 on @xmath208 . in particular , @xmath246 extends continuously to all points @xmath247 and @xmath248 , @xmath50 . one has @xmath249 and @xmath250 @xmath251 is analytic on @xmath252 for every @xmath172 . if @xmath146 is real , then for any @xmath253 and any real @xmath254 , @xmath255 clearly , @xmath256 is continuous on @xmath173 , strictly positive on @xmath183 , and vanishes at the boundary points . at the zero potential one has @xmath257{\lm}+\ii n\pi$ ] . \(i ) let us first consider the case where @xmath258 . since @xmath205 is analytic on @xmath259 , it suffices to show that @xmath260 is analytic on @xmath167 for some fixed @xmath261 near @xmath262 . choose the straight line @xmath263 $ ] as the path of integration with parametrization @xmath264 then by the product representation of @xmath205 @xmath265{\mu-\lm_{0}^{+}}}{\sqrt[+]{t } } \chi_{0}(\lm_{t})\,\dt,\qquad \chi_{0}(\lm ) = \prod_{k\ge 1 } \frac{\lm_{k}^{{\text{\tiny\ensuremath{\bullet}}}}-\lm}{\vs_{k}(\lm)}.\ ] ] since @xmath247 is analytic on @xmath180 and @xmath266 is analytic on @xmath267 cf . * appendix a ) and ( * ? ? ? * corollary 12.8 ) the claimed analyticity of @xmath246 follows . in lemma [ f - ana ] we obtain the analyticity of @xmath242 , @xmath50 . this proof is a bit more work due to the fact that in this case @xmath210 may be a double eigenvalue . \(ii ) note that @xmath268 . clearly , @xmath249 and @xmath269 for any @xmath50 by lemma [ w - closed ] , hence @xmath270 and @xmath271 to proceed , first consider the case where @xmath146 is real - valued . in this case @xmath272{\dl^{2}(\lm)-4 } > 0 $ ] for @xmath273 see ( * ? ? ? * section 5 ) so @xmath274{4-\dl^2(\lm)}}\,\dlm = \ii(-1)^{k } { \sin^{-1}}\frac{\dl(\lm)}{2}\bigg\|_{\lm_{k}^+}^{\lm_{k+1}^- } = -\ii \pi,\end{aligned}\ ] ] and hence @xmath275 . the function @xmath276 is analytic on @xmath167 by item ( i ) , therefore , @xmath275 holds true on all of @xmath167 in view of lemma [ ana - vanish ] . \(iii ) in view of item ( i ) it remains to show that @xmath277 admits also for @xmath278 an analytic extension from @xmath279 to all of @xmath174 . for @xmath258 write in the form @xmath280{\lm-\lm_{0}^{+}}}\chi_{0}(\lm)\,\dlm , \qquad \chi_{0}(\lm ) = \prod_{k\ge 1 } \frac{\lm_{k}^{{\text{\tiny\ensuremath{\bullet}}}}-\lm}{\vs_{k}(\lm)},\ ] ] and similarly , for @xmath50 write in the form @xmath281{\lm-\lm_{0}^{+ } } } \prod_{\atop{k\neq n}{k\ge 1 } } \frac{\lm_{k}^{{\text{\tiny\ensuremath{\bullet}}}}-\lm}{\vs_{k}(\lm)}.\ ] ] the mappings @xmath282 , @xmath172 , are analytic on @xmath174 cf . * appendix a ) . moreover , the roots @xmath186{\lm-\lm_{0}^{+}}$ ] and @xmath283 , @xmath50 , respectively , admit opposite signs on opposite sides of @xmath284 and @xmath173 , @xmath50 , respectively . therefore , in view of @xmath285 , for any @xmath172 and @xmath286 , @xmath287 consequently , @xmath277 is continuous and hence analytic on all of @xmath174 . \(iv ) if @xmath146 is real - valued , then for any @xmath286 @xmath288{\dl^{2}-4 } } \dlm = \pm \operatorname{arccosh}\frac{(-1)^{n}\dl(\lm)}{2}.\ ] ] \(v ) at the zero potential , @xmath289{\lm}\bigr)\dlm$ ] which follows directly from the product representation of @xmath205 . as @xmath290 one gets @xmath291{\lm}$ ] . since @xmath292 , we drop the subscript of @xmath246 in the sequel to simplify notation and denote @xmath293 by exactly the same arguments one can show that @xmath205 and @xmath242 , @xmath172 satisfy the properties stated in lemma [ w - closed ] & [ f - prop ] when @xmath294 is extended to the open neighborhood of @xmath295 within @xmath296 constructed in @xcite , and @xmath297 is chosen as an open neighborhood of @xmath146 within @xmath296 such that a common set of isolating neighborhoods exists . however , we will not make use of this fact . if @xmath146 is real - valued , then we can integrate by parts and subsequently shrink the contour of integration to the interval @xmath173 to obtain @xmath298 thus for real - valued potentials all action variables @xmath104 are real , and those on even levels are nonnegative . moreover , by the mean value theorem , @xmath299 for some @xmath300 . by the actions on level zero satisfy @xmath301 in the sequel we derive similar formulae expressing the actions on any level @xmath3 in terms of hamiltonians of the hierarchy . to this end , we need the following expansion of @xmath302 for real - valued finite - gap potentials , which follows from the expansion of the discriminant @xmath218 obtained in appendix [ a : dl - exp ] . [ f - exp ] suppose @xmath146 is a real - valued finite - gap potential with @xmath303 = 0 $ ] and let @xmath304 . then for any @xmath305 @xmath306 and @xmath307 as @xmath308 . since @xmath146 is real - valued , the function @xmath309 , for any @xmath172 , is strictly increasing from @xmath310 to @xmath311 on @xmath312 $ ] and the canonical root for @xmath313 is given by @xmath194{\dl^2(\lm ) - 4 } = ( -1)^{n+1 } \ii \sqrt[+]{4 - \dl^2(\lm)}.\ ] ] furthermore one computes for @xmath313 @xmath314{1-\dl^2(\lm)/4 } } = \frac{\dl^{\text{\tiny\ensuremath{\bullet}}}(\lm)}{\sqrt[c]{\dl^2(\lm ) - 4}}.\end{aligned}\ ] ] hence for any @xmath315 , @xmath316{\dl^2(\lm ) - 4 } } \ , \dlm \\ & = -\ii n \pi + \left[-\ii { \sin^{-1}}\left((-1)^{n+1}\frac{\dl(\lm)}{2}\right ) \right]_{\lm_n^+}^\lm \\ & = -\ii ( n + 1/2 ) \pi - \ii{\sin^{-1}}\left((-1)^{n+1}\frac{\dl(\lm)}{2}\right).\end{aligned}\ ] ] since @xmath146 as a finite - gap potential is smooth , it follows from lemma [ dl - exp ] that with @xmath304 for any @xmath317 @xmath318 where @xmath319 using that @xmath320{1-z^2}}$ ] and taking into account the estimates - one gets by the mean value theorem @xmath321 and hence @xmath322 finally , writing @xmath323 one gets by the addition theorem for the sine @xmath324 and hence in view of @xmath325 this gives the claimed expansion for @xmath326 . the claim for @xmath327 then follows by a straightforward computation . we are now in the position to prove the trace formula on an arbitrary level @xmath3 . the proof appears as a straightforward generalization from the case @xmath106 found in ( * ? ? ? * appendix e ) . [ tf ] suppose @xmath3 and @xmath328 , then @xmath329 where an empty sum denotes zero , and @xmath330 denotes the @xmath111th hamiltonian in the hierarchy . note that @xmath331 uniformly on bounded subsets of @xmath332 see @xcite and also appendix [ s : appendix ] . moreover , @xmath333 uniformly on bounded subsets of @xmath334 see @xcite . thus the sum of the actions on level @xmath116 converges locally uniformly to a real - analytic function on @xmath335 . since also the right hand side of is real - analytic on @xmath16 , it suffices to prove the claim on the dense subset of real - valued finite - gap potentials . suppose @xmath146 is a real - valued finite - gap potential , then there exists @xmath317 such that @xmath336 for @xmath337 . let @xmath338 denote a circle around the origin of sufficiently large radius @xmath339 so that all gaps @xmath173 , @xmath340 are enclosed . then @xmath341 is analytic outside @xmath338 by lemma [ f - prop ] and by contour deformation @xmath342 since @xmath277 is analytic on @xmath174 , the even powers of @xmath242 in the expansion of @xmath343 do not contribute to the contour integral , thus @xmath344 on the other hand , according to proposition [ f - exp ] for @xmath345 @xmath346 as @xmath308 . one infers directly from that for a finite - gap potential , @xmath347 and hence @xmath348 as @xmath349 . therefore , @xmath350 is meromorphic at infinity and we obtain from using cauchy s theorem @xmath351 this proves the trace formula . as an immediate consequence we obtain for @xmath352 @xmath353 where @xmath354 comprises only lower order derivatives of @xmath146 . in sections [ s : act - est ] and [ s : sob - est ] this identity is used to compare the sum @xmath355 of the actions on level @xmath116 with the corresponding sobolev norm of the potential . to proceed , we need an estimate of the action @xmath104 on level @xmath3 in terms of the weighted action @xmath356 . [ injn ] suppose @xmath357 , then for @xmath358 and @xmath3 , @xmath359 while the remaining actions for @xmath360 satisfy @xmath361 recall from that @xmath362 with @xmath300 . with proposition [ ev - as ] from the appendix we find for @xmath363 , @xmath364 while for @xmath360 , @xmath365 in this section we obtain an estimate of all weighted @xmath51-norms @xmath366 , @xmath27 , in terms of the sobolev norms @xmath367 of @xmath146 . this will be done in two steps . first , we estimate @xmath366 in terms of the sum @xmath368 of the actions on level @xmath116 and a remainder depending solely on the @xmath8-norm @xmath369 . second , we express @xmath368 through hamiltonians of the hierarchy using the trace formula . the polynomial structure of these hamiltonians then allows us to prove the claimed estimate . [ iqj - est ] uniformly for all @xmath370 with @xmath3 , @xmath371 clearly , by proposition [ injn ] @xmath372 for @xmath373 the actions @xmath104 may become negative , albeit , with the lower bound @xmath374 , hence @xmath375 the remaining actions may then be estimated by @xmath376 both estimates together with @xmath377 give the claim . lemma [ iqj - est ] and the trace formula yield @xmath378 using the @xmath379-estimate @xmath380 on @xmath381 , we obtain for the hamiltonian @xmath382 thus we arrive at @xmath383 to proceed with the general case @xmath384 , we need an estimate of the hamiltonian @xmath385 for @xmath384 . to this end , we recall some well known facts about the hamiltonians in the hierarchy see for example ( * ? ? ? * theorem 1 ) . [ h - form ] the @xmath116th hamiltonian in the hierarchy has the form @xmath386\,\dx,\ ] ] with @xmath114 being a homogeneous polynomial of degree @xmath387 , without constant term , where @xmath146 counts as 1 degree and differentiation as 1/2 degree . consequently , each monomial @xmath388 of @xmath120 may be estimated by @xmath389 with some positive constant @xmath390 and integers @xmath391 , where we denote @xmath392 to simplify notation . it turns out to be convenient to use exponents which are multiples of two , that is @xmath393 with @xmath394 . since @xmath388 is homogenous of degree @xmath387 , @xmath395 thus , we obtain the estimate @xmath396 with positive reals @xmath397 and @xmath398 being the set of all multi - indices satisfying the constraint . the majorant @xmath399 allows us to obtain detailed estimates of @xmath120 and therefore @xmath400 . for @xmath401 we have by @xmath402 hence @xmath399 is at most quadratic in @xmath403 . together with the @xmath404-estimate @xmath405 we obtain for the derivative of highest order @xmath406 estimating the remaining factors using the @xmath407-estimate @xmath408 then gives @xmath409 since @xmath410 , we obtain for the hamiltonian @xmath411 with some absolute constant @xmath412 . recall from lemma [ iqj - est ] that @xmath413 and combining the trace formula with the preceding estimate of the hamiltonians yields @xmath414{\sum_{n\ge 1 } ( 2n\pi)j_{n , m } } & \le \abs{h_m } + 2\sum_{k=0}^{m-2}\abs{h_{m-2-k}h_k}\\ & \le \frac{1}{2}\n{q}_{m}^{2 } + c_m(1+\n{q}_{m-1})^m\n{q}_{m-1}^2.\end{aligned}\ ] ] this proves estimate ( i ) of theorem [ act - sob - est ] . we now turn to the problem of controlling the sobolev norm @xmath367 of the potential in terms of weighted norms of its actions . as a starting point we use the identity @xmath415 inferred from lemma [ h - form ] , and proceed in two steps . first , we estimate the sum @xmath368 of the actions on level @xmath116 in terms of @xmath366 and a remainder depending solely on @xmath416 . from the trace formula we then obtain a bound of @xmath417 in terms of the weighted action norms . in the second step we use the polynomial structure of @xmath114 to obtain an estimate of the @xmath114-integral involving at most @xmath418 but not @xmath419 . here , we again use the notation @xmath420 . an inductive argument then gives the claim . [ j - i - est ] on @xmath16 for any @xmath27 , @xmath421 by proposition [ injn ] , for @xmath358 , @xmath422 while for @xmath423 , @xmath424 since @xmath425 by the trace formula , the claim follows . [ h - i - est ] for any @xmath27 there exists an absolute constant @xmath36 such that @xmath426 for all @xmath370 . we have @xmath427 by , and further by for any @xmath27 @xmath428 lemma [ j - i - est ] gives an estimate of the first summand , while we inductively obtain for the second @xmath429 this proves the claim . we begin with the case @xmath430 . clearly , @xmath431 with the @xmath407-estimate @xmath380 on @xmath381 we find @xmath432 and , together with @xmath433 , we therefore obtain @xmath434 finally , by the preceding theorem , @xmath435 such that @xmath436 to proceed with the general case @xmath384 , we recall from that @xmath437 theorem [ h - i - est ] provides an estimate of @xmath118 , while for @xmath120 we have by @xmath438 where the majorant @xmath399 only contains derivatives of @xmath146 up to order @xmath121 and the multi - indices @xmath439 satisfy the constraint . we use this to prove by induction for any @xmath384 @xmath440 clearly , this holds true for @xmath441 . consider the inductive step @xmath442 . recall from the proof of theorem [ act - sob - est ] ( i ) that @xmath399 is at most quadratic in @xmath443 for @xmath384 . thus , we have the following expansion @xmath444 where for @xmath445 the entity @xmath446 incorporates those multi - indices @xmath401 with @xmath447 . in particular , @xmath448 with some @xmath449 . these three terms will be estimated separately . we begin with @xmath450 . to avoid a flood a constants , we write @xmath451 if @xmath452 with an absolute constant @xmath453 which is independent of @xmath146 but may depend on @xmath116 . using the @xmath407-estimate and the induction hypothesis for @xmath121 , we get @xmath454 second , consider the term @xmath455 incorporating all multi - indices with @xmath456 . since @xmath457 , there has to be a nonzero @xmath458 for some @xmath459 . by cauchy - schwarz and the @xmath407-estimate @xmath460 for the remaining factors the @xmath407-estimate gives @xmath461 both estimates together yield @xmath462 as @xmath463 by , we conclude @xmath464 it remains to estimate the term @xmath465 which incorporates those @xmath439 with @xmath466 . first consider the case that @xmath467 for some @xmath468 . by cauchy - schwarz , @xmath469 while the remaining factors are estimated as usual , @xmath470 such that @xmath471 as before @xmath472 by . finally , consider the case where all @xmath458 except for @xmath473 with @xmath474 vanish . then @xmath475 , and consequently @xmath476 altogether we thus obtain @xmath477 which completes the induction step and proves theorem [ act - sob - est ] ( ii ) . the case of estimating the actions in arbitrary weighted sobolev spaces @xmath91 differs significantly from the case of integer sobolev spaces @xmath16 since for an arbitrary weight @xmath65 there is no identity known to exist relating @xmath478 to hamiltonians of the hierarchy . albeit , even in the case of weighted sobolev spaces , the regularity properties of @xmath146 are well known to be closely related to the decay properties of the gap lengths @xmath479 see e.g. @xcite and appendix [ s : appendix ] . moreover , the asymptotic relation @xmath480 is known to hold locally uniformly on @xmath34 see @xcite . in this section we obtain a quantitative version of which is uniform in @xmath369 on all of @xmath34 . this together with the estimates of the gap lengths given in the appendix allows us to prove theorem [ act - west ] . for @xmath357 we recall from that @xmath481 here the latter identity follows from the closedness of @xmath205 around the gap . in the case @xmath482 , or equivalently @xmath278 , we shrink the contour @xmath199 to the straight line @xmath483 $ ] and insert the product representation of @xmath205 , to obtain @xmath484{\gm_{n}^{2}/4-(\tau_{n}-\lm)^{2}}}\,\dlm , \qquad \chi_{n}(\lm ) = \frac{1}{\sqrt[+]{\lm-\lm_{0}^{+ } } } \prod_{m\neq n } \frac{\lm_{m}^{{\text{\tiny\ensuremath{\bullet}}}}-\lm}{\vs_{m}(\lm)}.\ ] ] parametrizing the gap @xmath485 $ ] by @xmath486 gives @xmath487{1-t^{2}}}(n\pi)\chi_{n}(\tau_{n}+t\gm_{n}/2 ) \,\dt,\ ] ] where we set @xmath488 . since @xmath489 we conclude @xmath490 , and hence @xmath491 the following uniform estimate of @xmath492 allows us to proof the desired quantitative version of in the sequel . on @xmath101 with @xmath493 for any @xmath494 , @xmath495 for @xmath496 with @xmath497 and @xmath286 we have by lemma [ sn - roots ] @xmath498 further , @xmath499 by proposition [ gap - est ] , hence @xmath500 since @xmath501 we have @xmath502 and thus @xmath503 suppose @xmath504 , then by proposition [ gap - est ] @xmath505 while on the other hand for any @xmath506 @xmath507 therefore , by cauchy - schwarz , @xmath508{\sum_{\atop{m\neq n}{m \ge n } } \frac{\gm_{m}}{4\abs{n^{2}-m^{2}}}}^{2 } & \le \frac{9}{16n^{2 + 2s}}\n{q}_{s}^{2 } + \frac{36}{n^{3 + 2s}}\n{q}_{s}^{4}.\end{aligned}\ ] ] clearly @xmath509 for @xmath493 . so if we increase @xmath510 such that @xmath511 , or more conveniently @xmath512 , then by the standard estimates for infinite products , @xmath513 to estimate the remaining part of the product we note that by the ordering of the eigenvalues @xmath514 and consequently ( if @xmath515 ) @xmath516 by proposition [ ev - as ] and lemma [ sn - roots ] we have for @xmath517 and @xmath286 , @xmath518 while on the other hand by proposition [ ev - as ] @xmath519 both estimates together yield ( if @xmath515 ) @xmath520 finally , since @xmath521 = 0 $ ] , we find @xmath522{\lm-\lm_{0}^{+ } } \ge \sqrt[+]{n^{2}\pi^{2}-4\n{q}_{0 } } \ge n\pi/2,\ ] ] and consequently for any @xmath517 @xmath523{\lm-\lm_{0}^{+}}}\prod_{m\neq n } \abs{\frac{\lm_{m}^{{\text{\tiny\ensuremath{\bullet}}}}-\lm}{\vs_{m}(\lm ) } } \le 8n \le 64(1+\n{q}_{s})^{3/2-s}.\qed\ ] ] on @xmath101 with @xmath493 for any @xmath524 , @xmath525 if @xmath336 , then @xmath526 and the estimate clearly holds . if @xmath278 , then by and the preceding lemma , @xmath527 suppose @xmath528 with @xmath529 where @xmath530 and @xmath493 . choose @xmath317 such that @xmath531 . then by the preceding corollary @xmath532 and for the gap lengths we obtain with proposition [ gap - est ] @xmath533 the remaining actions for @xmath534 may be estimated by @xmath535 altogether , we thus find @xmath536 [ section ] [ prop]corollary in this appendix we review , for the convenience of the reader , the localization of the periodic spectrum of hill s operator as well as an estimate of its gap lengths , which both are used in the various parts of this paper . we follow the exposition in @xcite see also @xcite . consider the operator @xmath144 on the interval @xmath145 $ ] endowed with periodic boundary conditions and @xmath146 being a complex - valued , @xmath221-periodic @xmath8-potential with vanishing mean value , that is @xmath537 . the spectrum of @xmath148 for @xmath538 consists of @xmath290 and the double eigenvalues @xmath539 , @xmath50 . for @xmath540 arbitrary and @xmath162 sufficiently large , the equation @xmath541 may be regarded as a perturbation of the free equation @xmath542 , hence one can expect the eigenvalues to come asymptotically in pairs @xmath248 satisfying @xmath543 as @xmath308 . the following localization of the eigenvalues is well known see e.g. @xcite . suppose @xmath550)$ ] is an @xmath8-normalized eigenfunction of @xmath247 , then @xmath551 with the @xmath8-inner product @xmath552 . with cauchy - schwarz , the @xmath379-estimate @xmath553 } + \n{f_{x}}_{0}$ ] , and @xmath554 } \le\n{f}_{0 } = 1 $ ] , @xmath555 } + \n{f_x}_0)\n{q}_0\\ & \le \n{q}_0 + \n{f_x}_0 ^ 2 + \frac{1}{4}\n{q}_0 ^ 2.\end{aligned}\ ] ] consequently , @xmath556 to prove the remaining part of proposition [ ev - as ] , we cover the right complex half - plane with the closed strips @xmath557 for @xmath558 the eigenspace of the double eigenvalue @xmath539 , @xmath50 , is spanned by @xmath559 and @xmath560 . when @xmath142 is sufficiently large , then for @xmath561 the dominant modes of a solution of @xmath562 are thus expected to be @xmath563 . therefore , it makes sense to separate these modes from the others by a lyapunov - schmidt reduction . to this end , denote the space of _ 2-periodic _ complex - valued @xmath564-functions by @xmath565 and consider the splitting @xmath566 the projections onto @xmath567 and @xmath568 are denoted by @xmath569 and @xmath570 , respectively . we write the eigenvalue equation @xmath562 in the form @xmath571 where @xmath572 denotes the operator of multiplication with @xmath146 . since @xmath573 is a fourier - multiplier , by writing @xmath574 we can decompose the equation @xmath575 into the two equations @xmath576 called the @xmath577- and the @xmath578-equation . we first solve the @xmath578-equation on each strip @xmath579 by writing @xmath580 as a function of @xmath1 . with foresight to estimating the gap lengths , we consider operator norms induced by _ shifted weighted norms _ @xcite . for @xmath1 in @xmath581 the @xmath582-shifted @xmath583-norm is given by @xmath584 write @xmath589 with @xmath590 . since @xmath591 , one checks that for @xmath592 @xmath593 hence , the restriction of @xmath573 to @xmath568 is boundedly invertible , and the function @xmath594 is well defined . by hlder s inequality we obtain for the weighted @xmath595-norm @xmath596 uniformly for @xmath597 . moreover , @xmath598 finally , by young s inequality for the convolution of sequences @xmath599 consequently , @xmath600 is a @xmath601-contraction on @xmath565 for all @xmath602 . if we multiply the @xmath578-equation from the left by @xmath603 , then @xmath604 which may be written as @xmath605 hence , for @xmath602 one finds a unique solution @xmath606 of the @xmath578-equation . substituting this solution into the @xmath577-equation yields @xmath607 writing the latter as @xmath608 we immediately conclude that there exists a one - to - one relationship between a nontrivial solution of @xmath609 and a nontrivial 2-periodic solution of @xmath562 . hence , a complex number @xmath597 is a periodic eigenvalue of @xmath547 if and only if the determinant of @xmath610 vanishes . recall that @xmath569 is the orthogonal projection onto the two - dimensional space @xmath567 . the matrix representation of an operator @xmath611 on @xmath567 is given by @xmath612 therefore , we find for @xmath610 the representation @xmath613 { \lm - \sg_{n}\\ & \lm - \sg_{n } } , \qquad\qquad p_n\hat{t}_n\phi = \mat[\bigg ] { a_n & c_n\\ c_{-n } & a_{-n } } , \end{aligned}\ ] ] with @xmath614 , and the coefficients of the latter matrix given by @xmath615 moreover , by inspecting the expansions of @xmath182 and @xmath616 using the representations of @xmath617 and @xmath572 in fourier space , one concludes that @xmath618 hence the diagonal of @xmath619 is homogenous , @xmath620{\lm - \sg_{n } - a_{n } & -c_{n}\\ -c_{-n } & \lm - \sg_{n } -a_{n}}.\ ] ] with @xmath625 we obtain @xmath626 . furthermore , since @xmath627 for any function @xmath628 , we conclude @xmath629 the claim follows with @xmath630 and @xmath631 . the proof for @xmath182 and @xmath632 is the same . the preceding lemma implies that the determinant of @xmath610 @xmath633 is an analytic function in @xmath162 , which is close to @xmath634 for @xmath142 sufficiently large . this is what we need to localize the eigenvalues . let @xmath637 . the preceding lemma together with lemma [ tn - est ] gives @xmath638 thus it follows from rouch s theorem that @xmath639 has a single root in @xmath640 , just as @xmath641 . in a similar fashion , we infer from @xmath642 that @xmath643 and @xmath644 have the same number of roots in @xmath640 , namely two , while @xmath644 clearly has no root in @xmath645 . for each @xmath646 lemma [ sn - roots ] applies giving us two roots @xmath647 and @xmath648 of @xmath644 which are contained in @xmath649 . since the strips @xmath579 cover the right complex halfplane , and @xmath543 as @xmath308 , it follows by a standard counting argument that these roots have to be the periodic eigenvalues @xmath650 . thus @xmath651 moreover , these are the only eigenvalues contained in @xmath652 . hence if we choose @xmath317 such that @xmath653 then for any @xmath654 , @xmath655 we now turn our attention to estimating the gap lengths @xmath656 , which by the preceding considerations satisfy for @xmath363 @xmath657 with @xmath658 being the complex roots of @xmath659 on @xmath174 . we write @xmath661 with @xmath662 where the branch of the root is immaterial . each root @xmath663 of @xmath644 is either a root of @xmath664 or @xmath665 , respectively , and thus satisfies @xmath666 . to estimate the distance of the two roots @xmath667 , we note that @xmath668 is analytic on @xmath579 , so cauchy s estimate gives @xmath669 the claim now follows with @xmath670 choose @xmath682 with @xmath683 , and consider the splitting , @xmath684 for the first term we use @xmath685 , to obtain @xmath686{\sum_{1\le n < n } \abs{\gm_{n}}}^{2}\\ & \le \p{w(8\n{q}_w)}^2\n{q}_w^{2 } \abs{\lm_{n-1}^{+}-\lm_{0}^{+}}^{2}\\ & \le ( 261)^2 \p{w(8\n{q}_w)}^2\n{q}_w^{2 } \n{q}_{w}^{4}.\end{aligned}\ ] ] the second term can be bounded by proposition [ gap - est ] , @xmath687 in this appendix we obtain an asymptotic expansion of the discriminant @xmath688 of @xmath689 for real - valued potentials in @xmath690 as @xmath691 along the real axis . such an expansion is needed in section [ s : setup ] . it is based on special solutions of @xmath692 studied in @xcite . to state the asymptotics for @xmath693 we first need to recall the definition of the hierarchy @xcite @xmath694 for @xmath3 arbitrary , one has @xmath695 with @xmath696 , @xmath50 , given recursively as in ( ? ? ? * section 2 ) by @xmath697 , @xmath698 and ( if @xmath515 ) , @xmath699 according to @xcite the @xmath700 are 1-periodic in @xmath701 , @xmath702 for all @xmath27 , and @xmath703 in @xcite , special solutions of @xmath692 with @xmath711 of the following form have been studied : @xmath712 with @xmath713 where the @xmath714 are given in . the error terms @xmath715 satisfy the estimates @xmath716 uniformly in @xmath717 , @xmath718 with @xmath719 , and uniformly on bounded subsets of @xmath690 . given @xmath710 there exists @xmath720 such that for @xmath721 , @xmath722 is uniformly bounded away from zero . this implies that the solutions @xmath723 and @xmath724 are linearly independent . the fundamental solutions @xmath725 , @xmath726 of @xmath727 can then be expressed as @xmath728 and it follows from that @xmath729 note that @xmath730 and hence @xmath731 since @xmath732 is 1-periodic in @xmath701 , and @xmath733 is uniformly bounded away from zero for @xmath734 , the discriminant @xmath735 can be computed as @xmath736 by definition of @xmath737 we have @xmath738 , where @xmath739 since all @xmath740 vanish , this formula for @xmath741 simplifies , @xmath742 and after changing the index of summation @xmath743 one concludes from and the assumption @xmath744 that @xmath745 as claimed . going through the arguments of the proof one verifies that the claimed uniformity statements hold . in this appendix we prove analyticity properties of the functionals @xmath242 , @xmath50 , introduced in section [ s : setup ] . we use the notation established in section [ s : setup ] . in particular , the open neighborhoods @xmath180 and @xmath167 . by the same line of arguments one obtains the analyticity of @xmath242 when @xmath294 is extended to the open neighborhood of @xmath295 within @xmath296 constructed in @xcite , and @xmath297 is chosen as an open neighborhood of @xmath146 within @xmath296 such that a common set of isolating neighborhoods exists . we want to apply ( * ? ? ? * theorem a.6 ) . first note that for any @xmath751 , @xmath752 is analytic on @xmath750 by lemma [ w - closed ] . hence at each step of the proof we might change @xmath707 at our convenience . in a first step we prove that @xmath753 is analytic on @xmath754 , where @xmath755 . note that @xmath756 is analytic on @xmath180 thus @xmath757 is an analytic subvariety of @xmath180 . fix @xmath758 . recall that @xmath759 , @xmath760 are lexicographically ordered and might therefore not even be continuous near @xmath761 . however , since the eigenvalues @xmath762 , @xmath763 are both simple , there exist a neighborhood @xmath764 of @xmath761 and two analytic functions @xmath765 satisfying the set inequality @xmath766 on @xmath767 and in addition @xmath768 , \\ & \abs{\rho_{2 } - \lm_n^+(p_0 ) } < \frac{1}{4}\min\left [ \dist\p{\lm_{n}^{+}(p_{0 } } , \partial u_{n } ) , \abs{\gm_{n}(p_{0 } ) } \right].\end{aligned}\ ] ] in particular , @xmath769 and @xmath770 . let @xmath771 , @xmath772 be elements of the form @xmath773 , @xmath774 with @xmath775 , @xmath776 . then for each @xmath777 , the line segments @xmath778 $ ] and @xmath779 $ ] are admissible paths and @xmath780 as already noted above , @xmath781 and @xmath782 are both analytic on @xmath750 . we now prove that for @xmath783 , @xmath784 is analytic on @xmath767 . since the cases @xmath785 and @xmath786 are treated in the same way we concentrate on @xmath785 . with the parametrization @xmath787 one gets @xmath788 where @xmath789{\lm - \lm_{0}^{+ } } } \ , \prod_{1\le m\neq n } \frac{\lm_m^{\text{\tiny\ensuremath{\bullet}}}- \lm}{\vs_{m}(\lm)}\ ] ] is analytic on @xmath790 see e.g. ( ? ? ? * appendix a ) . taking into account that @xmath791 one obtains @xmath792{t}\sqrt[+]{\frac{(\rho_1 - \nu_1 ) ( \rho_2 - \lm_{t})}{(\tau_{n}-\lm_{t})^{2}}}.\ ] ] moreover , at @xmath793 we have for any @xmath794 , @xmath795 thus , after possibly shrinking @xmath767 , the mappings @xmath796{\frac{(\rho_1 - \nu_1 ) ( \rho_2 - \lm_{t})}{(\tau_{n}-\lm_{t})^{2}}}$ ] and @xmath797 are continuous on @xmath798\times \vs$ ] , analytic on @xmath767 for every fixed @xmath799 , and uniformly bounded away from zero by some @xmath800 . since @xmath282 is analytic on @xmath801 , and @xmath802 , and @xmath803 are analytic on @xmath804 , it then follows that @xmath805{\frac{(\rho_1 - \nu_1 ) ( \rho_2 - \lm_{t})}{(\tau_{n}-\lm_{t})^{2 } } } } \ , \frac{\dt}{\sqrt[+]{t}},\ ] ] is analytic on @xmath767 as well . in a second step we show that the restriction of @xmath753 to @xmath806 is weakly analytic . note that on @xmath806 , @xmath753 coincides with the function @xmath807 with @xmath808 given by , where the path of integration is chosen in @xmath579 but otherwise arbitrary . thus @xmath809 is weakly analytic . in a third and final step we prove that @xmath753 is continuous on @xmath750 . by the considerations above , @xmath753 is continuous in each point of @xmath754 and the restriction of @xmath753 to @xmath806 is continuous . hence it remains to show that for any @xmath810 and any sequence @xmath811 with @xmath812 in @xmath750 it follows that @xmath813 as @xmath814 . by ( * proposition b.13 ) ( see also ( * ? ? ? * proposition 2.9 ) ) @xmath815 locally uniformly around @xmath816 . since @xmath817 as @xmath814 , there exists @xmath818 so that @xmath819 without loss of generality we may assume that @xmath820 . for each @xmath821 consider the parametrization of @xmath822 $ ] , given by @xmath798\to g_{n}$ ] , @xmath823 . then @xmath824 and thus @xmath825{1-t^2 } } \ , \abs{\chi_n(\lm_t ) } \ , \dt \le c \abs{\gm_n},\ ] ] where @xmath826 can be chosen independently of @xmath111 . since @xmath817 as @xmath827 , we conclude @xmath828 we can choose @xmath829 so that the straight line segment @xmath830 $ ] is admissible for any @xmath831 , @xmath832 . note that @xmath833 for any @xmath834 and @xmath835 . with the parametrization @xmath836 , @xmath837 , and since by the definition of the standard root @xmath838{t^{2}-\gm_{n}^{2}/4z_{k}^{2}}$ ] , one then obtains @xmath839{t^{2 } - \gm_{n}^{2}/4z_{k}^{2 } } } \ , \chi_n(\lm_t ) \ , \dt.\ ] ] as @xmath840 , one has @xmath186{t^{2}-\gm_{n}^{2}/4z_{k}^{2}}\big\|_{p_{k}}\to t$ ] and @xmath841 by . to see that the integral converges to @xmath842 let @xmath843 . then @xmath844{t^{2}-\ep_{k}^{2 } } } \ge \sqrt{t^{2}-\abs{\ep_{k}}^{2 } } \ge ( t-\abs{\ep_{k}})^{1/2}(t+\abs{\ep_{k}})^{1/2},\ ] ] whereas by @xmath845 . hence for any @xmath794 @xmath846{t^{2 } - \gm_{n}^{2}/4z_{k}^{2 } } } \ , \chi_n(\lm_t ) } \le cg_{k}(t),\qquad g_{k}(t ) = \frac{(t+\abs{\ep_{k}})^{1/2}}{(t-\abs{\ep_{k}})^{1/2}}.\ ] ] since @xmath847 is in @xmath848 $ ] for any @xmath834 and converges in @xmath848 $ ] to the constant function @xmath221 as @xmath827 , it follows from the generalized dominated convergence theorem that @xmath849{t^{2 } - \gm_{n}^{2}/4z_{k}^{2 } } } \ , \chi_n(\lm_t ) \ , \dt & = z_{\infty}\int_{0}^{1 } \chi_{n}(\lm_{t})\,\dt\bigg\|_{p_{\infty}}\\ & = \int_{\tau_{n}}^{\nu _ { * } } \chi_{n}(\lm)\dlm\bigg\|_{p\infty}.\end{aligned}\ ] ] in view of it then follows that @xmath850 . altogether we have shown that @xmath753 is continuous on @xmath750 . the claimed analyticity then follows from ( * ? ? ? * theorem a.6 ) . [ ana - vanish ] let @xmath851 be a real banach space and denote by @xmath852 its complexification . assume that @xmath853 is an open connected neighborhood of @xmath854 and that @xmath855 is an analytic map . if @xmath856 , then @xmath857 . let @xmath858 . then near @xmath1 , @xmath628 is represented by its taylor series , @xmath859 which converges absolutely and uniformly ( cf . e.g. ( * ? ? ? * theorem a.4 ) ) . as @xmath856 , it follows that for any @xmath860 and any @xmath50 , @xmath861 . as @xmath628 is analytic , @xmath862 is symmetric and @xmath863-multilinear . hence it follows that @xmath864 which implies that @xmath857 in a neighborhood @xmath865 of @xmath1 with @xmath866 . as @xmath867 is connected it follows from the identity theorem that @xmath868 . flaschka , h. , mclaughlin , d.w . : _ canonically conjugate variables for the korteweg - de vries equation and the toda lattice with periodic boundary conditions_. progr . . phys . * 55*(2 ) , 438456 ( 1976 )	the kdv - equation @xmath0 on the circle admits a global nonlinear fourier transform , also known as birkhoff map , linearizing the kdv flow . the regularity properties of @xmath1 are known to be closely related to the decay properties of the corresponding nonlinear fourier coefficients . in this paper we obtain two - sided polynomial estimates of all integer sobolev norms @xmath2 , @xmath3 , in terms of the weighted norms of the nonlinear fourier transformed , which are linear in the highest order . we further obtain quantitative estimates of the nonlinear fourier transformed in arbitrary weighted sobolev spaces . [ [ keywords - and - phrases . ] ] keywords and phrases . + + + + + + + + + + + + + + + + + + + + + nonlinear fourier transform , korteweg - de vries equation , integrable pdes [ [ mathematics - subject - classification-2000 . ] ] mathematics subject classification ( 2000 ) . + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + 37k15 ( primary ) 35q53 , 37k10 ( secondary ) = 1
despite recent progress in cosmology , very little is still known about the location of baryons in the local universe . the extraordinary wmap ( _ wilkinson microwave anisotropy probe _ ) measurements of the cosmic microwave background ( cmb ) anisotropies favor a @xmath3-cdm scenario in which the baryon density in the universe amounts to @xmath4 % of the total matter - energy density , with a baryon - to - dark - matter ratio of @xmath5 ( bennet et al . , 2003 ) . this number agrees well with predictions by the standard big bang nucleosynthesis when combined with light element ratios ( @xmath6 % : burles & tytler , 1998 ) , and also with the actual number of baryons at redshifts larger than 2 in the trees of the ly@xmath7 forest : @xmath8 % ( rauch , 1998 ; weinberg et al . , this concordance cosmology represents a major advance . however , the number of baryons detected in the `` virialized universe '' at @xmath9 ( i.e. stars , neutral hydrogen associated with galaxies , and x - ray emitting gas in clusters ) is more than a factor of 2 smaller than the concordance value , @xmath10 % ( 2@xmath11 ; e.g. fukugita , hogan , & peebles , 1998 ) . it has also been known since 1959 ( kahn & woltjer , 1959 ) that , locally , more than @xmath12 m@xmath13 are needed to dynamically stabilize our own local group . so , one fundamental question arises : where have the baryons been hiding for the last @xmath14 gyrs of life of the universe ? hydrodynamical simulations for the formation of structures in the universe , predict that in the present epoch ( @xmath15 ) , half of the normal baryonic matter in the universe ( the `` missing baryons '' ) is in a tenuous ( overdensities @xmath16 , relative to the mean baryon density in the universe)``warm phase '' ( the so called warm hot intergalactic medium : whim ) , shock - heated to temperatures of @xmath17 k during the continuous process of structure formation ( e.g. hellsten et al . , 1998 ; cen , & ostriker , 1999 ; dav et al . , 2001 ) . these hot filaments are now detectable with current instruments . two years ago we reported ( nicastro et al . , 2002 , hereinafter n02 ) the first detection of highly ionized o and ne absorption [ ovii@xmath19 , oviii@xmath20 , neix@xmath21 at redshift consistent with zero , in the _ chandra _ low energy transmission grating ( letg ) spectrum of the blazar pks 2155 - 304 ( fig 1 ) . = 2.4 in = 2.4 in detection of li - like oxygen absorption [ ovi@xmath22 was also reported in the far - ultraviolet fuse spectra of this ( sembach et al . , 2000 ; savage et al . , 2000 ; n02 ) and many other lines of sight ( sembach et al . , 2003 : hereinafter s03 ; savage et al . , 2003 ; nicastro et al . , 2003 : hereinafter n03 ) . the resolution of the fuse spectrometers is about 30 times better than that of the letg , and clearly sufficient to resolve the ovi complex in at least two components : a high velocity ( hv - ovi ) component at @xmath23 to -100 km s@xmath24 , and a narrower and more symmetric low velocity ( lv - ovi ) component at @xmath25 to + 100 km s@xmath24 ( see fig . 3a in n02 ) . figure 2 shows that the uv absorbers ( both , hv and lv ) are incompatible with the x - ray absorbers in the hypothesis of purely collisional ionized gas with typical interstellar medium ( ism ) densities ( solid lines and superimposed 2@xmath11 intervals for the ion abundance ratios ) . however , a single temperature solution ( at t@xmath26 k , dashed vertical line in fig . 2 ) can be found for the hv - ovi and the x - ray absorbers if the volume density of the , mainly collisional ionized , gas is low enough ( i.e. @xmath27 @xmath28 ) to allow photoionization by the extragalactic x - ray background to contribute ( to @xmath14 % at @xmath29 @xmath28 ) to the overall ionization balance in the gas . this is because the relative population of he - like and h - like ions of o and ne , compared to li - like o , starts rising at relatively lower temperatures ( i.e. t@xmath30 k ) , compared to pure collisional ionization model ( dashed lines and superimposed 2@xmath11 intervals for the ion abundance ratios . see n02 for further details ) . = 2.4 in = 2.4 in this solution implies a baryon volume density of @xmath31 @xmath28 , corresponding to overdensities of @xmath32 relatively to the average baryon density in the universe . assuming homogeneity , and an [ h / o ] ratio of 0.3 solar , this in turn gives a path - length of the absorber along this line of sight of @xmath33 mpc , so locating the absorber in the intergalactic space surrounding the galaxy . assuming a transverse size of this filament of 1 mpc , gives a baryonic mass of @xmath34 m@xmath13 . inhomogeneity would imply a smaller size of the absorber , and so a lower mass . if , for example , the diffuse intergalactic medium would condensate when approaching the virialized structures of the local group , so mixing up with the ism in the high galactic halo , up to densities of @xmath35 @xmath28 , and if this layer of denser gas provided , say , 75 % of the total observed equivalent baryon column , then the total filament would be about 1 mpc long along the line of sight to pks 2155 - 304 ( i.e. about the distance to m 31 ) , with the dense shell contributing about 10 % to the total mass of about @xmath36 m@xmath13 . the above solution ( and consequent interpretation ) is , of course , not unique . the alternative is that the x - ray and the ultraviolet absorbers are independently produced by a multiphase ism ( e.g. heckman et al . , 2002 ) , or by both a hot multiphase ism ( the uv lines ) and a diffuse local group medium or galactic corona ( the bulk of the x - ray lines ) . the importance of this controversy lies in the total mass implied by the two different interpretations : in one case ( local group whim : lg - whim ) the implied mass is comparable , if not larger , than that of the most massive visible constituents of the local group , with strong implications for the local group dynamics and its stability , while in the galactic scenario the gas mass would add up to a maximum of @xmath37 of the galaxy s mass . one of the main pieces of evidence supporting the extragalactic solution ( i.e. lg - whim solution ) , is the segregation of the hv - ovi in the local standard of rest ( lsr ) frame . about 85 % of the lines of sight towards extragalactic sources observed by fuse , show hv - ovi absorption ( s03 ) . moreover , as demonstrated by n03 , the majority of these absorbers appear to be at rest in the local group standard of rest ( see fig . 2 in n03 ) . the velocity distribution of these absorbers is highly segregated in the local standard of rest ( lsr ) , with all negative - velocity absorbers lying at galactic longitude @xmath38 , and all positive - velocity absorbers lying in the opposite hemisphere ( fig . 2 in n03 ) . this particular distribution is expected for objects at rest in the intergalactic space surrounding our galaxy , with the kinematics in the lsr being an effect of the rotation of the sun in the galaxy s plane . this suggestion is confirmed by the progressive reduction of the degree of segregation ( and so of the amplitude of the average velocity vector ) of the hv - ovi absorbers for translations of the velocity vectors into the galactic standard of frame ( gsf ) and the local group standard of frame ( lgsr ) . in the lgsr the hv - ovi distribution appears random , and the amplitude of the average velocity vector is close to the fuse resolution % confidence level . ] . n03 concluded that the hv - ovi absorbers are at rest in the local group and so trace either a very extended galactic corona ( galactocentric distances @xmath39 kpc ) or a lg - whim filament ( see also wakker , 2003 , these proceedings ) . s03 have argued against this interpretation . they noted that the above evidence is a necessary , but not sufficient condition to prove the lg - whim origin of the majority of the hv - ovi . while we certainly agree with this point , we want to stress here that their main argument against the lg - whim interpretation is the wide correspondence between some of the structures traced by the hv - ovi in the sky , and similar complexes seen in the hydrogen high velocity clouds ( hvcs ) distribution , and thought to be at distances of few kpc from the galaxy center . however , both n03 and s03 were able to tentatively identify ( based on very generous spatial and kinematics associations : up to tens of degrees in space and hundred km s@xmath24 in velocity ) only about 18 - 20 % of the hv - ovi with hvcs with known or estimated galactocentric distance . these objects have not been considered in the statistical analysis presented by n03 . for several other hv - ovi in their sample , s03 claim the association with known hvc structures ( i.e. the magellanic stream , complex c , complex a ) , only on the base of the vicinity ( few degrees or tens of degrees apart ) of these absorbers with the above hvcs structures . incredibly high quality _ chandra _ and/or _ newton_-xmm grating spectra are needed to detect x - ray absorption lines with typical whim intensities ( i.e. ew(@xmath40 ) @xmath41 m ) . the letg spectrum of the blazar pks 2155 - 304 ( n02 ) was collected with an exposure of 60 ks , and during a powerful source outburst ( @xmath42 mcrab @xmath43 erg s@xmath24 @xmath44 ) . this guarranteed us about 700 counts per resolution element at the wavelength of the ovii@xmath45 line , enough to detect absorption lines with ew@xmath46 m , at a significance larger than 3@xmath11 . unfortunately , however , the flux level of pks 2155 - 304 during this observation , was about 10 - 20 times that of the brightest nearby extragalactic sources in their quiescent levels . very long exposures are then normally needed to obtain a sizeable signal - to - noise selected sample of x - ray spectra of extragalactic sources , suitable to study the distribution of the highly ionized x - ray absorbers at @xmath47 . to date , only 5 extragalactic sources have such good quality spectra : pks 2155 - 304 , mkn 421 , 3c 273 , ngc 4593 and ngc 5548 . all of them show ovii@xmath45 absorption with similar intensities ( i.e. about 10 - 20 m : n02 , rassmussen et al . , 2003 , fang et al . , 2003 ) . despite efforts to look for galactic ( or between the galaxy and the two magellanic clouds ) hv - ovi or highly ionized x - ray absorbers , none has been found yet . we searched the literature for reports of , non - intrinsic , highly ionized ( i.e. ovii , oviii or neix ) absorbers at lsr velocity consistent with @xmath48 km s@xmath24 , in the x - ray grating spectra of binaries either in our galaxy or in the two magellanic clouds , and could not find any . we found about a dozen published _ chandra _ and/or _ newton_-xmm observations of x - ray binaries . in few cases variable ( i.e. intrinsic ) highly ionized emission and/or absorption , have been detected ( e.g. ness et al . , 2003 , apj , 594 , l127 ) . the remaining do not show any high ionization absoprption line either intrinsic or due to intervening ism . the situation is even clearer in the fuv . zsarg , et al . , ( 2003 ) report no detection of hv - ovi against any of the 22 o - b stars observed by fuse in the galactic halo . they also note that `` if hv - ovi absorption was as common toward halo stars as toward extragalactic sources , one would expect 10 - 12 detections of hv - ovi absorption in '' their `` sample '' . they further note that this evidence `` is very important because it justifies the association of the thick - disk ovi with the low - velocity gas '' . while we definitely agree with this conclusion , we also think that the above evidence directly suggests the association of the hv - ovi with extragalactic gas . howk , sembach & savage ( 2003 ) also report no detection of hv - ovi in the first 10 kpc , along the line of sight to the globular cluster m 3 . finally , only galactic ( i.e. low - velocity ) ovi absorption , or ovi absorption at the velocity of the two magellanic clouds , is detected against samples of o - b stars in the two clouds ( danforth et al . , 2002 ; hoopes et al . , 2002 ) . the few exceptions have galactic explanation ( i.e. a supernova remnant in our galaxy , along the line of sight to av 229 ) . one of the strongest arguments against the lg - whim interpretation for the hv - ovi and the x - ray absorbers , is the evidence that , along some lines of sight ( i.e. pks 2155 - 304 , mkn 509 : sembach et al . , 1999 ) , the hv - ovi have lower ionization counterparts ( si ii / iii / iv and c ii / iii / iv ) , with similar centroid velocity . this is not compatible with an homogeneous , diffuse and very tenuous whim surrounding our galaxy , and producing the x - ray and the hv - ovi absorption . one possibility is therefore that the different absorptions are produced by a multiphase ism , with typical ism densities . however , inhomogeneities in the lg - whim could also easily give rise to a multiphase medium . the most efficient source of cooling in a line emitting plasma at t@xmath49 k , is the ovi@xmath50 doublet . this mechanism is quite inefficient at typical whim densities , giving e - folding cooling times of 80 gyrs , i.e. larger than the hubble time . such a lg - whim obviously could not have had time to cool down to temperatures typical of the observed low - ionization species . however , an increase by a factor of 100 in density in compressed layers in proximity of the virialized structures of the local group , would reduce the cooling time to about 800 myrs , about 2/25 the galaxy s age . so , these denser layers of lg - whim , would have had time to cool down and produce the observed lower ionization species . moreover , in the local group scenario , the velocity of the absorber in the lsr , is mainly due to our rotational motion in the galaxy . so , along a given line of sight , any layer ( either mpc- or kpc - thick ) would produce absorption lines with roughly the same velocity in the lsr . the profile , however , would be different , depending mainly on the size of the absorber along the line of sight . this seems to agree with the profiles observed in low and high - ionization lines . while evidence for the @xmath0 filament is quite robust both in the fuv and in x - rays , until very recently there had been no clear detection in the x - rays of ovii - oviii whim absorption at @xmath2 ( either associated with the few known intervening ovi - savage , tripp , & lu , 1998 ; tripp , savage , & jenkins , 2000 ; tripp & savage , 2000 ; jenkins et al . , 2003 - or faint - i.e. non - damped - h ly@xmath7 systems in the fuv , or at independent redshifts ) . the deepest observation of one of the brightest quasars at @xmath51 ( the 500 ks letg observation of h 1821 + 643 ) is sensitive only to ew@xmath52 m(at @xmath53 level ) , and only 2 - 3 low - significance ( @xmath54 ) detections of such systems have actually been claimed ( mathur , weinberg , & chen . , 2003 ) along that line of sight , consistently with expectations . this is mostly due to the intrinsic steepness of the logn - log(ew) of the whim ovii - oviii absorption lines , at relatively large ew ( see , e.g. , fig . 4 in hellsten et al . , 1998 ) . a single ovii k@xmath7 system with @xmath55 m is expected along a random line of sight , up to @xmath56 . this number rises to 8 , for only twice weaker systems ( ew@xmath57 m ) . very high signal to noise _ chandra_-letg or _ newton_-xmm rgs spectra of extragalactic sources are therefore needed to detect the whim . these can be provided with either multi - megasecond observations of a single line of sight , or observing blazars when they undergo an outburst phase . we decided to pursue this second strategy , and during the past year triggered two target - of - opportunity observations ( toos ) of the blazar mkn 421 ( @xmath58 , i.e. @xmath59 mpc , de vaucouleurs , et al . , 1991 ) , during two exceptionally high flux outburst phases . the two observations were carried out with the _ chandra _ acis - letg ( 2002 , october 26 - 27 ) and the hrc - letg ( 2003 , july , 1 - 2 ) , lasted 100 ks each , and reported 0.5 - 2 kev fluxes of 100 and 40 mcrab , respectively . this allowed us to collect more than 10 million counts in the coadded first order spectrum , and more than 6000 counts per resolution element at 21.6 . in turn , this guaranteed a 3@xmath11-sensitivity of @xmath60 @xmath44 , more than an order of magnitude fainter than the lg - whim . between the two chandra too observations , mkn 421 has also been observed by fuse , following our request of a too . several absorption lines are clearly detected in the _ chandra _ spectrum of mkn 421 ( fig . 3a : left panel ) , from the relatively strong ovii k@xmath7 line from the lg - whim at @xmath61 down to the ovii k@xmath7 line at the source redshift ( @xmath58 ) . our preliminary analysis suggests the identification of these lines with 3 different absorbers , one of which ( @xmath62 , @xmath63 @xmath44 ) a truly intervening whim system at @xmath64 , close and the ovii@xmath45 systems are shifted in velocity by @xmath65 km s@xmath24 , which supports the existence of multiphase whim . ] to the redshift of a known weak hydrogen ly@xmath7 system ( shull , stocke , & penton , 1996 ) . upper limits ( at @xmath66 ) of @xmath67 @xmath44 and @xmath68 @xmath44 are also set on the oviii k@xmath7 ( fig . 3b : right panels ) and ovi@xmath50 transitions from this system . based on these measurements we estimate a temperature of the intervening whim ovii filament of @xmath69 k , centered around the peak of the temperature distribution of the whim ( e.g. dav et al . , 2001 ) . the number of ovii filaments per unit redshifts of @xmath70 , derived from this single detection , is in good agreement with expectations from hydrodynamical simulations ( hellsten et al . , 1998 ) . this is at least twice as much as the corresponding number for ovi ( @xmath71 , tripp , 2003 , private communication ) , and so by far the largest reservoir of baryons in the local universe .	in this contribution , i review the current observational evidence for the existence of filaments of warm - hot intergalactic medium ( whim ) . in particular , i first focus on the controversial issue of the identification of the @xmath0 highly ionized far - ultraviolet ( i.e. ovi ) and x - ray ( i.e. ovii , oviii and neix ) absorbers with either a very tenuous and diffuse whim filament , or with much denser condensations of material at large distances in the galactic halo . i then present our recent detection ( confidence level @xmath1 ) of the ovii whim at @xmath2 and derive an estimate of the total number of baryons contained in this hard - to - detect phase of the igm . # 1_#1 _ # 1_#1 _ = # 1 1.25 in .125 in .25 in
the phenomenon of coherent population trapping , originating from quantum optics , has recently been recognized as a useful and interesting concept in the electronic context as well.@xcite an all - electronic implementation , proposed in ref . , is based on destructive interference of single - electron tunneling between three quantum dots ( see fig.[fig:3dot ] ) . the trapped state is a coherent superposition of the electronic charge in two of these quantum dots , so it is destabilized as a result of decoherence by coupling to external charges . in the limit of weak decoherence one electron is transferred _ on average _ through the device for each decoherence event . in an experimental breakthrough,@xcite gustavsson have now reported real - time detection of single - electron tunneling , obtaining the full statistics of the number of transferred charges in a given time interval . the first two moments of the counting statistics give the mean current and the noise power , and higher moments further specify the correlations between the tunneling electrons.@xcite this recent development provides a motivation to investigate the counting statistics of coherent population trapping , going beyond the first moment studied in ref .. since the statistics of the decoherence events is poissonian , one might surmise that the charge counting statistics would be poissonian as well . in contrast , we find that charges are transferred in bunches instead of independently as in a poisson process . the fano factor ( ratio of noise power and mean current ) is three times the poisson value in the limit of weak decoherence . we identify the physical origin of this super - poissonian noise in the alternation of two decay processes ( tunnel events and decoherence events ) with very different time scales in accord with the general theory of belzig.@xcite for comparable tunnel and decoherence rates the noise becomes sub - poissonian , while the poisson distribution is approached for strong decoherence . the analysis of ref . was based on the lindblad master equation for electron transport,@xcite which determines only the average number of transferred charges . the full counting statistics can be obtained by an extension of the master equation due to bagrets and nazarov@xcite ( without phase coherence ) and to kielich et al.@xcite ( with phase coherence ) . in spite of the added complexity , we have found analytical solutions for the second moment at any decoherence rate and for the full distribution in the limit of weak or strong decoherence . ) and irreversible transitions ( rate @xmath0 ) are indicated by arrows.,width=264 ] the system under consideration , studied in ref . , is depicted schematically in fig . [ fig:3dot ] . it consists of three tunnel - coupled quantum dots connected to two electron reservoirs . in the limit of large bias voltage , which we consider here , electron tunneling from the source reservoir into the dots and from the dots into the drain reservoir is irreversible . we assume that a single level in each dot lies within range of the bias voltage . we also assume that due to coulomb blockade there can be at most one electron in total in the three dots . the basis states , therefore , consist of the state @xmath1 in which all dots are empty , and the states @xmath2 , @xmath3 , and @xmath4 in which one electron occupies one of the dots . the time evolution of the density matrix @xmath5 for the system is given by the lindblad - type master equation,@xcite @xmath6\\ + \sum_{x = a , b , c,\phi_a,\phi_b,\phi_c } \hspace{-0.8 cm } \big(l_x\rho l^\dagger_x - \tfrac{1}{2}l^\dagger_x l_x\rho - \tfrac{1}{2}\rho l^\dagger_x l_x \big).\end{gathered}\ ] ] the hamiltonian @xmath7 is responsible for reversible tunneling between the dots , with tunnel rate @xmath8 . for simplicity , we assume that the three energy levels in dots @xmath9 , @xmath10 , and @xmath11 are degenerate and that the two tunnel rates from @xmath9 to @xmath11 and from @xmath10 to @xmath11 are the same . the quantum jump operators @xmath12 model irreversible tunneling out of and into the reservoirs , with a rate @xmath0 ( which we again take the same for each dot ) . finally , the quantum jump operators @xmath13 model decoherence due to charge noise with a rate @xmath14 . as a basis for the the density matrix we use the four states @xmath15 if the initial state is @xmath16 most of the coefficients of @xmath5 remain zero . we collect the five non - zero real variables in a vector @xmath17 whose time evolution can be expressed as @xmath18 it is our goal to determine the full counting statistics , being the probability distribution @xmath19 of the number of transferred charges in time @xmath20 . irrelevant transients are removed by taking the limit @xmath21 . the associated cumulant generating function @xmath22 is related to @xmath19 by @xmath23 = \sum_{n=0}^\infty p(n ) \exp(in\chi).\ ] ] from the cumulants @xmath24 we obtain the average current @xmath25 and the zero - frequency noise @xmath26 , both in the limit @xmath21 . the fano factor is defined as @xmath27 . as described in refs . and , in order to calculate @xmath22 one multiplies coefficients of the rate matrix @xmath28 which are associated with tunneling into one of the reservoirs ( the right one in our case ) , by counting factors @xmath29 . this leads to the @xmath30-dependent rate matrix @xmath31 the cumulant generating function for @xmath21 can then be obtained from the eigenvalue @xmath32 of @xmath33 with the smallest absolute real part,@xcite @xmath34 low order cumulants can be calculated by perturbation theory in the counting parameter @xmath30 . the calculation in outlined in app.[sec : fano ] . for the average current we find @xmath35 in agreement with ref . . by calculating the noise power and dividing by the mean current we obtain the fano factor @xmath36 \big[\gamma^2 + 14t^2 + \beta\big]^{-2},\\ \beta & = & 2\gamma\gamma_\phi(1 + 2t^2/\gamma_\phi^2).\end{aligned}\ ] ] in fig . [ fig : fano ] the fano factor has been plotted as a function of @xmath37 for three different values of @xmath38 . the dependence of the fano factor on the decoherence rate is nonmonotonic , crossing over from super - poissonian ( @xmath39 ) to poissonian ( @xmath40 ) via a region of sub - poissonian noise ( @xmath41 ) . to obtain a better understanding of this behavior , we study separately the regions of weak and strong decoherence . on the normalized decoherence rate @xmath42 for three values of @xmath38 . ] for decoherence rate @xmath43 we have the limiting behavior @xmath44 hence one charge is transferred on average per decoherence event , but the fano factor is three times the value for independent charge transfers . there exists a simple physical explanation for this behavior . for zero decoherence the system becomes trapped in the state @xmath45 . the system is untrapped by `` decoherence events '' , which occur randomly at the rate @xmath14 according to poisson statistics . if @xmath14 is sufficiently small there is enough time for the system to decay into the trapped state between two subsequent events , so they can be viewed as independent . the super - poissonian statistics appears because a single decoherence event can trigger the transfer of more than a single charge . the probability of @xmath46 electrons being transferred in total as a consequence of one decoherence event is @xmath47 since a decoherence event projects the trapped state @xmath45 onto itself or onto @xmath48 with equal probabilities @xmath49 and each electron subsequently entering the dots has a 50% chance of getting trapped in the state @xmath45 . the number of electrons which have been transferred due to exactly @xmath50 decoherence events has distribution @xmath51 , the @xmath52th convolution of @xmath53 with itself . we get @xmath54 by definition , @xmath55 being the distribution of the transferred charges after no decoherence events have occurred . the decoherence events in a time @xmath20 have a poisson distribution , @xmath56 combining with eq . ( [ eq : rkn ] ) we find the probability that @xmath46 electrons have been transferred during a time @xmath20 , @xmath57 the corresponding cumulant generating function is @xmath58 which gives rise to the cumulants @xmath59 in agreement with eq . ( [ eq : weak ] ) . the probability distribution ( [ eq : cgf - low - decoh ] ) has been found by belzig in a different model.@xcite as shown in that paper , this superposition of poisson distributions with fano factor 3 arises generically whenever there are two transport channels with very different transport rates ( in our case slow transport via the trapped state @xmath45 , and fast transport via the untrapped state @xmath48 ) . we show that poisson statistics of the transferred charges is obtained for strong decoherence . consider the evolution equation ( [ eq : dotv ] ) of the system . for @xmath60 the coefficients @xmath61 , and @xmath62 will ensure that @xmath63 is equal to @xmath64 after a time which is short compared to the other characteristic times of the system . the trapped and the non - trapped states will be equally populated . let us therefore define @xmath65 and use @xmath66 . the evolution of @xmath67 is governed by @xmath68 , with @xmath69 the rate matrix @xmath70 is obtained by multiplying @xmath71 by the counting factor @xmath29 . an analytic expression can be found for the smallest eigenvalue @xmath72 of @xmath70 , leading to the cumulant generating function @xmath73 of a poisson distribution . in conclusion , we have shown that coherent population trapping in a purely electronic system has a highly nontrivial statistics of transferred charges . depending on the ratios of decoherence rate and tunnel rates , both super - poissonian and sub - poissonian statistics are possible . we have obtained exact analytical solutions for the crossover from sub- to super - poissonian charge transfer , and have calculated the full distribution in the limits of weak and strong decoherence . we hope that the rich behavior of this simple device will motivate experimental work along the lines of ref . and . we thank l. ament for useful discussions . this research was supported by the dutch science foundation nwo / fom . cwg acknowledges a cusanuswerk fellowship . to derive the result ( [ eq : fano ] ) for the fano factor it is sufficient to know the cumulant generating function to second order in @xmath30 . the eigenvalues of the rate matrix @xmath33 defined in eq.([eq : lchi ] ) have the expansion @xmath74 we seek the eigenvalue with the smallest real part in absolute value . that eigenvalue has @xmath75 . we also express the eigenvector @xmath76 corresponding to @xmath77 and the matrix itself in a power series in @xmath30 : @xmath78 inserting the above expansions into the eigenvalue equation @xmath79 yields the following relationships of respectively zeroth , first and second order : @xmath80 the coefficients @xmath81 are known , while @xmath82 and @xmath83 remain to be found by solving these equations sequentially . the first two cumulants then follow from @xmath84 in an analogue way it is possible to calculate higher cumulants . 3 t. brandes , phys . rep . * 408 * , 315 ( 2005 ) . j. siewert and t. brandes , adv . solid state phys . , 181 ( 2004 ) . b. michaelis , c. emary , and c. w. j. beenakker , europhys . lett . * 73 * , 677 ( 2006 ) . s. gustavsson , r. leturcq , b. simovi , r. schleser , t. ihn , p. studerus , k. ensslin , d. c. driscoll , and a. c. gossard , phys . rev . lett . * 96 * , 076605 ( 2006 ) . s. gustavsson , r. leturcq , b. simovi , r. schleser , p. studerus , t. ihn , k. ensslin , d. c. driscoll , and a. c. gossard , cond - mat/0605365 . m. blanter and m. bttiker , phys . rep . * 336 * , 1 ( 2000 ) . w. belzig , phys . b * 71 * , 161301(r ) ( 2005 ) . v. nazarov , physica b * 189 * , 57 ( 1993 ) . s. a. gurvitz , phys . b * 57 * , 6602 ( 1998 ) . d. a. bagrets and yu . v. nazarov , phys . b * 67 * , 085316 ( 2003 ) . g. kielich , p. samuelsson , a. wacker , and e. schll , phys . b * 73 * , 033312 ( 2006 ) .	destructive interference of single - electron tunneling between three quantum dots can trap an electron in a coherent superposition of charge on two of the dots . coupling to external charges causes decoherence of this superposition , and in the presence of a large bias voltage each decoherence event transfers a certain number of electrons through the device . we calculate the counting statistics of the transferred charges , finding a crossover from sub - poissonian to super - poissonian statistics with increasing ratio of tunnel and decoherence rates .
recently , many new resonances are discoveried experimentally . many of them have the proper quantum numbers of the @xmath3 meson states . however , their mass values do not fit the conventional @xmath3 states in the quark model . among them , @xmath0 was first observed in the @xmath4 channel by belle collaboration in 2003 @xcite , and has been confirmed by cdf @xcite , d0 @xcite and babar collaborations @xcite . its quantum numbers are probabely @xmath5 . the corresponding charmonium candidates in the quark model are @xmath6 and @xmath7 which are @xmath8 mev above @xmath9 mev . many people suggested that @xmath0 is mainly a @xmath10 molecular state @xcite . however , to bind the quarks and anti - quarks together in such a four quark state or other multi - quark states , we need to introduce new interaction into the quark model . swanson proposed that the @xmath0 is mainly a @xmath10 molecule bound by the meson - meson interaction derived from the one pion exchange and the quark exchange @xcite . in wong s work @xcite , the meson - meson interaction is derived from a qed - type effective interaction in terms of effective charges for quarks and antiquarks . in refs . @xcite , further investigations based on the molecule assumption were carried out . since the strength of the one pion exchange interaction seems not strong enough to bind the @xmath10 molecular state , other authors argued that @xmath0 may be a dominant @xmath1 charmonium with some admixture of @xmath10 @xcite . in ref . @xcite , after taking into account of the sigma meson exchange potential , the interpretation of @xmath0 as a loosely bound molecular state was further disfavored . we should notice that the color structure of a multi - quark state is much richer than that of a @xmath3 conventional meson state . unlike the conventional mesons or baryons , the @xmath3 and @xmath11 pairs in a multi - quark state can be in the color @xmath12 and @xmath13 representations respectively . some color interactions which have no effects in the @xmath3 or @xmath14 colorless system , may have significant contribution in a multi - quark system . so the complete interactions in the quark model can be quite different after we take into account of these multi - quark states . the @xmath2-channel one gluon exchange interaction is an interaction between quark and anti - quark of the same flavor which annihilate into a virtual gluon . it has no effect on the conventional @xmath3 mesons but acts in the hidden flavor multi - quark system like the charmonium - like moelcular states . in this work , we will investigate the hidden flavor molecular states by considering the @xmath2-channel gluon exchange interaction in the quark model . in the next section , we will model the potential of the @xmath2-channel one gluon exchange interaction starting from the non - relativistic reduction . then the @xmath15 quantum numbers of the molecular states are selected by an analysis of the spin dependence of the interaction strength . in sec . [ sect-3 ] , we will carry out the numerical calculation of @xmath0 and some other charmonium - like states as the molecular states . also we will make prediction about the similar bottomium - like molecular states . finally , we will give a brief summary . in our work , we will use the bhaduri quark model which is a rather simple non - relativistic quark potential model . in the bhaduri model , the hamiltonian can be written as@xcite @xmath17 @xmath18 are constituent quark masses and @xmath19 are the well - known @xmath20 gell - mann matrices . apart from a constant , here the central potential is the usual one gluon exchange coulomb potential plus the linear confinment : @xmath21 @xmath22 is the distance between quark @xmath23 and @xmath24 . the color - magnetic interaction reads @xmath25 where the @xmath26-interaction has been smeared smoothly with the prescription @xmath27 the model parameter values are @xmath28 @xmath29 the @xmath2-channel one gluon exchange interaction ( soge ) takes place when a @xmath3 pair annihilates into a virtual gluon ( fig . [ fig-1 ] ) . the non - relativistic reduction of the potential is : @xmath30 where @xmath31 is the one gluon exchange amplitude . the last @xmath32 represents that the quark and the anti - quark must have the same flavor . clearly , the two factors in the brackets mean that this interaction only occurs when the @xmath3 pair is in the color octet and with spin @xmath33 . in our model calculation , the @xmath26-interaction should be smeared smoothly with the same prescription ( [ smear - ansatz ] ) used in the color - magnetic interaction in the same quark model . however , the qcd amplitude @xmath31 is taken value in the timelike region , where the qcd behavior is still not clear at present time . in the spacelike region , this amplitude is well known from the gluon propagator in perturbative qcd and it reads @xmath34 however , in our study of the hidden flavor molecular states , in order to provide an attrative interaction to favor the formation of the bound states , we need that @xmath35 which means the above formula from perturbative qcd should not be directly used in the timelike region . this change of sign is first suggested in ref . @xcite in the study of @xmath36 and @xmath37 scattering . following ref . @xcite , we assume that @xmath38 where @xmath39 is an introduced strength factor . in our model , after the @xmath26-function being smeared smoothly , the soge potential turns to be : @xmath40 first we observe that the potential @xmath41 is proportional to @xmath42 , so the interaction mainly takes place between the light @xmath3 pairs of @xmath43 in any multi - quark system . next , we will analyse the spin dependence of @xmath41 in each @xmath15 channel of molecular state . for the molecular states we concern , let us assume that their flavor structure is @xmath44 , where @xmath45 and @xmath43 . since we can neglect the @xmath41 interaction between @xmath46 , the interaction acts only when the @xmath3 pair is in color octet and @xmath33 as we have mentioned before . more specifically , the interaction will be switched on if the spin coupling of the four quarks is @xmath47_j.\ ] ] the color structure of @xmath46 and @xmath3 should be octet obviously . however , the color ( re)coupling is irrelevant to our analysis here and will not be explicitly presented . furthermore , we will assume no spatial excitation of any quark . hence all quarks have zero orbital angular momentum . then the quantum numbers @xmath15 can be determined easily for @xmath48 . we have following four @xmath41 interaction channels : @xmath49 finally , we can recover the molecular states by the angular momentum recoupling . we obtain * @xmath50 @xmath51_0 = -\frac{\sqrt3}2 ( q\bar{q})_0(q\bar{q})_0 -\frac12 [ ( q\bar{q})_1(q\bar{q})_1]_0,\ ] ] * @xmath52 @xmath51_1 = -\frac1{\sqrt2 } ( q\bar{q})_1(q\bar{q})_0 + \frac1{\sqrt2 } ( q\bar{q})_0(q\bar{q})_1,\ ] ] * @xmath53 @xmath54_1,\ ] ] * @xmath55 @xmath51_2 = [ ( q\bar{q})_1(q\bar{q})_1]_2\ ] ] . then the factor of interaction strength can be read from the coefficients , which we present in table [ table1 ] . we see that the @xmath41 interaction favors to form the molecular states with @xmath56 . .[table1 ] the list of spin factor of @xmath41 interaction strength in a molecular state @xmath15 made of two meson states @xmath57 and @xmath58 . @xmath59 . [ cols="^,^,^",options="header " , ] to calculate the molecular states , we use the rayleigh - ritz variation principle . the test wave function will be taken to be a series of gaussian basis functions . the gaussian basis functions are often utilized in variational calculations of atomic and molecular problems . recently the method has been also used in the few body system in nuclear and particle physics @xcite . in our case of the @xmath44 molecular state , the test wave function of a molecular state between two clusters of @xmath3 meson states is a series @xmath60 where @xmath61 , @xmath62 , @xmath63 and @xmath64 are the coordinates of @xmath65 , @xmath66 , @xmath67 and @xmath68 , respectively . @xmath70 is the distance between the two meson clusters @xmath71 @xmath72 is the meson wave function which is also taken to be a gaussian function series @xmath73 the wave function of a molecular state is determined by the variation principle in two steps . we first determine the wave function ( [ eq - var-2 ] ) of each meson cluster . then the meson cluster functions @xmath74 are fixed in ( [ eq - var-4 ] ) to obtain the wave function of the molecular state and their masses . to reduce the amount of computation , the parameters @xmath75 and @xmath76 in a gaussian function series are determined in two steps by one - dimensional minimization . we first determined a average @xmath77 value using a single guassian function . then a set @xmath78 of @xmath79 elements is generated from scaling the @xmath77 value up and down by a scale factor @xmath2 @xcite : @xmath80 where @xmath81 . the coefficients @xmath76 are determined by diagonalization the model hamiltonian in the @xmath79-dimensional space spanned by these @xmath79 different gaussian functions . the final values of @xmath75 and @xmath76 and the mass of tetra quark states are determined by searching the scale factor @xmath2 for a minimum of system energy . in this way , we have calculated the possible @xmath50(the combination of two @xmath82 mesons with the spin factor @xmath83 ) , @xmath52 and @xmath55 molecular states in table [ table1 ] . we have calculated both the charmonium - like states @xmath84 which will be compared to recent experimental results and the bottomium - like states @xmath85 which can be investigated in further experiments . the results of relevant heavy quark @xmath86 mesons are listed in table [ table-2 ] . we see that the mass values calculated from the bhaduri quark model deviate at most @xmath87 mev away from the experimental data . the difference between the calculation from the variation method with gussian function series ( cal . ii ) and the exact numerical calculation ( cal . i ) is less than @xmath88 mev . so the variation method is an impresssive good approximation for the numerical calculation of the @xmath86 conventional mesons . c\|ddd meson state & exp . ( mev ) & cal . i ( mev ) & cal . ii ( mev ) + @xmath89 & 1869.62 & 1885.56 & 1886.14 + @xmath90 & 1864.84 & & + @xmath91 & 2006.97 & 2019.96 & 2020.07 + @xmath92 & 2010.27 & & + @xmath93 & 1968.49 & 1995.78 & 1996.36 + @xmath94 & 2112.3 & 2101.2 & 2101.4 + @xmath95 & 5279.15 & 5300.84 & 5301.18 + @xmath96 & 5279.53 & & + @xmath97 & 5325.1 & 5350.3 & 5350.5 + @xmath98 & 5366.3 & 5371.9 & 5372.4 + @xmath99 & 5412.8 & 5413.3 & 5413.6 the results of the molecular states are given in table [ table-3 ] and table [ table-4 ] . here the molecular state is presented by its binding energy : @xmath100 and the rms radius @xmath101 , where @xmath102 , @xmath103 are masses of the two compound mesons , @xmath104 is the mass of the molecular state . c\|ddddd @xmath39 & -0.8 & -1.0 & -1.5 & -2.0 & -3.0 + @xmath105 & - & - & 11.8 & 28.1 & 71.7 + @xmath106 & 1.8 & 6.2 & 24.2 & 48.7 & 109.3 + @xmath107 & - & 5.4 & 21.4 & 43.5 & 98.1 + @xmath108 & - & - & - & - & 5.3 + @xmath109 & - & - & - & - & 15.4 + @xmath110 & - & - & - & - & 14.0 + @xmath111 & 8.8 & 15.3 & 35.2 & 58.5 & 110.9 + @xmath112 & 16.8 & 26.8 & 55.8 & 88.5 & 160.4 + @xmath113 & 16.0 & 25.5 & 53.4 & 84.7 & 153.6 + @xmath114 & - & - & 3.4 & 9.3 & 25.7 + @xmath115 & - & 1.9 & 8.9 & 18.9 & 44.4 + @xmath116 & - & 1.8 & 8.5 & 18.2 & 42.9 c\|ddddd @xmath39 & -0.8 & -1.0 & -1.5 & -2.0 & -3.0 + @xmath105 & - & - & 1.33 & 1.00 & 0.76 + @xmath106 & 2.67 & 1.68 & 1.07 & 0.86 & 0.69 + @xmath107 & - & 1.79 & 1.12 & 0.91 & 0.72 + @xmath108 & - & - & - & - & 1.63 + @xmath109 & - & - & - & - & 1.12 + @xmath110 & - & - & - & - & 1.17 + @xmath111 & 1.11 & 0.94 & 0.74 & 0.64 & 0.54 + @xmath112 & 0.92 & 0.80 & 0.66 & 0.58 & 0.50 + @xmath113 & 0.94 & 0.82 & 0.67 & 0.59 & 0.51 + @xmath114 & - & - & 1.42 & 1.01 & 0.74 + @xmath115 & - & 1.77 & 1.03 & 0.81 & 0.63 + @xmath116 & - & 1.81 & 1.05 & 0.83 & 0.64 we can see from tables [ table-3 ] and [ table-4 ] that indeed the @xmath52 and @xmath55 molecular states are bound preferable since the @xmath41 are strong in these two channels . the binding of @xmath52 states are slightly deeper and tighter than that of @xmath55 s . so from our model calculation , the @xmath52 molecular states should be easily observed in experiments . first , let us look at the @xmath117 sector . if the @xmath0 is a molecular state of @xmath118 and @xmath119 , the binding energy can be estimated from experimental data@xcite @xmath120 which is very small . accounting the uncertainty of approximation in our model calculation , we think the reasonable range of @xmath39 value should be around @xmath121 . if it is small @xmath122 , then it may be the only molecular states in the @xmath117 sector . if @xmath123 , the @xmath55 state may also exists . from our numerical calculation , it mass should be about @xmath124 the belle collaboration has reported a @xmath125 state of @xmath55 at mass @xmath126 mev@xcite , which is a candidate of the @xmath1 charmonium @xmath127 excited state @xmath128 . however its mass is aboult @xmath129 mev below the quark model calculations ( in the bhaduri model , the @xmath128 mass is @xmath130mev ) . several authors have discussed the mass shifts of the coupled channel effect @xcite . if the above @xmath55 molecular state exists , the final state interaction will be important . now , we turn to the possible molecular states in other sectors . we observe that in @xmath131 sector , there are no such molecular states due to the mass increase of @xmath132 of the light quark pair . on the other hand , in the @xmath133 sector , the binding energy increased as the mass increase of @xmath134 of the heavy quark pair . so these bottomium - like molecular states with @xmath52 , @xmath55 , and even @xmath50 may also exists . we can also observe the @xmath135 molecular states @xmath52 and @xmath55 if @xmath123 . the quark model is extended by introducing the @xmath2-channel one gluon exchange interaction . the interaction has no effect on the inner quark structure of conventional @xmath3 mesons and @xmath136 baryons . since the interaction is short ranged a smeared @xmath26-interaction , the effect on the long ranged hadron - hadron interaction is expected to be very small . so the significant effect of this interaction is only in the so called hidden flavor states of multi - quark system . we have calculated the heavy quark molecular states of @xmath137 with @xmath45 and @xmath43 . we find that the interaction can be strong enough to bind the @xmath52 and @xmath55 states , and possibly @xmath50 states . to compare with the recent experiments , the @xmath0 is a candidate of the @xmath138 molecular state and the @xmath139 may be the @xmath128 state which couples to the @xmath140 @xmath55 molecular state . the calculation shows that it is more easy to bind the bottomium - like molecular states . thus we expect that the similiar bottomium - like molecular states of @xmath52 and @xmath55 should also exist .	the recently observed @xmath0 resonance , which is difficult to be assigned a conventional @xmath1 charmonium state in the quark model , may be interpreted as a molecular state . such a molecular state is a hidden flavor four quark state because of its charmonium - like quantum numbers . the @xmath2-channel one gluon exchange is an interaction which only acts in the hidden flavor multi - quark system . in this paper , we will study the @xmath0 and other similiar hidden flavor molecular states in a quark model by taking into account of the @xmath2-channel one gluon exchange interaction .
quantum turbulence @xcite consists of a disordered tangle of reconnecting quantized vortex filaments , each carrying one quantum of circulation @xmath3 . this form of turbulence is easily created by agitating superfluid liquid helium ( @xmath4he ) with propellers @xcite , forks@xcite , or grids @xcite , the same techniques which are used to create turbulence in ordinary fluids . there are also methods which are unique to superfluid liquid helium , for example heat flows @xcite and streams of ions @xcite . quantum turbulence is also studied in superfluid @xmath5he - b @xcite and , more recently , in atomic bose - einstein condensates@xcite . in the study of quantum turbulence , particular attention is devoted to the similarities with ordinary turbulence @xcite . recent progress has been boosted by the development of new flow visualization techniques , such as tracer particles@xcite , andreev scattering @xcite and laser - induced fluorescence@xcite . this article is concerned with numerical simulations of quantum turbulence . starting from the pioneering work of schwarz @xcite , numerical simulations have always played an important role in the field @xcite . the recent experimental progress , rather than reducing the need of numerical simulations , has highlighted their importance in interpreting visualization methods @xcite . in superfluid @xmath4he , the vortex core radius ( @xmath6 ) is many orders of magnitude smaller than the average separation between vortex lines ( typically from @xmath7 to @xmath8 ) or any other relevant length scale in the flow . this feature was early recognised by schwarz@xcite , who proposed to model vortex lines as spaces curves @xmath9 of infinitesimal thickness , where @xmath10 is the time and @xmath11 is arc length . this vortex filament approach is also valid in @xmath5he - b , although to a lesser extent because the core size is larger ( @xmath12 ) . the space curves are numerically discretized by a large , variable number of points @xmath13 ( @xmath14 , which hereafter we refer to as vortex points . at temperatures low enough that the normal fluid and mutual friction is negligible , the governing equation of motion of the superfluid vortex lines is the biot - savart ( bs ) law @xcite @xmath15 where the line integral extends over the entire vortex configuration @xmath16 . ( [ eq : bs ] ) formulates the classical euler dynamics of vortex filaments in incompressible inviscid flows . the assumption of incompressibility is justified , as all relevant speeds , of the order of few cm / s , are much smaller than the speed of sound , which is approximately @xmath17 ( but see the discussion of reconnections and dissipation in section [ sec : algo ] ) . unlike vortex lines in classical inviscid euler flows , however , superfluid vortex lines can reconnect with each other when they come sufficiently close @xcite . this effect is captured by a more microscopic model , the gross - pitaevskii equation ( gpe ) for a bose - einstein condensate@xcite . in the context of the vortex filament approach , reconnections are modelled by supplementing eq . ( [ eq : bs ] ) with an algorithmical reconnection procedure , as originally proposed by schwarz @xcite . two difficulties arise in seeking a numerical solution of eq . ( [ eq : bs ] ) . firstly eq . ( [ eq : bs ] ) is singular at @xmath18 . following schwarz @xcite , to circumvent this problem we split the velocity of the filament of the @xmath19 vortex point @xmath13 into a local contribution ( near the singularity ) and non - local contributions , writing @xmath20 here a prime denotes a derivative with respect to arc length ( hence @xmath21 and @xmath22 ) , @xmath23 and @xmath24 are the arclengths of the curves between points @xmath25 and @xmath13 and between @xmath13 and @xmath26 , @xmath27 is the original vortex configuration @xmath16 without the section between @xmath25 and @xmath26 , and @xmath28 is a cutoff parameter . for simplicity , hereafter we assume that the cutoff parameter is equal to the vortex core radius , @xmath29 . to numerically calculate the nonlocal term in eq . ( [ eq : bs_desing ] ) , we must perform further manipulations . we follow @xcite and split the integral into discrete parts , which account for the contribution of the vortex between points along @xmath27 . assuming a straight line between vortex points , the contribution to the nonlocal term due to the vortex segment between points @xmath30 and @xmath31 is given by , @xmath32\mathbf{p } \times \mathbf{q}.\ ] ] where @xmath33 , @xmath34 , @xmath35 , @xmath36 and @xmath37 . the total nonlocal contribution is then found by summing over all @xmath38 . the second difficulty , which we address in this paper , is the computational cost of the bs law . it is apparent from eq . ( [ eq : bs ] ) that the velocity at one vortex point depends on an integral over all other @xmath39 points , so the cpu time required to evolve the vortex configuration in time scales as @xmath40 . because of this cost , it has been difficult to numerically calculate vortex tangles with vortex line density @xmath41 ( vortex length per unit volume ) comparable to what is achieved in experiments . similarly , it has been difficult to perform sufficiently long numerical calculations which are needed , for example , to saturate the initial vortex configuration to a steady state tangle , or to produce long time series suitable for spectral analysis , or to study the decay of quantum turbulence . schwarz @xcite proposed the local induction approximation ( lia ) @xcite as a practical alternative to the exact ( but cpu - intensive ) bs law . neglecting the second non - local term in eq . ( [ eq : bs_desing ] ) , we have simply @xmath42 where @xmath43 is a constant of order unity and @xmath44 is the mean radius of curvature of the vortex filament . the lia greatly increases the efficiency of the numerical code because its computational cost scales with @xmath2 , as opposed to @xmath1 for the bs law . moreover the lia offers insight into vortex dynamics . using lia , it is easy to show that , in the first approximation , the motion of a point on the vortex filament is along the direction which is binormal to the filament at that point , with speed which is inversely proportional to the local radius of curvature @xmath45 . moreover the lia describes well the translational motion of a circular vortex ring and the rotation of a small amplitude kelvin wave . unfortunately , the lia does not capture the interaction between two vortex filaments , and the interaction which different parts of the same vortex have on each other . for example , two circular vortex rings , set coaxially close to each other , move in a leapfrogging fashion , but the effect is ignored by the lia . other limitations of the lia have been discussed in the literature @xcite . clearly the lia is not the best tool to study important properties of turbulence which arise from the vortex interaction , such as how the energy is distributed over the length scales ( the kolmogorov spectrum ) . in a recent paper @xcite we presented the spectrum of vortex density fluctuations in superfluid turbulence and compared it to experiments . this calculation was possible because we used a tree algorithm , which retains the long - range interaction of the bs law , but sensibly approximate it , so that its computational cost scaled as @xmath0 rather than @xmath1 . the difference between @xmath1 and @xmath0 can not be overstated : it is the same improvement of the fast fourier transform . for example , in the work reported in ref . @xcite , the typical number of vortex points which we used was of the order of @xmath46 . tree algorithms are popular in astrophysical n - body simulations @xcite . the numerical solution of the gravitational interaction of @xmath2 bodies , in fact , has the same difficulty of the bs law : the motion of one body depends on the other @xmath39 bodies , hence the computational cost scales as @xmath1 , as for vortex points . the aim of this paper is to present a detailed description of a tree algorithm tailored to vortex dynamics , provide quantitative error estimates associated with the method , and compare results with methods used in the literature . in particular we show that the tree method provides huge computational savings but without the inherent errors associated with the lia . the structure of the article is as follows . section [ sec : tree ] outlines the algorithms and methods required to create a tree structure for a set of points embedded in three - dimensional space . we pay careful attention to how this tree structure can be efficiently used to compute the velocity at a particular vortex point . in section [ sec : algo ] we provide further details on the algorithms required to apply the vortex filament method to the study of quantised turbulence . section [ sec : tree_test ] contains the results of numerical tests of the tree algorithm . we show that the @xmath1 scaling inherent in the evaluation of the bs integral can be approximated , with a tolerable loss of accuracy , by a tree method which exhibits an improved @xmath47 scaling . a recent , and very detailed , numerical study of counterflow turbulence by adachi _ et al . _ @xcite as again demonstrated that the lia is unsuitable for the study of quantum turbulence . in section [ sec : counterflow ] we verify that the tree method does not exhibit the same deficiencies of the lia . in all simulations presented , here we use parameters which refer to superfluid @xmath4he : circulation @xmath48 and vortex core radius @xmath49 . all our results and methods , however , can be generalised to vortex systems in low temperature @xmath5he - b . ( color online ) illustration of the tree construction in two dimensions ( quad - tree ) . the points ( blue dots ) are enclosed in the root cell ( a ) , which is divided into four cells of half size ( b ) , until ( c , d ) there is only one point per cell.,scaledwidth=40.0% ] building on the pioneering work of barnes and hut @xcite , the majority of astrophysical and cosmological n - body simulations have made use of tree algorithms to enhance the efficiency of the simulation with a relatively small loss in accuracy @xcite . the advantages of these methods are two - fold . firstly they have been shown to scale as @xmath50 , where @xmath2 is the number of particles used in the simulation . secondly there are now openly available parallel implementations of the tree algorithm , which have been used to compute simulations with up to 10 billion particles @xcite . our implementation of the tree algorithm for vortex dynamics is limited to a serial program , however the development of a parallel scheme is one of our future aims . we are not the first to apply tree methods to quantised vortex simulations . kivotedes @xcite used a tree method to enhance a vortex filament simulation , but did not provide any description , testing and evaluation of the performance of the method . kozik and svistunov @xcite introduced an interesting method to speed up the simulation of the kelvin wave cascade along quantised vortices using a scale separation scheme . whilst this is a significant improvement over the full bs integral the problem discussed in @xcite is inherently one - dimensional and it is not clear if there is an efficient implementation of the scheme that would enhance the speed of a three dimensional simulation . detailed descriptions of the tree algorithm can be found in a number of works @xcite , but their focus is on astrophysical simulations . in this paper we provide a detailed description of the method which is accessible to the fluid dynamics and the quantum turbulence community . at each time - step we have a set of orientated vortex points which represent the current position and direction of the vortex filaments . we first group the vortex points into a hierarchy of cubes which is arranged in a three - dimensional octree structure . the octree can be either be constructed top down or bottom up . we think that it is conceptually easier to work from the top down , repeatedly dividing the domain . to achieve this aim , we first divide the full computational box , the root cell , into eight cubes , and then continue to divide each cube into eight smaller ` children ' cubes , until either a cube is empty or contains only one vortex point . the numerical code , qvort we have developed is written in fortran 2003 and employs linked lists @xcite to efficiently create the tree structure . in the code each cell in the tree is a node which , using pointers , links to the 8 children within the domain of the parent node . the main node is set to the full computational box , we then repeatedly call a recursive subroutine to allocate the points in this box to the main nodes child cells . this process is then repeated dividing the main nodes child cells until the hierarchical tree structure is created . a cartoon of the calculation of the centre of circulation ( green ) triangle and the total circulation bold gray arrow . the ( red ) circles mark the position of the vortex points within a tree cell and the black line gives a representation of the vortex filament.,scaledwidth=50.0%,scaledwidth=50.0% ] as we create the tree , we calculate the total circulation contained within each cube and the ` center of circulation ' using the points that the cube contains . we denote the number of points the cube contains as @xmath51 , then the total circulation is given by @xmath52 likewise the center of circulation is calculated as , @xmath53 a schematic diagram for these quantities is displayed in fig [ fig : circ_cent ] . the time required for constructing the tree scales as @xmath54 , so it feasible ( and essential ) to ` redraw ' the tree at each time - step . [ fig : treecartoon ] illustrates this procedure in two dimensions for clarity . [ fig : octree ] shows the complex nature of the octree required for a single vortex ring in a three - dimensional calculation . the enhanced speed of the tree method appears when we calculate the induced velocity at each vortex point @xmath13 . as opposed to the bs method , we only take the full contribution of points near to the vortex point in question . the induced velocity from far vortex points is an average contribution , hence the number of evaluations required per point is significantly smaller than @xmath39 . therefore for each point we must ` walk ' the tree , and decide if a cell in the octree is sufficiently far . this is done using the concept of the opening angle @xmath55 . in their original work @xcite , barnes & hut calculated @xmath55 by evaluating the distance of the center of vorticity from @xmath13 , which we denote as @xmath56 . since each cube in the octree has a known width @xmath57 , then the opening angle is @xmath58 . we then must define a critical ( or maximum ) opening angle @xmath59 . if @xmath60 we accept the cube , and its contribution is used in computing the nonlocal contribution to the velocity via eq . ( [ eq : bs_schwarz_nl ] ) . the crucial difference between the full bs law and the tree approximation is that the vectors @xmath61 and @xmath62 must be redefined as @xmath63 and @xmath64 . if @xmath65 then each cell contains one point at the most , and it is clear that @xmath66 , @xmath67 and we recover @xmath61 and @xmath62 used in the full bs integral as one would expect . if @xmath68 , then we open the cube ( assuming it contains more that one vortex point ) and repeat the test on each of the child cubes that it contains . as a link list is used , from a programming point of view , it is relatively easy to start at the main node and follow the separate branches of the tree , at each stage evaluating whether the cells contribution can be used . the tree - walk ends when the contributions of all cubes have been evaluated . we note that the adjacent points are exempt from this walk , and their contribution is taken as in eq . ( [ eq : bs_desing ] ) so as to avoid the singularity . a correction to this method was proposed by barnes@xcite to avoid errors which would arise if the center of vorticity is near the edge of the cube . this requires the calculation of the distance from the center of vorticity and the geometrical center of the cube , which we denote as @xmath69 . the test criteria then becomes @xmath70 . in section [ sec : tree_test ] we evaluate the improvement caused by this new criterion . whatever opening criterion is used , it is crucial to have a reasonable choice for @xmath59 . if @xmath59 is too large , the discrepancy between the full bs integral and the tree approximation will damage to the accuracy of the simulation . in section [ sec : tree_test ] we shall also show how the errors in the calculation of the velocity at the vortex points depends on @xmath59 . in the next section we describe other necessary details of the vortex filament method which we used . the numerical simulations presented here use a @xmath71 order adams bashforth scheme . given an evolution equation of the form @xmath72 , the recursion formula is @xmath73 where @xmath74 is the time - step and the superscript @xmath75 refers to the time @xmath76 @xmath77 . lower order schemes are used for the initial steps of the calculations , when older velocity values are not available . as the vortex points move , the separation between neighbours along the same filament is not constant . in general this is not a problem , as we can use finite - difference schemes which work with an adaptive mesh size . however the distance between points sets the resolution of the calculation , therefore we must set some upper - bound on the distance between vortex points before we re - mesh the filaments by introducing new vortex points . our criterion is the following : if the separation between two vortex points , @xmath78 and @xmath26 , becomes greater than some threshold quantity @xmath79 ( which we call the minimum resolution ) , we introduce a new vortex point at position @xmath80 given by @xmath81 where @xmath82 . in this way @xmath83 , that is to say the insertion of new vortex points preserves the curvature . this property is desirable , as the curvature of the filament directly affects the velocity field via eq . ( [ eq : bs_desing ] ) . our procedure is thus superior to simpler linear interpolation . conversely , vortex points are removed if their separation becomes less than @xmath84 , ensuring that the shortest length - scale of the calculation remains fixed . this property is important , as the maximum time - step which can be used is related to this length - scale . in the long - wavelength approximation ( @xmath85 ) , the angular frequency of a kelvin wave , along a straight vortex is @xmath86,\ ] ] where @xmath87 is euler s constant . hence the timescale for the fastest motions , with a maximum wavenumber @xmath88 , is @xmath89 thus a reasonable maximum time - step that can be used is given by @xmath90 we note that at finite temperatures this condition may be relaxed as mutual friction will tend to dampen kelvin waves . setting a consistent minimum separation between vortex points makes the numerical resolution of the calculation transparent , and prevents the creation of too short length - scales which would considerably slow down the calculation . in order to model any turbulent system we must pay particular attention to the dissipation of kinetic energy . even at very low temperatures , in the absence of mutual friction with the normal fluid , there are still mechanisms which dissipate the kinetic energy of quantised vortices : reconnections , decay of very small vortex loops and direct phonon emission . using the gpe condensate model , leadbeater & al @xcite showed that , at vortex reconnection events , a small amount of kinetic energy is transformed into acoustic energy of a rarefaction pulse which moves away . since our model neglects sound waves and it would be impractical to determine with accuracy very small changes of kinetic energy , we interpret this dissipation of kinetic energy as a reduction of vortex length , essentially using length as a proxy for energy . we model reconnection events algorithmically in the following way . if two discretization points , which are not neighbours , become closer to each other than the local discretization distance , a numerical algorithm reconnects the two filaments by simply switching flags for the vortex points in front and behind the filament , subject to the criterion that the total length decreases . self - reconnections ( which can arise if a vortex filament has twisted and coiled by a large amount ) are treated in the same way . since reconnections involve only anti - parallel filaments , prior to reconnection we form local ( unit ) tangent vectors @xmath91 , and , using the inner product , we check that the two filaments are not parallel . clearly , testing for possible reconnections requires a distance evaluation between a vortex point and the vortex points in the vicinity . the most efficient way to do this is within the tree - walk itself , creating a linked list of all vortex points within a radius of @xmath84 from a vortex point in question . this avoids the need to re - loop over all discretization points , which would involve approximately @xmath1 operations . we distinguish two kinds of phonon emissions . the condensate model shows that very small vortex loops , of radius of the order of @xmath92 , which move at speed close to the speed of sound , are unstable and decay into sound@xcite . it would not be practical to mesh vortex filaments down to this atomic scale , so we remove any small vortex loop with less than five discretization points . phonon emission can also be induced by very short kelvin waves which rotate rapidly @xcite . again , since we do not mesh filaments down to this scale , we remove vortex points if the local wavelength is smaller than a specified value @xcite . since our smallest length - scale is maintained at @xmath84 , the maximum curvature is of the order of @xmath93 . we proceed in this way : at each time - step we compute the local curvature @xmath94 at each vortex point . if , at some location along a vortex filament , the local curvature exceeds the critical value @xmath95 ( @xmath96 of the maximum value @xmath93 ) , this vortex point is removed , smoothing the vortex filament locally . this loss of length as a proxy of energy is our model of phonon emission in the low temperature limit . we approximate all spatial derivatives using @xmath97 order finite difference schemes which account for varying mesh sizes along the vortex filaments @xcite . a high - order scheme is desirable as the spatial derivatives enter into the equation of motion via , eq . ( [ eq : bs_desing ] ) . let @xmath13 be the @xmath98 point on the vortex filament ; the two points behind have positions @xmath99 and @xmath25 , and the two points in front have positions @xmath26 and @xmath100 . we denote @xmath101 , @xmath102 , @xmath103 , and @xmath104 . we have : @xmath105 where the coefficients @xmath106 , @xmath107 , @xmath108 , @xmath109 and @xmath110 are given by , @xmath111 @xmath112 @xmath113 @xmath114 @xmath115 similarly we have : @xmath116 where the coefficients @xmath117 , @xmath118 , @xmath119 , @xmath120 and @xmath121 are given by @xmath122}{\ell_{i-1}(\ell_{i-1}+\ell_{i})(\ell_{i-1}+\ell_{i}+\ell_{i+1})(\ell_{i-1}+\ell_{i}+\ell_{i+1}+\ell_{i+2})}\ ] ] @xmath123}{\ell_{i-1}\ell_{i}(\ell_{i}+\ell_{i+1})(\ell_{i}+\ell_{i+1}+\ell_{i+2})}\ ] ] @xmath124 @xmath125}{\ell_{i+1}\ell_{i+2}(\ell_{i}+\ell_{i+1})(\ell_{i-1}+\ell_{i}+\ell_{i+1})}\ ] ] @xmath126}{\ell_{i+2}(\ell_{i+1}+\ell_{i+2})(\ell_{i}+\ell_{i+1}+\ell_{i+2})(\ell_{i-1}+\ell_{i}+\ell_{i+1}+\ell_{i+2})}\ ] ] note that if we set @xmath127 , then the above expressions reduce to familiar finite - difference schemes for a uniform mesh : @xmath128 @xmath129 the final issues to address in this section is the boundary conditions used in the simulations and how they are implemented . all numerical simulations presented here were performed with periodic boundary conditions . the computational domain was a cube . if a vortex point leaves the domain , then it is simply re - inserted at the opposite side of the cube . the presence of periodic boundary conditions affects the calculation of the distance between two vortex points ( two vortex points point at the opposite end of the domain may be actually close to each other ) . we must also use periodic ` wrapping ' to ensure that the velocity field is periodic ; in theory we should wrap the cube with an infinite number of copies in all directions . in practice , we only apply a single layer of wrapping , which still requires a very large amount of numerical evaluations . if we used the bs law , a single layer of wrapping would require @xmath130 evaluations per time - step . fortunately again the tree method eases the amount of numerical work , as the computational domain is wrapped with copies of the octree we have already computed . walking each of these trees , as well as the tree for the main domain , does not require a significant increase of cpu time we now test the error associated with the use of the tree approximation , and show that , with an appropriate choice of @xmath59 , the discrepancy between using the tree algorithm and the full bs law is very small . all tests are carried out in a periodic cube of side @xmath131 . the minimum resolution is set at @xmath132 cm , and the time - step is @xmath133s . for our first test we consider a system of ten random vortex loops with an increasing number of vortex points ( hence increasing radius ) . we determine the cpu time required for evolving the system for one hundred time - steps as a function of @xmath55 ; the results are presented in fig . [ fig : tree_timing ] . it is apparent that even for very small maximum opening angles , @xmath59 , there is a huge increase in cpu speed , and for larger values of @xmath59 the @xmath134 scaling is apparent . the cpu time required for one hundred time - steps , plotted as a function of the number of vortex points @xmath2 . the calculations were performed on an single processor with a clock speed of @xmath135 . the symbols represent different values of @xmath59 as follows : triangles , @xmath136 , asterisks , @xmath137 , squares , @xmath138 , and circles , @xmath139 ( that is , the full bs calculation).,scaledwidth=60.0% ] the scaling of the mean relative percentage error in the velocity field , @xmath140 , evaluated at discretization points plotted as a function of @xmath141 . the dashed line corresponds to the original opening angle criterion @xcite and the solid line shows the performance of the correction of barnes @xcite.,scaledwidth=60.0% ] the second test consists in determining the discrepancies between the ` true ' velocity field ( obtained using the bs law ) and the velocity filed obtained with the tree method , as a function of @xmath59 . let @xmath142 be the the velocity at the discretization point @xmath13 obtained using the bs integral , and @xmath143 the corresponding velocity obtained using the tree method . we define the mean relative percentage error of the tree method as @xmath144 again , we use a system of 10 randomly oriented and positioned vortex loops in the same periodic box , each vortex loop discretized into 200 vortex points . figure [ fig : tree_theta ] shows how @xmath140 varies as a function of @xmath141 , for both the original tree opening criterion and the correction of barnes . whilst the improvement is modest , since the extra computational cost is negligible , hereafter we include barnes s modification . the next test investigates how this error varies as a function of @xmath2 for a fixed value of @xmath141 . this test is particularly important in view of computing very dense vortex tangles , which are not accessible using the exact bs law . as before , we consider a system of 10 random vortex loops with an increasing number of vortex points in the same periodic box , and compute @xmath140 . figure [ ebarvsn ] shows the results for fixed @xmath137 . it is apparent that the error grows with the number of vortex points , but its magnitude is still small considering the advantages of the tree method . we obtain the same result with a different set up to increase @xmath2 : we fix the vortex loop size but increase the number of vortex loops . the mean relative percentage error of the tree approximation for a fixed maximum opening angle , @xmath145 , with an increasing number of discretization points and therefore vortex length in the system.,scaledwidth=60.0% ] in the previous tests we have only considered the errors at a single time - step of a simulation . it is important to quantify how the difference between the exact bs law and the tree method varies in time . for this test we deliberately choose a vortex configuration which the lia fails to accurately model , namely two ` leap - frogging ' vortices . the initial condition consists of two co - axial vortex rings set in the @xmath146-plane , separated by a small distance in the @xmath147 direction . given the orientation of the vortex lines , the local velocity component - see eq . ( [ eq : bs_desing ] ) - drives the vortices in the negative @xmath147 direction . the non - local velocity component makes the vortex ring which is initially behind to squeeze through the vortex ring which is in front , overtaking it ; the process repeats over and over again , in a leap - frogging fashion ( note that the vortices maintain the circular shape ) . this remarkable behaviour is visible in fig . [ fig : leap_frog ] . we stress that this behaviour is not captured by the lia . we repeat the calculation using the exact bs law , keeping the same number @xmath2 of vortex points and the same minimum resolution @xmath79 . we then quantity the difference between the tree method and the bs law by computing @xmath148 where @xmath149 is the position of the @xmath19 vortex point in the bs calculation , and @xmath150 is the corresponding position in the tree method calculation . the temporal duration of these calculations captures four separate leap - frogging events . figure [ fig : dist_error ] shows the evolution of @xmath151 for two different values of @xmath59 : @xmath152 and @xmath153 . it is apparent that in both cases the difference between the tree method and the exact bs evolution is small ( smaller than the minimum discretization level @xmath79 ) , particularly for @xmath137 . the evolution of the normalised distance error statistic @xmath151 ( see eq . ( [ eq : dbars ] ) ) , for a system of leap - frogging vortex rings , plotted as a function of @xmath10 ( seconds ) . the solid line corresponds to a simulation with @xmath137 , and the dashed line to @xmath154.,scaledwidth=60.0% ] having benchmarked the tree method for simple vortex configurations , we change our focus to numerical simulations of quantum turbulence . we concentrate our attention to turbulence sustained by the relative motion ( counterflow ) of normal fluid and superfluid components , driven by a heat flow at non nonzero temperature . this form of turbulence which has no analogy in ordinary fluids was originally studied experimentally by vinen @xcite , and , numerically , by schwarz @xcite . counterflow turbulence has been recently re - examined by adachi & al . @xcite , who compared results obtained using the lia ( as in the pioneering work of schwarz@xcite ) and the full bs law . counterflow turbulence is particularly suitable for testing the tree method , as adachi & al . used it to discuss the shortcoming of lia . we want to make sure that the tree method does not suffer from the same problems of the lia . at finite temperatures we must account for the mutual friction between the superfluid and the normal fluid . the equation of motion at position @xmath155 is given by @xmath156 , \label{eq : schwarz}\ ] ] where @xmath157 , @xmath158 are temperature dependent friction coefficients @xcite , @xmath159 is the normal fluid s velocity , and the velocity @xmath160 is calculated using the tree approximation to the de - singularised bs integral . as usual , the evolution includes the algorithmical vortex reconnection procedure . the calculation is performed in a periodic cube with sides of length @xmath131 as done by adachi _ et al . _ superfluid and normal fluid velocities @xmath159 and @xmath160 are imposed in the positive and negative @xmath161 directions respectively , where @xmath162 is proportional to the applied heat flux . simulations are performed with the same numerical resolution @xmath163 and time - step @xmath164 as used by adachi _ et al . _ at various value of @xmath165 and @xmath166 . _ et al . _ used the bs law , we use our tree method with @xmath137 . the number of vortex points used in these calculations is not large , in order to overlap with calculations which used the bs law : ) for example a typical value is @xmath167 at @xmath168 and @xmath169 . it must be stressed that , as in the works of schwarz and adachi _ et al . _ , the back - reaction of the superfluid vortex lines onto the normal fluid velocity is neglected ( an effect which may be significant at low temperatures ) . the initial condition consists of six vortex loops set at random locations , as for adachi _ et al . we find that the results do not depend on the initial condition . during the evolution we monitor the vortex line density @xmath170 where @xmath171 is the volume of the computational domain and @xmath172 is the entire vortex configuration . we find that , after an initial transient , the vortex line density saturates to a statistical steady state value , see figure [ fig : cf_line_density ] ( left ) , which agrees well with the value reported by adachi _ et al . _ in their figure 2 at the same temperature and counterflow velocity . figure [ fig : cf_line_density ] ( right ) shows the expected relation @xmath173 between the steady state value of @xmath41 and the driving counterflow . more precisely , we obtain @xmath174 , which compares well with @xmath175 obtained by adachi _ et al . _ , where @xmath176 . another important quantity considered by schwarz and adachi _ et al . _ is the anisotropy parameter @xmath177 d\xi,\ ] ] where @xmath178 is the unit vector parallel in the direction of the normal fluid . figure [ fig : cf_anisotropy ] ( right ) shows that @xmath179 is almost independent of @xmath180 , in agreement with experiments and with the numerical calculations of adachi _ et al . _ performed using the full bs law ( shown in their fig . for example at @xmath181 we have @xmath182 , compared to @xmath183 - @xmath184 of adachi _ et al . figure [ fig : cf_anisotropy ] also shows that this anisotropy is not captured by the lia . a final qualitative test is shown in fig . [ fig : cf_snaps ] : the left panel shows the bundling of vortex lines wrongly predicted by the lia , and the right panel shows the vortex tangle computed using the tree method , in agreement with adachi _ et al . _ . we conclude that the tree method is an ideal tool to perform numerical simulations of quantum turbulence . it does not suffer from the known shortcoming of the lia , retaining to a good approximation the nonlocal interaction which is typical of the exact bs law . unlike the bs law , whose computational cost scales as @xmath1 where @xmath2 is the number of discretization points , the cost of tree method scales as @xmath0 , thus making possible numerical calculations of intense vortex tangles for a sufficiently long time . finally , the tree method can be further speeded up by parallelization . v.springel , s.d.m . white , a. jenkins , c.s . frenk , n. yoshida , l. gao , j. navarro , r. thacker , d. croton , j. helly , j.a . peacock , s. cole , p. thomas , h. couchman , a. evrard , j. colberg , & f. pearce , nature * 435 * , 629 ( 2005 )	we present a numerical method to compute the evolution of vortex filaments in superfluid helium . the method is based on a tree algorithm which considerably speeds up the calculation of biot - savart integrals . we show that the computational cost scales as @xmath0 rather than @xmath1 , where @xmath2 is the number of discretization points . we test the method and its properties for a variety of vortex configurations , ranging from simple vortex rings to a counterflow vortex tangle , and compare results against the local induction approximation and the exact biot - savart law .
in the recent years , experiments with ultracold trapped fermionic atoms like @xmath0li or @xmath1k have become a powerful tool for improving our understanding of strongly correlated fermi systems . for example , the continuous tunability of the interatomic interaction via feshbach resonances made it possible to study the crossover from a bose - einstein condensate ( bec ) of diatomic molecules to a bcs - type superfluid . the creation of vortex lattices in rotating fermi gases with attractive interaction ( scattering length @xmath2 ) showed unambiguously that the superfluid bcs phase was indeed reached @xcite . by studying the collective modes of these systems , one can obtain a lot of interesting information which can not be obtained from static properties alone . near the bcs phase transition , several collective modes of the trapped fermi gases show interesting features like an extremely strong damping or a sudden change of the frequency @xcite . however , most of these experiments were done in the vicinity of the feshbach resonance ( @xmath3 , where @xmath4 is the fermi momentum in the center of the trap ) , and in this strongly interacting regime no theory is available so far which could quantitatively explain the observed damping effects . an obvious qualitative difference between the strongly and the weakly interacting regimes is that in the latter ( @xmath5 ) the normal phase is in the collisionless regime , whereas in the former ( @xmath3 ) collisions can produce a hydrodynamic behavior of the normal phase @xcite . the aim of the present paper is to discuss the properties of experimentally accessible collective modes , namely of the radial quadrupole and scissors modes , in the bcs regime ( @xmath5 ) . in particular , we want to see how the properties of the modes change across the superfluid - normal phase transition . at zero temperature , the collective modes can be described within the framework of superfluid hydrodynamics , but this is no longer true at non - zero temperature , where superfluid and normal components coexist . in ref . @xcite it was proposed to use landau s two - fluid hydrodynamics in this case . however , although this approach may be valid in the strongly interacting regime , it can not be applied in the weakly interacting regime because of the lack of collisions . the theoretical framework which will be used in the present paper is the semiclassical transport theory developed for clean superconductors by betbeder - matibet and nozires @xcite , which has recently been applied to the case of trapped fermi gases @xcite . in the case of a quadrupole oscillation in a spherical trap , this theory predicts a strong damping of the hydrodynamic mode in the superfluid phase due to its coupling to the collisionless normal component of the system , made of thermally excited quasiparticles . the article is organized as follows . [ method ] gives a brief summary of the theoretical method ( for details , see refs . @xcite and @xcite ) . in sec . [ quadrupole ] , the results for the quadrupole mode are discussed , and it is shown that the frequency jump observed in a recent experiment at innsbruck @xcite can be understood . in sec . [ scissors ] , results for the scissors mode are presented and sec . [ conclusions ] contains further discussions and conclusions . the method used in the present work to describe collective modes in the bcs phase at finite temperature is based on the quasiparticle transport theory by betbeder - matibet and nozires @xcite . the basic degrees of freedom are the phase @xmath6 of the superfluid order parameter @xmath7 , describing the collective flow of the cooper pairs , and the quasiparticle distribution function @xmath8 , describing the normal - fluid component . in equilibrium , one can choose @xmath9 , and the quasiparticle distribution function @xmath10 reduces to the usual equilibrium fermi function @xmath11 , where @xmath12 with @xmath13 , @xmath14 and @xmath15 being the trap potential and chemical potential , respectively . in the non - equilibrium case , one writes @xmath16 , such that the collective velocity of the cooper pairs is given by @xmath17 . in this case , @xmath10 is defined in the rest frame of the cooper pairs , and the density @xmath18 and current @xmath19 are given by @xmath20 the subscripts @xmath21 and @xmath22 denote the time - even and time - odd parts of a given function ( e.g. , @xmath23 ) , @xmath24 is the hamitonian and @xmath25 the quasiparticle energy in the rest frame of the cooper pairs . the equation of motion of @xmath10 can be written in the compact form @xmath26 where @xmath27 denotes the poisson bracket of @xmath28 and @xmath10 , similar to the usual vlasov equation , which can be written as @xmath29 , where @xmath30 denotes the usual distribution function . the equation for @xmath31 can be obtained from the continuity equation @xmath32 which becomes a second - order differential equation for @xmath31 if one inserts the explicit expressions for @xmath18 and @xmath19 . in order to describe collective modes , one writes @xmath33 , @xmath34 , and @xmath35 , with @xmath36 and @xmath37 , the subscripts @xmath38 and @xmath39 referring to equilibrium quantities and deviations from equilibrium , respectively . assuming that the perturbation @xmath40 is weak , one keeps only terms which are linear in the deviations from equilibrium . the explicit equations of motion for @xmath41 and @xmath42 are quite cumbersome and can be found in refs . @xcite and @xcite . note that , contrary to ref . @xcite , the mean - field ( hartree ) potential @xmath43 ( @xmath44 being the coupling constant ) has been neglected in the definitions of @xmath45 and @xmath46 . if one wants to approach the crossover region ( @xmath47 ) , the expression @xmath43 for the mean - field potential is not applicable . one possible solution of this problem is to replace @xmath43 by the real part of the single - particle self - energy in ladder approximation @xcite . anyway , previous experimental and theoretical studies showed that the mean - field shift does not modify the qualitative behaviour of the modes and leads only to minor changes of the mode frequencies . for example , in the recent innsbruck experiment on the radial quadrupole mode @xcite , the frequency shift in the collisionless normal phase from @xmath48 ( @xmath49 being the radial trap frequency ) to @xmath50 was attributed to the mean field . however , the mean field can considerably change the equilibrium density profile , thereby modifying the gap , the critical temperature , etc . for a quantitative study , the inclusion of the mean field is therefore desirable , but beyond the scope of the present paper . except in very simple cases ( e.g. , a sound wave propagating in a uniform system @xcite ) , the equations of motion for @xmath42 and @xmath41 have to be solved numerically . in ref . @xcite , a test - particle method was proposed which is similar to the one often used for solving the boltzmann equation . in the context of the linearized theory , it is useful to write @xmath42 in the form @xmath51 the advantage of this rewriting is that @xmath52 is a smooth function in phase space , contrary to @xmath42 [ this is analogous to writing @xmath53 in the context of the linearized boltzmann equation , which is a rather common trick ] . the test - particle method consists in replacing the continuous function @xmath54 by a sum of @xmath55 functions , @xmath56\,\delta[{{\bm{\mathrm{p}}}}-{{\bm{\mathrm{p}}}}_i(t)]\ ] ] where @xmath57 $ ] . the trajectories of the test particles in phase space , @xmath58 and @xmath59 , satisfy the following equations of motion : @xmath60 these equations of motion do not conserve the energy @xmath61 , but the quasiparticle energy @xmath62 . this can lead to surprising effects which make the numerical solution more involved than that of usual newtonian equations of motion . a well - known example is andreev reflection : if a particle ( @xmath63 ) arrives at a point with @xmath64 , it is reflected as a hole ( @xmath65 ) , and vice versa . in the present calculations , the number of test particles is set to @xmath66 as in ref . however , practically identical results are already obtained with @xmath67 . in fact , the large number of test particles was only needed in ref . @xcite in order to obtain a stable mean field , while here , where the mean field is neglected , a much smaller number of test particles is sufficient . the time dependence of the coefficients @xmath68 can rather easily be solved . for the time dependence of the phase , @xmath31 is expanded in a basis and the time dependence of the coefficients is determined by minimizing the violation of the continuity equation . in the case of the modes under consideration , the velocity field and therefore the form of @xmath31 is well known . in this case , one or two basis functions are completely sufficient @xcite . let us start by discussing the radial quadrupole mode . as it is usually the case in the present experiments , the unperturbed trap potential is supposed to be cigar - shaped : @xmath69 $ ] with @xmath70 . the radial quadrupole mode is excited at time @xmath71 by applying a perturbation of the form @xmath72 with @xmath73 . the equations of motion are then solved for @xmath74 . the excitation spectrum is obtained as the fourier transform of the time evolution of the quadrupole moment @xmath75 . we choose here parameters similar to the recent innsbruck experiments : @xmath76 atoms , trap frequencies @xmath77 hz , @xmath78 hz . since on the one hand our theory is only valid in the bcs limit , but on the other hand the scattering length of @xmath0li saturates at quite a large value at high magnetic fields , we choose as a compromise @xmath79 . for these parameters , the critical temperature within bcs theory is @xmath80 nk ( where @xmath81 is the fermi temperature ) . however , one should remember that bcs theory can not be trusted quantitatively . to give an example , the inclusion of the correlated density according to nozires and schmitt - rink @xcite , keeping the number of atoms @xmath82 and the parameter @xmath79 fixed , reduces the critical temperature by almost @xmath83 to @xmath84 ( @xmath4 and @xmath85 being defined as usual by the density of a non - interacting system with the same number of atoms ) . the resulting excitation spectra for different temperatures @xmath86 are shown in fig . [ fig1 ] . ( in units of the radial trap frequency @xmath49 ) for different temperatures ranging from @xmath87 to @xmath88 ( from bottom to top ) . for a better visibility of the strength contained in the narrow peaks at @xmath48 , all spectra have been folded with a lorentzian of width @xmath89.,width=260 ] at @xmath90 , the spectrum is dominated by a peak situated at the hydrodynamic frequency , @xmath91 . there is , however , a considerable broadening of the peak , i.e. , damping , due to the coupling of the hydrodynamic mode to the thermally excited quasiparticles . at higher temperatures , the peak is further broadened and slightly shifted towards lower frequencies . at the same time , a second peak appears , which is situated at @xmath48 , the frequency of the radial quadrupole mode in an ideal ( collisionless normal - fluid ) fermi gas . close to @xmath92 , this mode is practically undamped ( the width of the peak at @xmath48 in the uppermost spectrum in fig . [ fig1 ] is due to an artificial broadening which has been introduced in order to improve the visibility of the peaks ) . as one can clearly see , the peak position does not move from @xmath93 to @xmath48 with increasing temperature , but the strength is shifted from the lower to the upper peak . there seems to be no simple explanation for the downshift of the hydrodynamic mode , but it can partially be understood as a consequence of the usual repulsion between two coupled modes . the general behavior is very similar to that found in previous work on the quadrupole mode in spherical traps @xcite . this finding suggests the following interpretation of the results of the innsbruck experiment @xcite , where a sudden `` jump '' of the frequency as a function of the interaction strength was found . in the analysis of the experiment , the deformation of the atom cloud was fitted with an ansatz function containing only a single frequency and damping rate . let us try what happens if we apply the same analysis to our theoretical results . in order to follow as closely as possible the experiment , we will do the calculation as a function of the coupling strength for fixed temperature , which we choose rather arbitrarily as @xmath94 . we also simulate the experimental initial conditions , which are different from the @xmath55-like perturbation used above . in the experiment , the potential is first slowly deformed from @xmath95 to @xmath96 , with @xmath97 , giving enough time to the collisions to keep the momentum distribution spherical , and then the deformation @xmath98 is suddenly switched off at @xmath99 . unfortunately there is one step in the experiment which can not be simulated within the present linearized theory , namely the expansion of the cloud before the picture is taken . however , if one assumes that the picture taken after the expansion reflects in some way the state of the system at the moment when the trap was switched off , it seems plausible to assume that the expansion is not important for the observed frequency and damping , but only for the observed amplitude and phase of the oscillation . in fig . [ fig2 ] the obtained time dependence of the quadrupole moment @xmath100 is shown for different values of @xmath101 . the solid lines are the calculations , while the dashed lines correspond to fits with functions of the form @xmath102 the parameters @xmath103 , @xmath104 , @xmath105 , @xmath31 , @xmath106 , and @xmath107 are determined by minimizing the error @xmath108 ^ 2\,,\ ] ] where @xmath109 ms is chosen as in ref . it can be seen that for @xmath110 and for @xmath79 the fit reproduces almost exactly the full calculation , while for the intermediate couplings , in particular for @xmath111 , the fit is quite bad . in fig . [ fig3 ] the fitted frequency @xmath105 and damping rate @xmath104 are shown as functions of the parameter @xmath101 . the third panel shows the corresponding rms error of the fit , defined as @xmath112 . there is a couple of striking similarities with the experimental results : at strong coupling , where @xmath92 is large and hence @xmath113 is small , the fit gives the hydrodynamic frequency . with decreasing coupling , i.e. , increasing @xmath113 , the damping increases and the frequency of the hydrodynamic mode is shifted downwards . the quality of the fit becomes quite bad ( this corresponds probably to the shaded area in ref . @xcite ) , and at some critical coupling strength the fit jumps to the collisionless ( normal - fluid ) mode . as seen in the experiment , the damping rate of the collisionless mode is weaker than that of the hydrodynamic mode before the jump . ranging from @xmath114 to @xmath115 ( from bottom to top ) for fixed temperature @xmath86.,width=260 ] , damping rate @xmath104 , and corresponding rms error of the fit ( definition see text ) for the radial quadrupole mode as functions of @xmath101 for fixed temperature @xmath86 . the dotted lines in the upper panel indicate the prediction of superfluid hydrodynamics ( @xmath93 ) and the prediction for an ideal fermi gas ( @xmath48).,width=260 ] it should be noticed that with the present choice of parameters , the jump happens at @xmath116 , corresponding to @xmath117 . this means that in the center of the trap the system is still superfluid , but since this region contributes only very little to the quadrupole moment , the quadrupole mode behaves almost as if the system was in the collisionless normal - fluid phase . in spite of the surprising qualitative agreement with the findings of ref . @xcite , there is no exact agreement on a quantitative level . first of all , in the experiment the sudden transition from hydrodynamic to collisionless normal - fluid behavior happens already at a stronger coupling than in the calculation . the position of the jump depends on the temperature , and if the temperature in the experiment had been higher than ours ( @xmath118 ) , this could explain the difference . in addition , one should remember that the critical temperature predicted by bcs theory is too high . the second problem is the damping rate : the experimental damping rates are larger than the theoretical ones by roughly a factor of 2 near the jump , and even worse in the case of the collisionless normal - fluid mode far away from the transition . in fact , the experiment is not in the collisionless limit and collisions still play a role . in order to improve the theory in this respect , one would have to include a collision term similar to that of the boltzmann equation into the quasiparticle transport equation . the small anharmonicity of the experimental trap potential could be an additional source of damping . however , a calculation where the radial harmonic potential , @xmath119 , was replaced by a gaussian one , @xmath120 , with a realistic width @xmath121 mm , showed that this effect can explain only @xmath122 of the experimentally observed damping in the normal phase . the third point concerns the frequencies : on the one hand , although the theory gives a down - shift of the hydrodynamic mode before the transition , it is not as strong as the down - shift observed in the experiment . on the other hand , the theoretical frequency of the collisionless normal - fluid mode above the transition is too low , because the mean - field has been neglected . in the case of the radial quadrupole mode the velocity field is always irrotational and it does not change significantly from the superfluid to the normal - fluid phase . the different frequencies are a consequence of the fermi surface distortion during the oscillation in the collisionless normal - fluid phase . this is different from the `` scissors mode , '' which has also been studied in a recent experiment @xcite . in this experiment , the trap was deformed in the @xmath123 plane : @xmath124 $ ] , with @xmath125 . the scissors mode corresponds to a motion where the deformation of the atom cloud in the @xmath123 plane is not aligned with that of the trap , so that the angle between them oscillates . in the normal - fluid phase , such a motion can be realized by rotating the gas back and forth around the @xmath126 axis [ velocity field @xmath127 , but also by an irrotational motion [ velocity field @xmath128 . in an ideal gas , these two kinds of motion correspond to two distinct modes with frequencies @xmath129 and @xmath130 , respectively , given by @xmath131 . in the superfluid phase , of course , the rotational mode is suppressed and only the irrotational one exists . in the zero - temperature limit , its frequency approaches the hydrodynamic value @xmath132 @xcite . the excitation of the scissors mode consists in turning the trap potential around the @xmath126 axis by a small angle . in the small - amplitude limit , this corresponds to an excitation operator @xmath133 . the excitation spectra of the scissors mode for different temperatures , which are obtained analogously to those of the radial quadrupole mode ( i.e. fourier transform of @xmath134 after a @xmath55-like perturbation ) , are shown in fig . [ fig4 ] for different temperatures ranging from @xmath135 to @xmath136 ( from bottom to top ) . the spectra have been folded with a lorentzian of width @xmath137 hz . the arrows indicate the predicted frequencies in the hydrodynamic ( @xmath138 ) and collisionless limit ( @xmath139 ) , respectively , width=260 ] for the case of @xmath140 atoms in a trap with @xmath141 hz , @xmath142 hz , @xmath143 hz , and @xmath79 . at @xmath144 , the spectrum has a single broad peak at the hydrodynamic frequency . the damping rate ( width ) is comparable with that of the hydrodynamic radial quadrupole mode at @xmath144 . with increasing temperature , the two collisionless normal - fluid modes appear at the predicted frequencies , while the strength of the hydrodynamic mode decreases and its frequency is slightly shifted downwards . at @xmath145 and @xmath88 the scissors - mode spectrum has three pronounced peaks . there is a striking analogy between these results and the experimental observations concerning the scissors mode in a bec with a thermal cloud , where the bec oscillates with @xmath146 which is slightly shifted downwards while the thermal cloud oscillates with @xmath147 @xcite . the results of fig . [ fig4 ] are quite different from those found in the experiment @xcite . however , this is not surprising , since the experiment was not done in the bcs limit but for @xmath148 , where in particular the collisions are very important . as one can see in fig . 2 of ref . @xcite , the phase transition in the experiment does not lead directly from the superfluid to the collisionless regime , but the system enters first a collisionally hydrodynamic regime ( i.e. , the mode frequency does not change ) and reaches the collisionless phase only at much higher temperature . as shown in ref . @xcite , the transition from collisional hydrodynamic to the collisionless regime leads to a continuous increase of the upper ( irrotational ) branch of the scissors mode . this explains why in the experiment no sign of a sudden jump of the scissors mode frequency has been observed . it would be interesting to see if in the bcs regime , e.g. at @xmath79 , the experiment would observe a similar jump as in the case of the quadrupole mode . in the present paper , the behavior of the radial quadrupole and of the scissors mode across the bcs phase transition was theoretically studied in the framework of a semiclassical theory , in which superfluid hydrodynamics is coupled to a quasiparticle transport theory for the normal component . the theory proved to be applicable to realistic situations ( large numbers of atoms , arbitrary trap geometry ) , and it provides an explanation of the surprising behavior of the radial quadrupole mode as a function of @xmath101 as observed by the innsbruck group @xcite : the damping near the bcs phase transition is due to the coupling between the hydrodynamic mode and the thermally excited quasiparticles . in addition to the damping , this coupling leads to a two - peak spectrum , the two peaks belonging , respectively , to the hydrodynamic mode of the superfluid component and the collisionless mode of the normal - fluid component . the sudden jump from the hydrodynamic to the collisionless frequency reported in ref . @xcite is probably a consequence of the fit with a single frequency in the analysis of the experiment . it should be mentioned that an alternative explanation of the damping was suggested in ref . @xcite : the frequency shift and damping are not necessarily a signal for a superfluid - normal phase transition . they could also be related to the breakdown of superfluid hydrodynamics when @xmath149 is of the same order of magnitude as the pairing gap @xmath150 . in ref . @xcite , effects of this kind were quantitatively studied by using a fully quantum mechanical formalism [ quasiparticle random phase approximation , ( qrpa ) , corresponding to the time - dependent bogoliubov - de gennes equations in the case of small deviations from equilibrium ] , and it was shown that indeed , even at zero temperature , considerable damping and deviations from the hydrodynamic frequency can occur if @xmath151 is not small . although such effects may also play a role in the recent experiments ( maybe they can give rise to an additional downshift of the hydrodynamic quadrupole mode , which would help to quantitatively explain the data ) , they are probably less important than the coupling between the hydrodynamic mode and the thermally excited quasiparticles considered here : in ref . @xcite it was shown that for a ratio of mode frequency to zero - temperature gap of @xmath152 ( which is similar to the most unfavorable ratio of all the cases considered in the present paper ) , the results of the semiclassical theory and of the quantum mechanical qrpa calculation for the quadrupole mode in a spherical trap are in good agreement , even if the finite - temperature gap is smaller . possible extensions of this theory which are needed for giving more quantitative results at the limit between the bcs and the crossover regime ( @xmath153 ) have been mentioned . these are , on the one hand , the inclusion of quasi - particle collisions in order to reproduce the observed damping , and on the other hand the inclusion of the single - particle self - energy in order to account for the `` mean field '' energy shift . however , these extensions may not be enough to give a complete description of the bcs - bec crossover ( @xmath3 ) , including the unitary limit ( @xmath154 ) . in this regime , the normal - fluid component of the system is not only made of thermally excited quasiparticles , which are included in the present theory , but also of pair fluctuations @xcite , which persist even above the critical temperature @xmath92 up to the `` pair breaking temperature '' @xmath155 and which can not easily be accounted for in the present theoretical framework . i thank g.m . bruun and e.r . snchez guajardo for discussions , comments and suggestions . schunck , m.w . zwierlein , a. schirotzek , and w. ketterle , phys . * 98 * , 050404 ( 2007 ) . j. kinast , s.l . hemmer , m.e . gehm , a. turlapov , and j.e . thomas , phys . * 92 * 150402 ( 2004 ) . m. bartenstein , a. altmeyer , s. riedl , s. jochim , c. chin , j. hecker denschlag , and r. grimm , phys . lett . * 92 * , 203201 ( 2004 ) . j. kinast , a. turlapov , and j.e . thomas , phys . a * 70 * , 051401(r ) ( 2004 ) . a. altmeyer , s. riedl , m.j . wright , c. kohstall , j. hecker denschlag , and r. grimm , phys . a * 76 * , 033610 ( 2007 ) . wright , s. riedl , a. altmeyer , e.r . snchez guajardo , j. hecker denschlag , and r. grimm , phys . lett . * 99 * , 150403 ( 2007 ) ; e. taylor and a. griffin , phys . a * 72 * , 053630 ( 2005 ) . o. betbeder - matibet and p. nozires , ann . ( n.y . ) * 51 * , 392 ( 1969 ) . m. urban and p. schuck , phys . a * 73 * , 013621 ( 2006 ) . m. urban , phys . a * 75 * , 053607 ( 2007 ) . a. perali , p. pieri , g.c . strinati , and c. castellani , phys . b * 66 * 024510 ( 2002 ) . p. nozires and s. schmitt - rink , j. low . phys . * 59 * , 195 ( 1985 ) . bruun and b.r . mottelson , phys . . lett . * 87 * , 270403 ( 2001 ) . m. grasso , e. khan , and m. urban , phys . a * 72 * , 043617 ( 2005 ) . d. guery - odelin and s. stringari , phys . * 83 * , 4452 ( 1999 ) . o. marag , g. hechenblaikner , e. hodby , and c. foot , phys . rev . * 86 * , 3938 ( 2001 ) . bruun and h. smith , phys . a * 76 * , 045602 ( 2007 ) . r. combescot and x. leyronas , phys . rev . 93 * , 138901 ( 2004 ) . y. ohashi , n. fukushima , h. matsumoto , e. taylor , and a. griffin , physica c * 468 * , 599 ( 2008 ) .	the excitation spectra of the radial quadrupole and scissors modes of ultracold fermi gases in elongated traps are studied across the bcs superfluid - normal phase transition in the framework of a transport theory for quasiparticles . in the limit of zero temperature , this theory reproduces the results of superfluid hydrodynamics , while in the opposite limit , above the critical temperature , it reduces to the collisionless vlasov equation . in the intermediate temperature range , the excitation spectra have two or three broad peaks , respectively , which are roughly situated at hydrodynamic and collisionless frequencies , and whose strength is shifted from the hydrodynamic to the collisionless modes with increasing temperature . by fitting the time dependent quadrupole deformation with a damped oscillation of a single frequency , we can understand the `` jump '' of the frequency of the radial quadrupole mode as a function of interaction strength which has recently been reported by the innsbruck group .
in the hierarchical picture of galaxy formation , massive galaxies are formed through merger and accretion . satellite galaxies associated with massive host galaxies , memorize the history of their accretion and interaction with host galaxies . in addition , they are distributed out to a few virial radii of the host galaxy . therefore , they are excellent probes of the dark matter halos surrounding the host galaxies . a great deal of effort has been made to understand the spatial distribution of satellite galaxies around their host galaxies . the radial distribution of the observed satellites was found to be more concentrated than that of subhalos around galaxy - sized halos in n - body simulations @xcite , but consistent with dark matter distribution @xcite . the studies of color dependence on the radial distribution of satellites revealed that the radial distribution of red satellites is more centrally concentrated than that of blue satellites @xcite . similarly , the radial distribution of early morphological type satellites is more centrally concentrated than that of late - type satellites @xcite . the angular distribution of satellite galaxies is also an issue , which has been extensively addressed ( e.g. , @xcite ) . another interesting issue is the kinematic properties of satellite galaxies . satellite galaxies can be divided into two classes based on their orbital motion : those in prograde orbit that rotate around the host in the same direction as the rotation of the host , and those in retrograde orbit that rotate in the opposite direction . several important questions have been raised for these two classes : ( 1 ) what the number ratio of galaxies in prograde orbit to those in retrograde orbit is ; ( 2 ) what the amplitude of mean bulk rotation of satellites is ; and ( 3 ) whether there is a difference in physical properties between them . @xcite searched for the satellites around late - type spiral galaxies with a luminosity similar to the milky way , and found 69 satellites around 45 host galaxies ( @xmath71.5 satellites per host ) . for 35 satellites whose orbital velocity directions were determined , they detected a small amount of rotation for the satellite systems , @xmath8 km s@xmath3 , in the same sense as the rotation of hosts disk . they also found that the number of satellites in prograde orbit is nearly equal to that in retrograde orbit . however , in ngc 5084 , @xcite detected nine satellite galaxies , and found a clear excess of satellites in retrograde orbit ( eight in retrograde orbit and one in prograde orbit ) . using the expanded catalog of 115 satellites around 69 host galaxies , @xcite found that the number of galaxies in prograde orbit is almost the same as that of galaxies in retrograde orbit . recently , @xcite measured the direction of rotation for some host galaxies in the list of satellite systems of @xcite , and secured 76 satellites associated with 43 host galaxies of which orbital directions were determined . with this sample , they found that the fraction of galaxies in prograde orbit ( @xmath9 ) is @xmath10 , which is slightly larger than the previous results ( @xmath11 in @xcite ) , but is broadly consistent with those from their @xmath12-body simulation ( @xmath13 ) . in the catalog of satellite systems extracted from the sloan digital sky survey ( sdss ; @xcite ) data release 6 @xcite , @xcite determined the direction of rotation of some host galaxies by spectroscopic observations . they obtained the data for 78 satellites associated with 63 hosts . combining these data with those for the satellites in @xcite , they found that the fraction of galaxies in prograde orbit is still @xmath14 , and mean bulk rotation of satellites is @xmath15 km s@xmath3 , in the same sense as the rotation of hosts disk . interestingly , they found that the peculiar velocity distribution of satellites is not a single gaussian , but is described as the sum of two gaussians . as summarized above , most studies focused on the first two issues , the fraction of galaxies in prograde orbit and the amplitude of mean bulk rotation of satellites . the difference in physical properties between them , which is very important to understand the formation of satellite systems , has not been explored much . moreover , the direction of rotation of host galaxy that is essential to know the orbit of satellites , has been determined by the expensive spectroscopic observations . in this paper , we determine the direction of rotation of host galaxies using the color images provided by the sdss without conducting spectroscopic observations ( see [ spin ] for details ) , and present the results of a study on the dependence of various physical properties of satellite galaxies on their orbits ( prograde and retrograde orbits ) . in this paper we mean by ` satellites ' as the galaxies that are much fainter than their host galaxies that are isolated from other bright galaxies . therefore , they are like the conventional satellites associated with the milky way or andromeda , and are rather different from the ` satellites ' used in the n - body simulations or group / cluster studies where all galaxies ( or dark halos ) other than the brightest one ( or the most massive one ) are called satellites . section [ data ] describes the observational data used in this study . orbital dependence of the physical parameters of satellites are given in [ results ] . discussion and summary are given in [ discuss ] and [ sum ] , respectively . we use a spectroscopic sample of galaxies in the sdss dr7 @xcite . the physical parameters of galaxies that we consider in this study are @xmath16-band absolute petrosian magnitude ( @xmath17 ) , morphology , ( @xmath6 ) color , equivalent width of @xmath18 emission line [ w(h@xmath19 ) ] , color gradient in ( @xmath20 ) color , concentration index ( @xmath21 ) , internal velocity dispersion ( @xmath22 ) , and petrosian radius in @xmath23-band ( @xmath24 ) . here we give a brief description of these parameters . the @xmath16-band absolute magnitude @xmath17 was computed using the formula , @xmath25-k(z)+e(z),\ ] ] where @xmath26 is a comoving distance at redshift @xmath27 , @xmath28 is @xmath29-correction , and @xmath30 is the luminosity evolution correction . we adopt a flat @xmath31cdm cosmology with density parameters @xmath32 and @xmath33 . the rest - frame absolute magnitudes of individual galaxies are computed in fixed bandpasses , shifted to @xmath34 , using galactic reddening correction @xcite and @xmath29-corrections as described by @xcite . the evolution correction given by @xcite , @xmath35 , is also applied . the distance @xmath26 has a unit of @xmath36mpc and the corresponding @xmath37 in @xmath17 will be omitted in this paper . we adopt the galaxy morphology information from the sdss dr7 release of the korea institute for advanced study value - added galaxy catalog ( kias - vagc ; choi 2010 , in prep . ) . in this catalog galaxies are divided into early ( ellipticals and lenticulars ) and late ( spirals and irregulars ) morphological types based on their locations in the ( @xmath6 ) color versus ( @xmath20 ) color gradient space and also in the @xmath23-band concentration index space @xcite . the resulting morphological classification has completeness and reliability reaching 90% . we perform an additional visual check of the color images of galaxies to correct misclassifications by the automated scheme . in this procedure we changed the types of the blended or merging galaxies , blue but elliptical - shaped galaxies , and dusty edge - on spirals . the @xmath38 color was computed using the extinction and @xmath29-corrected model magnitude . the superscript 0.1 means the rest - frame magnitude @xmath29-corrected to the redshift of 0.1 , and will subsequently be dropped . we adopt the values of @xmath39 color , concentration index ( @xmath21 ) , and petrosian radius @xmath24 computed for the galaxies in kias - vagc ( choi 2010 , in prep . ) . the @xmath39 color gradient was defined by the color difference between the region with @xmath40 and the annulus with @xmath41 , where @xmath42 is the galactocentric radius and @xmath24 is the petrosian radius estimated in the @xmath23-band image . to account for the effects of flattening or inclination of galaxies , elliptical annuli were used to calculate the parameters . the ( inverse ) concentration index is defined by @xmath43 , where @xmath44 and @xmath45 are semimajor axis lengths of ellipses containing @xmath46 and @xmath47 of the petrosian flux in the @xmath23-band image , respectively . the velocity dispersion values of galaxies are adopted from the dr7 release of new york university ( nyu ) vagc @xcite and values of @xmath18 equivalent width are taken from the dr7 release of max - planck - institute for astrophysics ( mpa)/john hopkins university ( jhu ) vagc @xcite . finding isolated satellite systems consists of two steps : identifying isolated galaxies and finding satellites associated with the isolated galaxies . we consider all sdss galaxies with redshifts less than 0.03 as the host candidates . only the relatively nearby galaxies are considered because the visual determination of the sense of rotation is feasible only when the galaxy image shows detailed internal features . for each galaxy its neighbors are searched . the neighbors of a target galaxy with a @xmath16-band absolute magnitude @xmath17 are the galaxies with an absolute magnitude brighter than @xmath17 + 1 and a velocity difference less than @xmath48 . we adopt @xmath49 km s@xmath3 for early - type target galaxies , and @xmath50 km s@xmath3 for late - type galaxies based on the distributions of pairwise velocity difference of the sdss galaxies ( see fig . 2 of @xcite and fig . 1 of @xcite ) the neighbors are searched in the redshift interval from @xmath51 to @xmath52 . when a target galaxy has a projected distance to its neighbor galaxies ( @xmath53 ) larger than max[@xmath54,@xmath55 for all neighbors , it is selected as an isolated galaxy . here @xmath56 is the virial radius of a galaxy ( see eq . ( 5 ) of @xcite ) . once isolated host candidates are determined , the satellites associated with them are searched for . a galaxy qualifies as a satellite if it has an absolute magnitude fainter than @xmath17(host)+1 , velocity difference less than @xmath57 km s@xmath3 , and projected separation of the galaxy with the host @xmath58min[@xmath53 @xmath59/(@xmath60 ) , @xmath61 where @xmath53 is a projected separation between the host and host s neighbor . the first limit divides the distance between the host and neighbor in proportion to their virial radii . the second limit prohibits the galaxies within the virial radii of the neighbor galaxies from being selected as satellites . in our sdss galaxy sample we find 8904 satellites associated with 3515 isolated host galaxies . to determine the sense of rotation of host galaxies from their images we first assume that all spiral galaxies have trailing arms @xcite and determine the receding side of the major axis from their color images . the part of minor axis behind the bulge often appears smooth due to the foreground bulge stars while the side of the minor axis closer than the bulge often shows interstellar dust patches of relatively higher sharpness @xcite . when it is not possible to determine the sense of rotation due to various reasons , we simply drop the host galaxy . very early - type galaxies with smooth disks , very late - type galaxies with vanishing bulge , and nearly face - on or edge - on disk galaxies are difficult to assign the spin direction . after we remove all these cumbersome cases , we can determine the spin vector directions for 165 host galaxies that have 535 satellite galaxies . we supplemented our sample by adding host galaxies of which rotation sense was determined from the spectroscopic observations in the literature @xcite . we compiled 390 satellites associated with 116 hosts . there are 51 hosts in common between the hosts of which rotation sense was determined in this study and those in the literature . we have compared our determination of the rotation sense with that in the literature for these common objects , and found that there is only one galaxy ( ngc 5894 ) that shows a disagreement between the two . we kept the rotation sense for this object determined from the spectroscopic observation in the literature . in total , we have 727 satellites assigned to 230 hosts whose spin vector directions are known . among 727 satellites , we eliminated 45 spurious sources ( e.g. , faint fragments of bright galaxies , diffraction spikes of bright stars ) that were included in the list of satellites ( originally included in the spectroscopic sample of galaxies in sdss ) by inspecting color images of the satellites . in the result , 682 satellites with 221 host galaxies remain . it is noted that our list contains some satellite systems whose host galaxies are early types because their spin directions were determined from the spectroscopic observations . since the morphology of satellite galaxies is strongly affected by the morphology of host galaxies as found in @xcite , we further divide the satellites into two categories ( the cases of early- and late - type hosts ) and obtain 197 late - type host galaxies having 579 satellite galaxies and 24 early - type hosts with 103 satellites . our analysis will be restricted to the satellite systems with late - type hosts except [ samp ] . in figure [ fig - ncut ] , we plot peculiar velocities of satellite galaxies as a function of absolute magnitudes of host galaxies . the peculiar velocity of a satellite is defined by , @xmath62 , where @xmath63 and @xmath64 are observed velocities of the satellite and the host , respectively . @xmath65 is defined by @xmath66 @xmath67 where pared is a position angle of the receding side of the host s major axis , and pas is the position angle of the satellite with respect to the host . all position angles are measured from north to east . therefore , the satellites with @xmath68 rotate around the host in the same direction as the rotation of the host ( prograde orbit ) and those with @xmath69 rotate in the opposite direction ( retrograde orbit ) . since the probability for a galaxy to become a satellite of a host decreases as the mass ( luminosity ) of the host decreases , we exclude some probable interlopers located outside the region bounded by dashed lines in figure [ fig - ncut ] . the dashed lines are determined as follows . the escape velocity of a satellite can be given by @xmath70 where @xmath71 is a mass of the host and @xmath4 is a projected distance to the host . the mass of the host galaxy is computed by @xmath72 where @xmath73 is the mass - to - light ratio and @xmath74 is the @xmath16-band luminosity of the host . we assume that @xmath73(early)=@xmath75(late ) at the same @xmath16-band luminosity , and that @xmath73 is constant with galaxy luminosity for a given morphological type @xcite . then we obtain @xmath76 in a given distance to the host galaxy ( e.g. , @xmath77 ) . we adopt @xmath78 km s@xmath3 and @xmath79 mag , and plot @xmath80 versus @xmath17 as dashed lines in figure [ fig - ncut ] . ) . , width=321 ] vs. projected distance to the host galaxy . median values in each radial bin and their errors are connected by solid lines . peculiar velocity distribution for the satellites ( b ) at @xmath81 and ( c ) at @xmath82 . the vertical dotted lines indicate zero velocity . solid lines indicate the best fit exponential functions of the distributions of satellites in prograde or retrograde orbits separately . solid line for the satellites in prograde orbit at @xmath81 is the best fit gaussian function . , width=321 ] the virial radius of a host galaxy ( @xmath59 ) is defined by the projected radius where the mean mass density @xmath83 within the sphere of radius @xmath16 is 200 times the critical density or 740 times the mean density of the universe , namely , @xmath84 since we adopt @xmath85 , @xmath86 where @xmath87 is the mean density of the universe . the value of mean density of the universe , @xmath88 , was adopted , where @xmath89 is the mass of a late - type galaxy with @xmath90 @xcite . according to our formula the virial radii of galaxies with @xmath91 and @xmath92 are 260 , 300 , and 350 @xmath36 kpc for early types , and 210 , 240 , and 280 @xmath36 kpc for late types , respectively . by rejecting probable interlopers as shown in figure [ fig - ncut ] , we finally secured 534 satellite galaxies associated with 191 late - type hosts and 101 satellites associated with 24 early - type hosts . we present the peculiar velocity distribution of all 534 satellite galaxies in figure [ fig - hist ] . the fraction of galaxies in prograde orbit ( @xmath9 ) is found to be @xmath93 , and the median value of the peculiar velocity distribution is @xmath94 km s@xmath3 . when we use only 262 satellites located near the major axis of the hosts ( @xmath95 ) , @xmath9 does not change and the median value of the distribution is @xmath96 km s@xmath3 . we fit the peculiar velocity distribution using the sum of two gaussian functions , and find the best - fit function with mean values of @xmath97 km s@xmath3(@xmath98 km s@xmath3 ) and @xmath99 km s@xmath3(@xmath100 km s@xmath3 ) . when we fit the peculiar velocity distribution of satellites in prograde or retrograde orbits separately using a single gaussian or exponential function , we find that the exponential function fits the distribution better than the gaussian function . the reduced @xmath101 for satellites in prograde orbit , are 2.1 and 0.5 for gaussian and exponential functions , respectively . the reduced @xmath101 for satellites in retrograde orbit are 2.5 and 0.7 for gaussian and exponential functions , respectively . to study whether or not the peculiar velocity distribution changes depending on the distance to the host we show the distribution for satellites at @xmath81 or at @xmath82 separately in figure [ fig - histrad ] . we use gaussian and exponential functions to fit the distributions of satellites in prograde and retrograde orbits , and show best - fit functions in figure [ fig - histrad](b - c ) . it is seen that the best - fit function for satellites in retrograde orbit is exponential both in the inner and outer regions , while that for satellites in prograde orbit changes from gaussian ( inner region ) to exponential ( outer region ) . ) computed from eq . ( [ eq - mag ] ) in @xcite . the satellites associated with the host galaxies within the solid box are selected as a luminosity subsample of satellites . , width=321 ] ) are filled circles , while the rests are crosses . number outside the parenthesis means the number of galaxies in each panel , and the number in the parenthesis indicates the number of galaxies within the luminosity subsample . , width=321 ] we plot @xmath16-band absolute magnitudes of host galaxies and their satellites as a function of redshift in figures [ fig - host ] and [ fig - sat ] . we made a luminosity subsample of satellites whose hosts have absolute magnitudes in the range of @xmath102 . vs. projected distance to the host galaxy for early - type satellites in prograde orbit ( ep ) , for early - type satellites in retrograde orbit ( er ) , for late - type satellites in prograde orbit ( lp ) , and for late - type satellites in retrograde orbit ( lr ) in the luminosity subsample . global median values are shown by horizontal dot - dashed lines , and the median value in each radial bin is connected by the solid line . vertical dashed line indicates one virial radius of the host.,width=321 ] in figure [ fig - vpec ] , we plot peculiar velocities of satellites as a function of a distance to host galaxies . it shows that global median values of @xmath103 do not significantly vary depending on the orbit or morphology . however , there exists an interesting trend that @xmath103 of the late - type satellites in prograde orbit ( lp ) increases as they approach the host , while @xmath103 of those in retrograde orbit ( lr ) decreases . at the merger scales with @xmath104 the orbital velocity decreases for both lp and lr cases . ) among early - type satellites ( _ b _ ) as a function of the projected distance to the host galaxies . distribution of late - type satellites in prograde ( filled triangles ) and retrograde ( open upside - down triangles ) orbits ( _ c _ ) and @xmath9 among late - type satellites ( _ d _ ) . , width=321 ] ( cos@xmath105 ) ) and perpendicular ( @xmath22(sin@xmath105 ) ) to the major axis of hosts as a function of the host axis ratio . solid and dashed lines indicate satellites in prograde and retrograde orbits , respectively . , width=321 ] in figure [ fig - frac ] , we plot the radial distribution of satellites in prograde and retrograde orbits . we also show the fraction of galaxies in prograde orbit ( @xmath9 ) as a function of the projected distance to host galaxies . the top panel shows that satellites in prograde orbits are located closer to their hosts compared to those in retrograde orbits for the case of early - type satellites . however , we do not find such dependence of the spatial distribution of late - type satellites on their orbits ( see panel c ) . as a result , @xmath9 of early types increases as @xmath4 decreases , but @xmath9 of late types remains constant ( @xmath106 ) within the error bar . to study whether or not satellites are concentrated on the plane of host galaxies disk , we plot the ratio of the dispersion of satellite distribution along ( @xmath22(cos@xmath105 ) ) and perpendicular ( @xmath22(sin@xmath105 ) ) to the major axis of the host as a function of the ratio of minor axis to the major axis of the host ( @xmath107 ) in figure [ fig - prob ] . it shows neither of the distributions of prograde or retrograde orbit satellites is aligned with late - type host galaxies . , but for absolute magnitude @xmath17 . , width=321 ] figure [ fig - mag ] shows absolute magnitudes of the satellites in the luminosity subsample as a function of the projected distance to host galaxies . each panel represents different subsamples depending on morphology and orbit . for both early and late types , galaxies in retrograde orbit , on average , appear to be slightly brighter than those in prograde orbit . a statistical test with monte carlo simulation to compute the significance of the difference in the median of the physical quantity ( to be discussed in the end of this section ) confirms this difference for only late types with a significance level of 97@xmath108 . when we compare satellites in each radial bin , the monte carlo test indicates with a significance level of 99@xmath108 that median absolute magnitudes of late - type galaxies in the outer region ( @xmath109 ) are different depending on the orbit . to investigate the difference in the galaxy luminosity depending on the orbit in more detail , we show the distribution of absolute magnitudes in figure [ fig - lum ] . we count the number of satellites within the redshift limit defined by the apparent magnitude limit in figure [ fig - sat ] , and do not correct the count for different volumes . it is seen that bright satellites ( @xmath110 ) in retrograde orbit are more numerous compared to those in prograde orbit . the test with monte carlo simulation confirms this result with a significance level of 97@xmath108 in the sense that median absolute magnitudes of late types in prograde and retrograde orbits are significantly different . we summarize the results of statistical comparison of several physical parameters between satellites in prograde and retrograde orbits and between satellites in the inner and outer regions in tables [ tab - all ] and [ tab - inout ] , respectively . the statistical significance of the difference in the median or the dispersion of physical quantities , is computed by monte carlo simulations . for a given physical parameter , we construct two subsamples ( with the same number of galaxies as in the actual prograde and retrograde samples ) by randomly drawing the observed parameters from the whole early - type or late - type satellite galaxy sample and compute the median and the dispersion of each subsample . the resulting two subsamples will have the same medians and standard deviations on average . these random subsamples will tell us about whether or not the difference in the parameter between prograde and retrograde satellite samples is statistically significant . we construct 1000 trial data sets and compute the fraction of simulated data sets in which the difference of the physical quantity is larger than or equal to that based on the real data ( f@xmath111 ) . the significance levels of the difference defined by 100(1@xmath112f@xmath111 ) ( % ) are given in tables . we have applied the statistical tests to all the physical parameters , but only show statistically significant results ( significance level above 95@xmath108 ) in the tables . [ cols="^,^,^,^,^,^,^,^,^ " , ] _ column descriptions_. ( 1 ) : physical parameters . ( 2 ) : morphological types ( e : early types , l : late types ) . ( 3 ) : significance level ( @xmath108 ) of the difference of the median of the physical parameter of prograde orbit satellites in the inner and outer regions determined from monte carlo simulation . ( 4 ) : significance level ( @xmath108 ) of the difference of the dispersion of the physical parameter of prograde orbit satellites in the inner and outer regions determined from monte carlo simulation . ( 5 ) : same as column ( 3 ) , but for retrograde orbit satellites . ( 6 ) : same as column ( 4 ) , but for retrograde orbit satellites . ( 7 ) : relevant figure . statistically significant values ( @xmath113 ) are represented in bold face . , but for ( @xmath6 ) color.,width=321 ] figure [ fig - ur ] shows ( @xmath6 ) colors of satellite galaxies divided by their morphologies and orbital motion as a function of the projected distance to the hosts . the late - type satellites in prograde and retrograde orbits have a similar global median value , and both show a hint of an increase in ( @xmath6 ) color with decreasing @xmath4 . in the monte carlo test for the radial gradient we divide each sample into inner ( @xmath114 ) and outer ( @xmath115 ) regions , and test if the difference in the physical parameter in two regions is statistically significant . the random subsamples are generated in such a way that a parameter value is randomly drawn from the combined sample and assigned to a galaxy in inner or outer region . the resulting subsamples will have no radial gradient in the parameter and have the same dispersions on avergae . a test with such monte carlo simulation confirms that the change of ( @xmath6 ) color with @xmath4 for late types in prograde orbit ( see tab . [ tab - inout ] ) is significant at a 99% confidence level . it is noted that the increase of ( @xmath6 ) colors for those in retrograde orbit is very strong in the innermost region ( @xmath116 ) even though the overall gradient is not signifiant . , but for equivalent width of h@xmath19 line.,width=321 ] in figure [ fig - halpha ] , we plot the equivalent width of h@xmath19 lines of satellite galaxies , which is a useful measure of star formation activity ( sfa ) . for late - type satellites , the median values of w(h@xmath19 ) for two subsamples are similar , and those do not seem to change with @xmath4 for both samples even though their ( @xmath6 ) colors become redder at smaller @xmath4 as seen in figure [ fig - ur ] . however , the test with monte carlo simulation shows that the difference in the dispersion of w(h@xmath19 ) between late - type satellites in prograde and retrograde orbits is remarkable with a significance level of of 97% ( see tab . [ tab - all ] ) . the difference in the dispersion of w(h@xmath19 ) of retrograde orbit satellites in the inner and outer regions is also noted ( see tab . [ tab - inout ] ) . , but for ( @xmath20 ) color gradient.,width=321 ] figure [ fig - gigrad ] represents ( @xmath20 ) color gradients of satellite galaxies . the positive @xmath39 color gradient means a bluer color in the central region of a galaxy than that in the outer region . for early types , the difference of median ( @xmath20 ) color gradients of satellites in prograde orbit ( ep ) between in the inner and outer regions is noticeable , which is supported by the test with monte carlo simulation ( see tab . [ tab - all ] ) . for late types , the test with monte carlo simulation suggests that the median and its dispersion of ( @xmath20 ) color gradients of satellites in prograde and retrograde orbits are different significantly in the whole / outer region . , but for concentration index @xmath21.,width=321 ] figure [ fig - cin ] shows the distribution of the concentration index @xmath21 for satellite galaxies divided by their morphologies and orbital motion . small value of @xmath21 means that the light of a galaxy is more concentrated in the central region . it is found that the global median and the scatter of @xmath21 values for early - type satellites are not significantly different depending on their orbits . however , it is noted that there are few early types in retrograde orbit with large @xmath21 values ( @xmath117 ) in the inner region ( @xmath118 ) , and few early types in prograde orbit with large @xmath21 values ( @xmath117 ) are found in the outer region ( @xmath109 ) . for late types , it is seen that the difference in the dispersion of @xmath21 between late - type satellites in the prograde and retrograde orbits ( see tab . [ tab - all ] ) . however , the global median values of @xmath21 for two subsamples are similar , and the median values for both subsamples do not show any significant variation with @xmath4 . , but for central velocity dispersion @xmath22 . , width=321 ] , but for petrosian radius @xmath24 . , width=321 ] in figure [ fig - vdisp ] , we plot central velocity dispersions of satellite galaxies . the median values for early types in prograde and retrograde orbits are similar and do not show any significant variation with @xmath4 . for late types , any significant difference between two subsamples of late - type satellites is not found except for the scatter of @xmath22 values . figure [ fig - rpet ] represents the @xmath23-band petrosian radius @xmath24 of satellite galaxies . the @xmath24 for the early - type satellites in the inner region ( @xmath119 ) appears to be larger than that in the outer region ( @xmath109 ) ( see tab . [ tab - inout ] ) , while no significant difference between the late - type satellites in prograde and retrograde orbits is not seen . , but for the satellites associated to host galaxies whose spin direction is determined from the spectroscopic observations.,width=321 ] , but for the satellites associated to early - type hosts.,width=321 ] we found in figures [ fig - halpha]-[fig - gigrad ] and table [ tab - all ] that the median or its dispersion of w(h@xmath19 ) and ( @xmath20 ) color gradients can be different depending on the orbital motion of late - type satellites . the sfa of early - type satellites in retrograde orbit ( er ) also appears to be stronger than those in prograde orbit in the outer region ( @xmath120 ) as seen in figures [ fig - ur ] and [ fig - halpha ] . moreover , ( @xmath6 ) colors of late - type satellites increase as they approach their host galaxies for both satellites in prograde and retrograde orbits , and ( @xmath6 ) colors of late - type satellites in retrograde orbit in the innermost region ( @xmath116 ) are remarkably large . these differences depending on the orbit can be attributed to the different origin of satellite galaxies , or to the different strength of physical processes that they experience through interactions with host galaxies . to check whether or not these results are strongly affected by the satellites associated to the host galaxies whose spin direction is determined from the sdss color images in this study , in figure [ fig - urspec ] , we plot ( @xmath6 ) colors versus the projected distance to the hosts using the satellites associated to the host galaxies whose spin direction is determined from the spectroscopic observations . for both early and late types , we do not find any significant difference of ( @xmath6 ) colors between satellites in prograde and retrograde orbits , which is consistent with the result when we use the combined sample of satellites in the host galaxies whose spin direction is determined from the color images or the spectroscopic observations ( see fig . [ fig - ur ] ) . in addition , the late - type satellites in prograde orbits ( lp ) show an increase in ( @xmath6 ) color as decreasing @xmath4 , which is again similar to the case of late types in prograde orbits using the combined sample of satellites ( see lp in fig . [ fig - ur ] and tab . [ tab - inout ] ) . in the result , the results of sfa using the combined sample of satellites are not affected strongly by the satellites associated to the host galaxies whose spin direction is determined from the color images in this study . since the physical parameters of satellite galaxies can be strongly affected by the morphology of host galaxies ( e.g. , @xcite ) , it is also interesting to compare satellite systems that have early- and late - type host galaxies . to investigate the sfa of satellites around early - type hosts we use a sample of 101 satellites associated with 24 early - type hosts , and show their ( @xmath6 ) colors in figure [ fig - urearly ] . it is seen that ( @xmath6 ) colors of late - type satellites become redder as @xmath4 decreases , which was also seen in figure [ fig - ur ] or [ fig - urspec ] . for early - type satellites , the small number of satellites prevents us from drawing any firm conclusion . previously , @xcite and @xcite found that galaxy properties strongly depend on the distance and the morphology of the nearest neighbor galaxy . for example , the sfa of galaxies is enhanced as they approach late - type neighbors of comparable luminosity , while it is reduced as they approach early - type neighbors . the structure parameters of galaxies change with the distance to the nearest neighbor galaxy in the same direction independently of neighbor s morphology . galaxy properties start to change when galaxies approach their nearest neighbor galaxy closer than the virial radius ( @xmath121 ) of the nearest neighbor galaxy and to change significantly at @xmath122 where the galaxies in pair start to merge . they interpreted this phenomenon as a result of hydrodynamic interactions within the virial radius of the galaxy plus dark halo system . the dependence of satellite properties as a function of the distance to the host galaxy in this study can be compared with the case of late - type neighbors in @xcite . since we restrict our analysis to the case of late - type host galaxy , it is expected that the sfa of both early- and late - type satellites increases as they approach the host . however , we found in [ sfa ] that ( @xmath6 ) color becomes redder ( fig . [ fig - ur ] ) and @xmath123 ) hardly changes ( fig . [ fig - halpha ] ) even when late - type satellites approach their late - type hosts . this may be because , unlike the galaxies considered by @xcite , the satellites in this study are much fainter and less massive than their hosts . as a result they may not be able to pick up cold gas from their host much , and the enhancement of sfa through interactions with their host is counterbalanced by the quenching by the hot halo gas of the host . on the other hand , the environment of satellite galaxies can be compared with that of galaxies in galaxy clusters because the member galaxies in a galaxy cluster are bounded within the gravitational potential well of the cluster as the satellites are bounded by their host galaxy . @xcite studied the dependence of galaxy properties on the clustercentric radius and the environment attributed to the nearest neighbor galaxy using the sdss galaxies associated with the abell galaxy clusters . they found that the sfa of late - type galaxies decreases as they approach the cluster center , and suggested that hydrodynamic interactions with nearby early - type galaxies in clusters is the main drive to quenching the sfa of late types . most cluster galaxies are early types including the brightest cluster galaxies ( or cd galaxies ) , which might be comparable with the case of early - type host galaxy in the satellite system as seen in figure [ fig - urearly ] . the late - type satellites ( bottom panels in fig . [ fig - urearly ] ) become redder as they approach early - type host galaxies , which agrees to the case of galaxy clusters . previously , @xcite pointed out that there are two distinct populations in the halo globular clusters in the milky way : old and young halo clusters . two populations were originally divided by the horizontal branch morphology , but were found to be also separated by their orbital motion in the milky way : prograde orbit for old halo clusters and retrograde orbit for young halo clusters @xcite . therefore , the formation history of old halo clusters with prograde orbit is thought to be different from that of young halo clusters with retrograde orbit ( e.g. , @xcite vs. @xcite ) . for example , it was suggested that the clusters in retrograde orbit may have been accreted into the milky way ( e.g. , @xcite ) , and further that these clusters and galactic dsphs may have the same origin in accretion or in partially disruption of formerly large parent satellite galaxies ( e.g. , @xcite ) . therefore , two populations of satellite galaxies with different orbits in this study are also expected to have formed and evolved in different paths , which may be responsible for the different sfa depending on the orbit . the galactic satellite galaxies also can give us important clues for testing this scenario , but it is not achievable at this moment because of small number ( @xmath124 ) of galactic satellites whose orbital motion is determined . we have studied the orbital dependence of various galaxy properties in the satellite systems of galaxies in sdss . our primary results are summarized as follows . 1 . we have found the satellite systems of galaxies at @xmath0 , which consist of 8904 satellites associated with 3515 isolated host galaxies . we have determined the spin direction of late - type host galaxies by inspecting the sdss color images . by combining our results with those from the previous spectroscopic observations in the literature , we obtained 635 satellites associated with 215 host galaxies whose spin directions are known . the number ratio of satellites in prograde orbit to those in retrograde orbit is nearly equal to one , and the peculiar velocity distribution is found to be symmetric . the number ratio of satellites in prograde orbit to those in retrograde orbit ( @xmath9 ) appears to increase as the projected distance to the host galaxy decreases . however , we do not find any significant difference in the azimuthal distribution of satellites between those in prograde and retrograde orbits . the satellites in retrograde orbit , on average , appear to be slightly brighter than those in prograde orbit . the latter seems to get brighter as they approach late - type hosts , while the former does not show any significant change with the distance to the hosts . 5 . for early - type satellites in retrograde orbit , the sfa seems to decrease as they approach host galaxies . for late - type satellites , ( @xmath6 ) colors of both samples in prograde and retrograde orbits appear to increase as they approach the hosts , while the radial variations of the equivalent width of h@xmath19 lines for both samples are not significant . the structure parameters of satellites do not seem to be different significantly between galaxies in prograde and retrograde orbits . the authors are grateful to the anonymous referee for useful comments that improved the original manuscript . cbp acknowledges the support of the national research foundation of korea(nrf ) grant funded by the korea government(mest ) ( no . 2009 - 0062868 ) . funding for the sdss and sdss - ii has been provided by the alfred p. sloan foundation , the participating institutions , the national science foundation , the u.s . department of energy , the national aeronautics and space administration , the japanese monbukagakusho , the max planck society , and the higher education funding council for england . the sdss web site is http://www.sdss.org/. the sdss is managed by the astrophysical research consortium for the participating institutions . the participating institutions are the american museum of natural history , astrophysical institute potsdam , university of basel , cambridge university , case western reserve university , university of chicago , drexel university , fermilab , the institute for advanced study , the japan participation group , johns hopkins university , the joint institute for nuclear astrophysics , the kavli institute for particle astrophysics and cosmology , the korean scientist group , the chinese academy of sciences ( lamost ) , los alamos national laboratory , the max - planck - institute for astronomy ( mpia ) , the max - planck - institute for astrophysics ( mpa ) , new mexico state university , ohio state university , university of pittsburgh , university of portsmouth , princeton university , the united states naval observatory , and the university of washington .	we study the dependence of satellite galaxy properties on the distance to the host galaxy and the orbital motion ( prograde and retrograde orbits ) using the sloan digital sky survey ( sdss ) data . from sdss data release 7 we find 3515 isolated satellite systems of galaxies at @xmath0 that contain 8904 satellite galaxies . using this sample we construct a catalog of 635 satellites associated with 215 host galaxies whose spin directions are determined by our inspection of the sdss color images and/or by spectroscopic observations in the literature . we divide satellite galaxies into prograde and retrograde orbit subsamples depending on their orbital motion respect to the spin direction of the host . we find that the number of galaxies in prograde orbit is nearly equal to that of retrograde orbit galaxies : the fraction of satellites in prograde orbit is @xmath1% . the velocity distribution of satellites with respect to their hosts is found almost symmetric : the median bulk rotation of satellites is @xmath2 km s@xmath3 . it is found that the radial distribution of early - type satellites in prograde orbit is strongly concentrated toward the host while that of retrograde ones shows much less concentration . we also find the orbital speed of late - type satellites in prograde orbit increases as the projected distance to the host ( @xmath4 ) decreases while the speed decreases for those in retrograde orbit . at @xmath4 less than 0.1 times the host virial radius @xmath5 the orbital speed decreases in both prograde and retrograde orbit cases . prograde satellites are on average fainter than retrograde satellites for both early and late morphological types . the @xmath6 color becomes redder as @xmath4 decreases for both prograde and retrograde orbit late - type satellites . the differences between prograde and retrograde orbit satellite galaxies may be attributed to their different origin or the different strength of physical processes that they have experienced through hydrodynamic interactions with their host galaxies .
it is believed that active galactic nuclei ( agn ) are powered by accretion of matter on to massive black holes , and the observed uv / optical emission of agn is thought to be a thermal emission from the standard geometrically thin , optically thick accretion discs ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) . on the other hand , the power - law hard x - ray spectra of agn are most likely due to the inverse - compton scattering of soft photons on a population of hot electrons @xcite . in the accretion disc - corona model , such soft photons are from the cold disc , a fraction of which are compton scattered by the hot electrons in the corona above the cold disc to the hard x - ray energy band . the disc - corona model was extensively explored in many previous works ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) . in this disc - corona scenario , most gravitational energy is generated in the cold disc , probably through turbulence produced by the magneto - rotational instability @xcite . the magnetic fields generated in the cold disc are strongly buoyant , and a substantial fraction of magnetic energy is transported vertically to heat the corona above the disc with the reconnection of the fields ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? it was found that the temperature of the hot electrons in the corona is roughly around @xmath8 k , which can successfully reproduce a power - law hard x - ray spectrum as observed ( e.g. , * ? ? ? the physical processes of turbulence triggered by the amplified magnetic fields in the disc are very complicated , and quite unclear , though they are revealed to some extent with numerical magneto - hydrodynamical ( mhd ) simulations @xcite . the so - called @xmath9-prescription " is widely adopted in most of the works on accretion discs @xcite , in which the magnetic stress tensor is assumed to be proportional to the total pressure ( @xmath10 ) , gas pressure @xmath11 , or @xmath12 @xcite . the magnetic stress tensor @xmath3 was suggested in @xcite , however , the disc is thermal unstable if the radiation pressure dominates over the gas pressure @xcite(but also see * ? ? ? it was argued that the magnetic field amplification is likely to be limited to that the magnetic pressure is less than the gas pressure even in the inner radiation - pressure - dominated regions , i.e. , @xmath13 @xcite . the disc with this stress tensor is thermal stable @xcite . the stress tensor @xmath4 was initially suggested by @xcite based on the idea that the viscosity is proportional to the gas pressure , but the size of turbulence should be limited by the disc scaleheight that is given in terms of the total pressure . this is also supported by the analysis on the local dynamical instabilities in magnetized , radiation - pressure - supported accretion discs @xcite , which also leads to @xmath14 , where @xmath15 is the magnetic pressure , and @xmath16 is a constant of the order of unity ( see * ? ? ? * for a detailed discussion ) . @xcite argued that the accretion disc model with @xmath2 is too stable to explain the unstable behaviour observed in grs 1915 + 105 , and they further suggested that it is possible to couple the radiation to the particles through collisions and thereby allow the radiation pressure to contribute to the stress tensor to some extent . as the complexity of the physics in the radiation - pressure - dominated fluids , any one of these stress tensors should only be regarded as a possible option in accretion discs . the hard x - ray observations on agn may provide useful clues on the accretion disc - corona models . @xcite compiled a sample of radio - quite agn , and found a strong correlation between @xmath17 and @xmath0 , which was conformed by @xcite with a different agn sample . this correlation was then used to constrain the disc - corona models with different magnetic stress tensors , and they found that the model with magnetic stress tenor @xmath18 is favored by the correlation of @xmath19 . it was also found that the spectral index of hard x - ray continuum emission is correlated with the eddington ratio @xcite . @xcite found a strong correlation between the compton reflection and the hard x - ray spectral index for seyfert galaxies and x - ray binaries . in the accretion disc - corona scenario , the hard x - ray emission originates from the compton scattering of the soft photons by the hot electrons in the corona . therefore , the correlations of the hard x - ray spectral index with the eddington ratio / compton reflection , together with the correlation of @xmath19 , provide important constraints on the accretion disc - corona model . the parallel plane homogeneous corona is unable to produce an x - ray spectrum with @xmath20 in 210 kev . the x - ray photons radiated from the corona are reprocessed in the cold disc , and the reprocessed photons irradiate the corona , which cool the corona and lead to rather soft x - ray spectra with @xmath21 . in the patchy corona model proposed by @xcite , the corona appears as individual blobs above the cold disc . most of the reprocessed photons do not enter the blob , so the cooling is significantly reduced and the blobs are hotter than the parallel plane homogeneous corona , which leads to harder x - ray spectra . this model can explain the observed hard x - ray spectra with @xmath20 in some agn ( e.g. , * ? ? ? * ) , however , it is still unable to explain the correlation between the compton reflection and the hard x - ray spectral index for seyfert galaxies and x - ray binaries @xcite , which is also the case for the parallel plane homogeneous corona model . @xcite suggested that a central hot plasma surrounded by a cold disc may explain the correlation of the compton reflection with the hard x - ray spectral index . in this scenario , the cold disk is truncated at a certain radius @xmath22 , within which the hot plasma may possibly correspond to an advection dominated accretion flow ( adaf ) ( e.g. , * ? ? ? the radiation of the hot plasma is dominated by the inverse compton scattering of the soft photons from the outer cold disc , while the cold disc is irradiated by the inner hot plasma . thus , the compton reflection component decreases with increasing the inner radius of the cold disc , which leads to less soft seed photons from the cold disc entering the hot plasma and then the harder x - ray spectrum from the hot plasma . the correlation between the compton reflection and the hard x - ray spectral index can be reproduced by this model ( see * ? ? ? * for the details ) . an alternative model was proposed by @xcite to explain the correlation between the compton reflection and the x - ray spectral index . in this model , the hot plasma above the cold disc is assumed to move away from the cold disc at a mild relativistic velocity . such an outflow reduces the downward flux , and then reduces both the reflection and reprocessing in the cold disc . the reduction of the reprocessing leads to less incident soft seed photons and cooling in the hot plasma above the cold disc , and in turn leads to harder x - ray spectrum @xcite . in this work , we take these different magnetic stress tensors as candidates in our disc - corona model calculations , which could be tested with the x - ray observations of agn . we summarize the disc - corona model in sect . 2 , and the numerical results of the model calculations are given in sect . 3 . in sect . 4 , we discuss the physical implications of the results . the gravitational power dissipated in unit surface area of the accretion disc is given by @xmath23 , \label{q_dissi}\ ] ] where @xmath24 is the mass accretion rate of the disc , @xmath25 is the keplerian velocity at radius @xmath26 , and @xmath27 @xcite . the schwarzschild radius @xmath28 , where @xmath29 is the black hole mass . the corona is assumed to be heated by the reconnection of the magnetic fields generated by buoyancy instability in the disc . the power dissipated in the corona is @xcite @xmath30 where @xmath15 is the magnetic pressure in the disc , and @xmath31 is the velocity of the magnetic flux transported vertically in the disc . the rising speed @xmath32 is assumed to be proportional to their internal alfven velocity , i.e. , @xmath33 , in which @xmath34 is of the order of unity for extremely evacuated magnetic tubes . the soft photons from the disc are compton scattered by the hot electrons in the corona to x - ray bands , and about half of the scattered photons are intercepted by the disc . the reflection albedo @xmath35 is relatively low , @xmath36 , and most of the incident photons from the corona are re - radiated as blackbody radiation ( e.g. , * ? ? ? thus , the energy equation for the cold disc is @xmath37 where @xmath38 is the effective temperature in the mid - plane of the disc , and @xmath39 is the optical depth in vertical direction of the disc . in this work , we adopt @xmath40 in all our calculations . as the detailed physics for generating magnetic fields in the accretion disc is still quite unclear , we adopt different magnetic stress tensors as : @xmath41 in our model calculations , respectively . we summarize the equations describing the disc as follows : the continuity equation of the disc is @xmath42 where @xmath43 is the half thickness of the disc , @xmath44 is the mean density of the disc , and @xmath45 is the radial velocity of the accretion flow at radius @xmath26 . the equation of state for the gas in the disc is @xmath46 where @xmath47 , @xmath48 and @xmath49 are adopted corresponding to the plasma consisting of 3/4 hydrogen and 1/4 helium . the vertical hydrodynamical equilibrium requires @xmath50 , where the sound speed @xmath51 . the angular momentum equation for the disc is @xmath52=4\pi h_{\rm d}\tau_{r\varphi } , \label{angular}\ ] ] where the magnetic stress tensor is given by equation ( [ viscosity ] ) . solving equations ( [ q_dissi])-([angular ] ) numerically , the structure of the disc and the power dissipated in the corona @xmath53 can be derived as functions of radius @xmath26 . the ratio of the power dissipated in the corona to the total for such a disc - corona system is available : @xmath54 the equation of state for the hot gas in the corona is @xmath55 where @xmath56 and @xmath57 are the temperatures of the ions and electrons in the two - temperature corona , and the magnetic pressure @xmath58 . in this work , the magnetic fields are assumed to be equipartition with the gas pressure in the corona . the vertical hydrodynamical equilibrium in the corona requires @xmath59 , where the sound speed @xmath60 . the energy equation describing the two - temperature corona is @xmath61 where @xmath62 is the cooling rate in unit surface area of the corona and @xmath63 is the energy transfer rate from the ions to the electrons in the corona via coulomb collisions , which is given by @xcite . the fraction of the energy directly heat the electrons @xmath64 can be as high as @xmath65 by magnetic reconnection , if the magnetic fields in the plasma are strong @xcite . in this work , we adopt @xmath66 in our calculations . for the plasma consisting of different elements , the coulomb interaction between the electrons and ions is given by ( see * ? ? ? * ) @xmath67 @xmath68,\label{q_ie}\ ] ] where @xmath69 is the mass number of the @xmath70th element , @xmath71 , @xmath72 , @xmath73 , and @xmath74 are the dimensionless temperatures . although the cooling rate of the corona is dominated by the compton scattering of the soft incident photons from the disc , we include the synchrotron , bremsstrahlung and compton emission in our calculations . the cooling terms @xmath75 and @xmath76 are the functions of electron number density , temperature , and the magnetic field strength of the gas in the corona , which are taken from @xcite . assuming the corona to be a parallel plane , the mean probability of the soft photons injected from the cold disc experiencing the first - order scattering in the corona is @xmath77 where the constant specific intensity of the soft photons from the disc is assumed , @xmath78 is the angle of the motion of the soft photons with respect to the vertical direction of the disc and @xmath79 is the optical depth of the corona for electron scattering in the vertical direction . for simplicity , we assume the first - order scattered photons are radiated uniformly throughout the vertical direction of the corona . the mean probability for these first - order scattered photons experiencing the second - order scattering can be estimated as @xmath80{\rm d}\cos\theta , \label{prob_scat2}\ ] ] where @xmath81 , and the first - order scattered photons are assumed to be radiated isotropically . the probability of the scattered photons experiencing next higher order scattering approximates to @xmath82 , so we simply adopt @xmath83 for @xmath84 . thus , using the method suggested by @xcite , we can calculate the comptonized spectrum with equations ( [ prob_scat1 ] ) and ( [ prob_scat2 ] ) , if the density , temperature of electrons , and the incident spectrum of the disc are known . integrating the derived comptonized spectrum over the frequency , the cooling of the electrons in the corona due to the compton scattering @xmath85 is available . the disc structure and the power dissipated in the corona can be calculated as functions of @xmath26 , provided the black hole mass @xmath86 , mass accretion rate @xmath24 and the viscosity parameter @xmath9 are specified . the scaleheight of the corona is a function of @xmath56 , @xmath57 , and @xmath87 , on the assumption of static hydrodynamical equilibrium in the vertical direction . there are two equations describing the energy equilibrium in the corona at @xmath26 ( equation [ cor_energy ] ) , which contain three physical quantities of the corona : @xmath56 , @xmath57 , and @xmath87 . @xcite calculated the vertical structure of the two - temperature corona above a thin disc based on the evaporation mechanism , and their results show that the temperature of the ions in the corona is always in the range of @xmath88 virial temperature ( the virial temperature is defined as @xmath89 in their work ) . in this work , we adopt a slightly different definition : @xmath90 , as that adopted by @xcite . to avoid the complexity of calculating the vertical structure of the corona , we simply adopt @xmath91 [ equivalent to @xmath92 in @xcite s work ] in our model calculations . thus , the structure of the corona , i.e. , the temperature and density of the electrons , can be derived , and the spectra of the disc - corona system are available based on the derived disc - corona structure with different magnetic stress tensors given in equation ( [ viscosity ] ) . we calculate the disc - corona structure as described in section 2 . the black hole mass @xmath93 is adopted in all our calculations , because the main features of the hard x - ray spectra are almost independent of @xmath29 for massive black holes in agn . in fig . [ fig1 ] , we plot the ratios of the power radiated in the corona to the total power @xmath94 as functions of accretion rate @xmath95 predicted by the models with different magnetic stress tensors . the dimensionless accretion rate @xmath95 is defined as @xmath96 , where @xmath97 , and a conventional radiative efficiency @xmath98 is adopted . the model with @xmath99 always leads to constant ratios @xmath94 independent of accretion rate @xmath95 , while the disc - corona model calculations with @xmath13 show that @xmath100 becomes extremely small for high accretion rates ( e.g. , @xmath101 for @xmath102 ) . the model with @xmath4 shows that the ratios @xmath103 at @xmath104 , while @xmath105 at @xmath106 , for different values of @xmath9 . the temperature and optical depth for compton scattering of the hot electrons in the vertical direction of the corona are plotted in fig . the electron temperatures are in the range of @xmath107 k for different values of @xmath95 . the temperature of the hot electrons in the corona tends to decrease with accretion rate @xmath95 . when the accretion rate @xmath95 is as high as @xmath108 , the electron temperature of the corona decreases to @xmath109 k. in fig . [ fig3 ] , we plot the spectra of the disc - corona systems calculated with different magnetic stress tensors . the hard x - ray emission indeed exhibits a power - law feature for all the models . the photon spectral indices and the ratio of the bolometric luminosity to the x - ray luminosity in 210 kev as functions of accretion rate for different magnetic stress models are plotted in fig . the photon indices @xmath110 do not change much with accretion rate @xmath95 for the disc - corona models with either @xmath99 or @xmath13 , while the hard x - ray spectral index @xmath110 increases significantly with accretion rate @xmath95 for the model with @xmath111 . we compare the spectra of the accretion disc / corona systems with different ion temperature @xmath56 adopted in fig . it is found that the x - ray spectra change little with the value of @xmath56 , provided all other parameters are fixed . in fig . [ fig6 ] , we plot the compton @xmath112-parameter ( @xmath113 ) varying with @xmath56 , and find that it is insensitive to @xmath56 . it is believed that the power generated in the disc is transported vertically with the buoyancy of the magnetic fields in the disc , and the fraction of the power dissipated in the corona to the total is mainly regulated by the magnetic fields ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? @xcite found that the fraction @xmath94 is about 0.5 for the sources with low eddington ratio @xmath6 , while it decreases to 0.1 for @xmath114 , for a sample of agn . from fig . [ fig1 ] , we find that the model with @xmath3 always predict a constant @xmath94 , which seems to be inconsistent with the observation . the models with other two magnetic stress tensors can roughly reproduce the trend of @xmath115 decreasing with accretion rate @xmath95 , though the model with @xmath2 always underpredicts the @xmath94 at high eddington ratio end , even if @xmath116 is adopted . it seems that the model with @xmath4 is better than the other two models . the ratio @xmath94 decreases with accretion rate @xmath95 , which means the fraction of the power radiated from the cold disc increases with @xmath95 . the cooling mechanism of the hot corona is dominated by the inverse compton scattering of the soft photons from the cold disc by the hot electrons in the corona . there are more soft photons supplied by the cold disc for high-@xmath95 cases , and the fraction of the power radiated from the corona decreases with @xmath95 . thus , it can be easily understood that the temperature of the hot electrons in the corona should decrease with accretion rate @xmath95 ( see fig . [ fig2 ] ) . it is found that the temperature of the electrons in the corona decreases to @xmath117 k for the models with @xmath13 or @xmath118 when @xmath119 , which implies that the hot corona is nearly to be suppressed when @xmath95 is as high as @xmath120 . we calculate the spectra of the accretion disc - corona systems , and find that their hard x - ray continuum emission indeed exhibits a power - law spectrum ( see fig . [ fig3 ] ) . in fig . 4 , we plot the photon spectral indices @xmath110 in 210 kev and the x - ray correction factors @xmath1 as functions of accretion rate @xmath95 with different magnetic stress tensors . we find that the photon spectral index @xmath110 predicted by the model with @xmath99 is within @xmath121 for different values of @xmath95 , and it increases slightly with @xmath95 . the x - ray correction factor @xmath122 remains constant with accretion rate @xmath95 for this model . the photon spectral index @xmath110 increases with @xmath95 while @xmath123 , and it then decreases with @xmath95 , for the model with @xmath13 . we find that the bremsstrahlung emission contributes some to the hard x - ray energy band when @xmath95 is high , and the photon spectral index @xmath110 is slightly affected by the bremsstrahlung emission ( see the middle panel of fig . we find that the x - ray correction factor @xmath1 for this model becomes extremely high for high-@xmath95 cases , which seems to be inconsistent with that derived from the observations ( see fig . 12 in * ? ? ? the x - ray spectra of the disc - corona systems with @xmath4 show that both the photon spectral index @xmath110 and the x - ray correction factor @xmath1 increase with accretion rate @xmath95 , which are qualitatively consistent with the observations , except for the low @xmath95 end . the temperature of the electrons in the corona decreases with @xmath95 ( see fig . [ fig2 ] ) , and the inverse compton scattered x - ray spectrum therefore becomes softer for a higher @xmath95 . similarly , anticorrelations between the x - ray spectral indices and x - ray fluxes were found in some x - ray binaries ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) , which seem to be roughly consistent with our present disc corona model . the photon spectral index @xmath110 is found to increase with the eddington ratio , and it could be as low as @xmath124 when @xmath125 ( see * ? ? ? * ; * ? the hard x - ray photon spectral index @xmath126 derived from our model calculations with any magnetic stress tensor when @xmath125 , which are significantly higher than the observed @xmath127 ( e.g. * ? ? ? * ; * ? ? ? although the patchy corona model proposed by @xcite can produce the observed hard x - ray spectra with @xmath127 , this patchy corona model is unable to explain the correlation between the compton reflection and the hard x - ray spectral index . in an alternative model , the hot plasma above the cold disc is assumed to move away from the cold disc at a mild relativistic velocity @xcite , which reduces both the reflection and reprocessing in the cold disc . this naturally leads to harder x - ray spectrum ( see the discussion in sect . 1 ) . however , the detailed physics for producing such mildly relativistic outflows is still unclear . it was argued that the central engines in these low - luminosity sources with very hard x - ray emission may be different from their high - luminosity counterparts , i.e. , the advection dominated accretion flows ( adafs ) may be present in these low - luminosity sources @xcite . the soft incident photons for comptonization are mainly due to the synchrotron@xmath7bremsstrahlung emission in the adafs . the energy density of the soft photons in the adaf is much lower than that in the disc - corona model , which leads to inefficient cooling and relatively higher electron temperature in the adafs ( e.g. , * ? ? ? thus , the x - ray spectra of the adafs can be much harder than those of the disc - corona systems . the very hard x - ray spectra observed in these low luminosity sources can be well modelled with the adafs accreting at low rates ( see e.g. , * ? ? ? * ; * ? ? ? * for the spectral modeling for the hard x - ray emission from the low luminosity agn with adafs ) . our present model calculations are limited to the disc - corona model , in which the cold discs extend to the marginal stable orbits . the model of a disc - corona connecting with inner adaf may resolve the photon index discrepancy at the low @xmath95 end . the geometry of this adaf@xmath7disc / corona scenario is quite similar to the hot plasma@xmath7cold disc model proposed by @xcite . when the accretion rate decreases to a critical value @xmath128 , the inner cold disc may be truncated at radius @xmath129 and it transits to an adaf within this radius . the transition radius @xmath129 may be close to the marginal stable orbits soon after the accretion mode transition ( e.g. , * ? ? * ; * ? ? ? * ; * ? ? ? * ) , and then it may increase with decreasing accretion rate @xmath95 ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? the x - ray spectrum of such an adaf@xmath7disc / corona system consists of emission from the inner adaf and outer corona . the ratio of the x - ray emission from the adaf to that from the corona increases with decreasing @xmath95 , and therefore the photon spectral index @xmath110 may decrease smoothly with decreasing @xmath95 provided the initial truncated radius of the cold disc is not very large compared with the marginal stable orbits . similar to @xcite s model , the correlation between the compton reflection and the hard x - ray spectral index can also be naturally explained by this adaf@xmath7disc / corona model , if the truncated radius @xmath129 increases with decreasing accretion rate @xmath95 . this model can also successfully explain the spectral behaviours of x - ray binaries ( see * ? ? ? * for a review and references therein ) . the detailed calculations on such adaf@xmath7corona systems will be reported in our future work . in this work , we simply assume @xmath130 , motivated by the previous work on the disc - corona model calculations @xcite . we also check how @xmath56 may affect our results by tuning the value of @xmath56 , and find that the x - ray spectra of the disc / corona systems change very little if all other disc parameters are fixed ( see fig . [ fig5 ] ) . the cooling of the corona is dominated by the inverse compton radiation , which is roughly proportional to compton @xmath112-parameter . thus , it is not surprising that our calculations show the compton @xmath112-parameter varying little with @xmath56 if all other parameters are fixed ( see equation [ cor_energy ] and fig . [ fig6 ] ) . the thickness of the corona is mainly regulated by the ion temperature @xmath131 , and therefore the electron density decreases with increasing @xmath56 , which leads to significant difference in synchrotron / bremsstrahlung spectra for different @xmath56 ( see fig . [ fig5 ] ) . we find that the resulted @xmath132 relations are almost not changed if a different value of @xmath56 is adopted . unlike the adafs , almost all the power dissipated in the hot corona with magnetic reconnection is radiated away locally . this means the radiated power in the corona is independent of the value @xmath64 , and the temperature and density of the electrons in the corona are almost insensitive with the value of @xmath64 . we also perform the same model calculations for different black hole masses ( e.g. , @xmath133 ) . it is found that our results are almost independent of @xmath29 for massive black holes . in all our calculations , we assume the magnetic fields to be equipartitioned with the gas pressure in the corona . as the cooling of the corona is dominated by the inverse comptonization of the soft photons from the cold disc , the structure of the disc - corona and its spectrum ( except in the radio wavebands ) is almost independent of the magnetic field strength in the corona . recently , @xcite found there is a strong correlation between the radio luminosity ( @xmath134 ) and x - ray luminosity ( @xmath135 ) with @xmath136 , for the radio quiet palomar - green ( pg ) quasar sample . the spectra of our disc - corona model show that the radio emission is correlated with the x - ray emission , which is roughly consistent with the correlation between @xmath134 and @xmath135 discovered by @xcite . i thank the referee , zdziarski a. a. , for his helpful suggestions / comments , b. f. liu , t. g. wang , q. w. wu and w. yuan for helpful discussion . this work is supported by the nsfc ( grants 10773020 , 10821302 and 10833002 ) , the cas ( grant kjcx2-yw - t03 ) , and the national basic research program of china ( grant 2009cb824800 ) . balbus s. a. , hawley j. f. , 1991 , apj , 376 , 214 balbus s. a. , hawley j. f. , 1998 , rev . phys . , 70 , 1 beloborodov a. m. , 1999 , apj , 510 , l123 bisnovatyi - kogan g. s. , lovelace r. v. e. , 1997 , apj , 486 , l43 bisnovatyi - kogan g. s. , lovelace r. v. e. , 2000 , apj , 529 , 978 blaes o. , socrates a. , 2001 , apj , 553 , 987 cao x. , xu y .- d . , 2007 , mnras , 377 , 425 coppi p. s. , blandford r. d. , 1990 , mnras , 245 , 453 del santo m. , malzac j. , jourdain e. , belloni t. , ubertini p. , 2008 , mnras , 390 , 227 di matteo t. , 1998 , mnras , 299 , l15 di matteo t. , celotti a. , fabian , a. c. , 1999 , mnras , 304 , 809 done c. , gierlinski m. , kubota a. , 2007 , a&arv , 15 , 1 galeev a. a. , rosner r. , vaiana g. s. , 1979 , apj , 229 , 318 haardt f. , maraschi l. , 1991 , apj , 380 , l51 haardt f. , maraschi l. , 1993 , apj , 413 , 507 haardt f. , maraschi l. , ghisellini g. , 1994 , apj , 432 , l95 hirose s. , krolik j.h . , blaes o. , 2008 , apj accepted ( arxiv:0809.1708 ) kawaguchi t. , shimura t. , mineshige s. , 2001 , apj , 546 , 966 laor a. , behar e. , 2008 , mnras accepted ( arxiv:0808.0637 ) liu b. f. , mineshige s. , ohsuga k. , 2003 , apj , 587 , 571 liu b. f. , mineshige s. , shibata k. , 2002 , apj , 572 , l173 liu b. f. , yuan w. , meyer f. , meyer - hofmeister e. , xie g. , z. , 1999 , apj , 527 , l17 lu y. , yu , q. , 1999 , apj , 526 , l5 malkan m. a. , sargent w. l. w. , 1982 , apj , 254 , 22 malzac j. , beloborodov a. m. , poutanen j. , 2001 , mnras , 326 , 417 merloni a. , fabian a. c. , 2001 , mnras , 328 , 958 merloni a. , fabian a. c. , 2002 , mnras , 332 , 165 narayan r. , yi i. , 1995 , apj , 452 , 710 nayakshin s. , rappaport s. , melia f. , 2000 , apj , 535 , 798 quataert e. , di matteo t. , narayan , r. , ho , l.c . , 1999 , apj , 525 , l89 rozanska a. , czerny b. , 2000 , a&a , 360 , 1170 sakimoto p.j . , coroniti f.v . , 1981 , apj , 247 , 19 shakura n.i . , sunyaev r.a . , 1973 , a&a , 24 , 337 shakura n.i . , sunyaev r.a . , 1976 , mnras , 175 , 613 shemmer o. , brandt w.n . , netzer h. , maiolino r. , kaspi s. , 2006 , apj , 646 , l29 shemmer o. , brandt w.n . , netzer h. , maiolino r. , kaspi s. , 2008 , apj , 682 , 81 shields g. a. , 1978 , nature , 272 , 706 spruit h. c. , deufel b. , 2002 , a&a , 387 , 918 stella l. , rosner r. , 1984 , apj , 277 , 312 stepney s. , guilbert , p.w . , 1983 , mnras , 204 , 1269 sun w. -h . , malkan m. a. , 1989 , apj , 346 , 68 svensson r. , zdziarski a. a. , 1994 , apj , 436 , 599 taam r.e . , lin d.n.c . , 1984 , apj , 287 , 761 vasudevan r.v . , fabian a.c . , 2007 , mnras , 381 , 1235 wang j .- m . , watarai k .- y . , mineshige s. , 2004 , apj , 607 , l107 xu y .- d . , cao x. , 2008 , chjaa in press ( arxiv:0809.1793 ) yu w. , klein - wolt m. , fender r. , van der klis m. , 2003 , apj , 589 , l33 yuan f. , narayan r. , 2004 , apj , 612 , 724 zdziarski a.a . , 1998 , mnras , 296 , l51 zdziarski a.a . , gierlinski m. , gondek d. , magdziarz p. , 1996 , a&as , 120 , 553 zdziarski , a.a . , gierlinski m. , mikolajewska j. , wardzinski g. , smith d.m . , harmon b.a , kitamoto s. , 2004 , mnras , 351 , 791 zdziarski a.a . , lubinski p. , smith d.a . , 1999 , mnras , 303 , l11	the hard x - ray emission of active galactic nuclei ( agn ) is believed to originate from the hot coronae above the cold accretion discs . the hard x - ray spectral index is found to be correlated with the eddington ratio @xmath0 , and the hard x - ray bolometric correction factor @xmath1 increases with the eddington ratio . the compton reflection is also found to be correlated with the hard x - ray spectral index for seyfert galaxies and x - ray binaries . these observational features provide very useful constraints on the accretion disc - corona model for agn . we construct an accretion disc - corona model and calculate the spectra with different magnetic stress tensors in the cold discs , in which the corona is assumed to be heated by the reconnection of the magnetic fields generated by buoyancy instability in the cold accretion disc . our calculations show that the magnetic stress tensor @xmath2 fails to explain all these observational features , while the disc - corona model with @xmath3 always leads to constant @xmath1 independent of the eddington ratio . the resulted spectra of the disc - corona systems with @xmath4 show that both the hard x - ray spectral index and the hard x - ray bolometric correction factor @xmath1 increase with the eddington ratio , which are qualitatively consistent with the observations . we find that the disc - corona model is unable to reproduce the observed very hard x - ray continuum emission from the sources accreting at low rates ( e.g. , @xmath5 for @xmath6 ) , which may imply the different accretion mode in these low luminosity sources . we suggest that the disc - corona system transits to an advection - dominated accretion flow@xmath7disc corona system at low accretion rates , which may be able to explain all the above - mentioned correlations . [ firstpage ] ( galaxies : ) quasars : general accretion , accretion discs black hole physics
nucleus - nucleus ( a+a ) collisions at relativistic energies have been intensely studied over the last two decades . the main goal of these efforts is to understand the properties of strongly interacting matter under extreme conditions of high energy and baryon densities when the creation of the quark - gluon plasma ( qgp ) is expected @xcite . experimental results obtained in a broad range of collision energies indicate that an extended zone of strongly interacting dense matter indeed occurs in the collision process . various collision characteristics and their collision energy dependence suggest @xcite that a transient state of deconfined matter may be created at collision energies as low as 40 @xmath0gev . qgp formation is expected to occur at the early collision stage when the system is sufficiently hot and dense . in the course of further evolution , the system dilutes and cools down , hadronizes and finally decays into free hadrons . therefore the final state hadrons carry only indirect information about the early stage of the collision . thus firm conclusions about the creation of deconfined matter require studying a variety of collision characteristics . as fluctuations are sensitive to the dynamics of the system , in particular at the phase transition , the analysis of event - by - event fluctuations has been proposed as an important tool in the study of a+a collisions @xcite . large acceptance detectors , which allow the observation of a significant fraction of the final state particles , have made this suggestion truly attractive @xcite . first pioneering studies were carried out by the na49 collaboration on the fluctuations of the average transverse momentum @xcite and the @xmath25 ratio @xcite in central pb+pb collisions at the cern sps . there are a number of collision characteristics which can be inferred from experimental data on event - by - event fluctuations . first of all , the fluctuation analysis can help to resolve the long standing problem whether , or rather to what extent , the strongly interacting matter , emerging from the early collision stage , achieves both thermal @xcite and chemical @xcite equilibrium . in the na49 investigation of event - wise fluctuations in the @xmath25 ratio @xcite in central pb+pb collisions at 158 @xmath0gev , no significant deviations from the hadro - chemical equilibrium ratio were found . if the equilibrium is indeed reached , the system s heat capacity @xcite and its compressibility @xcite can , in principle , be deduced from the temperature and multiplicity fluctuations respectively . while large multiplicity fluctuations can be treated as a signal of particle production via clusters @xcite or super - cooled droplets of deconfined matter @xcite , small final state fluctuations of the conserved charges , electric or baryonic , can occur when fluctuations generated in the plasma phase are frozen due to the rapid expansion of the system @xcite . on the other hand , significant transverse momentum and multiplicity fluctuations can result if the system hadronizes from a qgp near the predicted second order critical qcd end point @xcite . the latter has been shown by recent lattice qcd studies to occur at a substantial baryochemical potential @xcite , characteristic of the cern sps energy range . this paper extends the previous study @xcite of the na49 experiment of event - by - event transverse momentum fluctuations . the main objective is to observe how the fluctuation pattern changes with increasing number of nucleons participating in a collision , i.e. with the system size . for this purpose , not only pb+pb collisions at various centralities are studied , but also p+p , c+c and si+si interactions at a beam energy of 158 gev per nucleon . in particular , it will be checked , whether the correlations present in the final state of p+p interactions survive in the collisions of heavier systems , as expected if the nucleus+nucleus ( a+a ) collision is a simple superposition of nucleon+nucleon ( n+n ) interactions . moreover a test will be performed of the reasonable expectation that the fluctuations become more similar to those of an equilibrated system when the number of participating nucleons increases . measuring event - by - event fluctuations in a+a collisions one should consider the influence of two trivial sources of fluctuations . the first one is caused by event - by - event fluctuations of the collision geometry and the second one by the finite number of particles ( statistical fluctuations ) . the dynamical fluctuations of interest have to be extracted from the noise caused by these trivial sources . in such a situation , a suitable choice of statistical tools for the study of event - by - event fluctuations is really important . in this work , mainly the fluctuation measure @xmath26 introduced in @xcite will be employed . however , other fluctuation measures have also been proposed and studied , e.g. @xmath27 @xcite , @xmath28 @xcite and @xmath29 @xcite , which can all be related . @xmath26 equals exactly zero when inter - particle correlations are absent . it also eliminates ` geometrical ' fluctuations due to the impact parameter variation . thus , @xmath26 is ` deaf ' to the statistical noise and ` blind ' to the collision centrality . the @xmath26 measure was already used in the previous experimental study @xcite of @xmath30 fluctuations . it was also calculated within various models of nuclear collisions @xcite . however , in these model considerations , the effects of experimental acceptance , which significantly influence the results , were usually not taken into account . consequently , most of these results can not be compared to the data . for better understanding of the structure of the correlations contributing to @xmath26 , two - particle transverse momentum correlations ( as proposed in @xcite ) are studied as well . a preliminary analysis was presented in @xcite . this paper is organized as follows : in sec . [ s : measures ] the statistical tools used in this analysis are introduced and briefly discussed . the experimental set - up and procedures are presented in sec . [ s : experiment ] . experimental effects such as detector acceptance and two - track resolution are discussed in sec . [ s : data ] . the results on the system dependence of the @xmath30 fluctuations are presented in sec . [ s : results ] . discussion of the results and their comparison with theoretical models is given in sec . [ s : discussion ] . a summary closes the paper . there are numerous observables which can be used to measure @xmath30 fluctuations in high energy collisions . a natural one is the distribution of the average transverse momentum of the events defined as @xmath31 where @xmath32 is the multiplicity of accepted particles in a given event and @xmath33 is the transverse momentum of the @xmath34-th particle . the distribution of @xmath35 is usually compared to the corresponding distribution obtained for ` mixed events ' in which the particles are independent from each other and follow the experimental inclusive spectra ( the multiplicity distribution for mixed events is the same as for the data ) . a difference between the two distributions signals the presence of dynamical fluctuations . since the @xmath35 distribution crucially depends on the particle multiplicity , the method can not be used to compare systems of significantly different multiplicities . a more appropriate measure is the quantity @xmath26 @xcite which , by its construction , is insensitive to the system size . in this paper the @xmath26 measure is used for the analysis of @xmath36 fluctuations ( @xmath1 ) . following the authors of @xcite , one defines the single - particle variable @xmath37 with the bar denoting averaging over the single - particle inclusive distribution . one easily observes that @xmath38 . further , one introduces the event variable @xmath39 , which is a multi - particle analog of @xmath40 , defined as @xmath41 where the summation runs over particles in a given event . note , that @xmath42 , where @xmath43 represents averaging over events . finally , the @xmath1 measure is defined as @xmath44 the second part of eq . ( [ eq_phi ] ) is simply the dispersion of the inclusive @xmath30 distribution ( further denoted as @xmath45 ) . it can be easily shown that @xmath46 , when no inter - particle correlations are present and the single - particle spectrum is independent of multiplicity . as already mentioned , @xmath1 is insensitive to centrality . this property may be expressed as follows : @xmath1 is independent of the distribution of the number of particle sources if the sources are identical and independent from each other @xcite . in particular , @xmath1 does not depend on the impact parameter if the a+a collision is a simple superposition of n+n interactions . in spite of the above mentioned advantages , there is an important disadvantage of using @xmath1 in the fluctuation analysis . while @xmath1 is sensitive to the presence of dynamical correlations in a system , it does not disentangle their nature . fluctuations of very different character contribute to @xmath1 . in order to achieve a better understanding of the fluctuation structure one needs to apply a more differential method @xcite . the correlations can be studied by plotting the cumulative @xmath30 variables of particle pairs . namely , for a given particle , instead of its @xmath36 one introduces the variable @xmath47 defined as @xcite @xmath48 where @xmath49 is the inclusive @xmath36 distribution , normalized to unity , which is obtained from all particles used in the analysis . by construction , the @xmath47 variable varies between 0 and 1 with a _ flat _ probability distribution . the two - particle correlation plots , as presented in this paper , are obtained by plotting @xmath50 points for all possible particle pairs within the same event . the number of pairs in each @xmath50 bin is divided by the mean number of pairs in a bin ( averaged over all @xmath50 bins ) . this two - dimensional plot is uniform when no inter - particle correlations are present in the system . correlations due to the bose statistics produce a ridge along the diagonal of the @xmath50 plot , which starts at @xmath51 and ends at @xmath52 , whereas temperature fluctuations lead to a saddle shaped structure @xcite . as will be seen in the figures , the distribution of @xmath53 or @xmath54 obtained from the two - dimensional @xmath50 plots by projecting on the @xmath53 or @xmath54 axis is not flat . this is due to the method by which the plots are constructed . namely , each pair of particles is represented by a point on the plot . therefore , the events with higher multiplicities are represented by a larger number of pairs than those with smaller multiplicities . since the shape of the @xmath30 distribution depends on the event multiplicity , the projection of the two - dimensional plot on @xmath53 or @xmath54 is no longer flat . however , it should be stressed that in the absence of any correlations the @xmath50 plot is uniformly populated and the @xmath53 and @xmath54 projections are flat . the na49 experiment is a large acceptance hadron spectrometer at the cern - sps used to study the hadronic final states produced by collisions of various beam particles ( p , pb from the sps and c , si from the fragmentation of the primary pb beam ) with a variety of fixed targets . the main tracking devices are four large volume time projection chambers ( tpcs ) ( fig . [ setup ] ) which are capable of detecting 80% of approximately 1500 charged particles created in a central pb+pb collision at 158 @xmath0gev . two of them , the vertex tpcs ( vtpc-1 and vtpc-2 ) , are located in the magnetic field of two super - conducting dipole magnets ( 1.5 and 1.1 t , respectively ) and two others ( mtpc - l and mtpc - r ) are positioned downstream of the magnets symmetrically to the beam line . the results presented here are analyzed with a global tracking scheme @xcite , which combines track segments that belong to the same physical particle but were detected in different tpcs . the na49 tpcs allow precise measurements of particle momenta @xmath55 with a resolution of @xmath56 ( gev / c)@xmath57 . up is supplemented by two time of flight ( tof ) detector arrays and a set of calorimeters . the targets , c ( 561 mg / cm@xmath14 ) , si ( 1170 mg / cm@xmath14 ) and pb ( 224 mg / cm@xmath14 ) for ion collisions and a liquid hydrogen cylinder ( length 20 cm ) for elementary interactions , are positioned about 80 cm upstream from vtpc-1 . pb beam particles are identified by means of their charge as seen by a helium gas - cherenkov counter ( s2 ) and p beam particles by a 2 mm scintillator ( s2 ) . both of these are situated in front of the target . the study of c+c and si+si reactions is possible through the generation of a secondary fragmentation beam which is produced by a primary target ( 1 cm carbon ) in the extracted pb - beam . with the proper setting of the beam line magnets a large fraction of all @xmath59 fragments at @xmath60gev are transported to the na49 experiment . on - line selection based on a pulse height measurement in a scintillator beam counter ( s2 ) is used to select particles with @xmath61 ( carbon ) and @xmath62 ( al , si , p ) . in addition , a measurement of the energy loss in beam position detectors ( bpd-1/2/3 in fig . [ setup ] ) allows for a further selection in the off - line analysis . these detectors consist of pairs of proportional chambers and are placed along the beam line . they also provide a precise measurement of the transverse positions of the incoming beam particles . for p , c and si beams , interactions in the target are selected by an anti - coincidence of the incoming beam particle with a small scintillation counter ( s4 ) placed at the beam axis between the two vertex magnets . for p+p interactions at 158 @xmath0gev this counter selects a ( trigger ) cross section of 28.5 mb out of 31.6 mb of the total inelastic cross section . for pb beams , an interaction trigger is provided by an anti - coincidence with a helium gas - cherenkov counter ( s3 ) directly behind the target . the s3 counter is used to select minimum bias collisions by requiring a reduction of the cherenkov signal by a factor of about 6 . since the cherenkov signal is proportional to @xmath63 , this requirement ensures that the pb projectile has interacted with a minimal constraint on the type of interaction . this setup limits the triggers on non - target interactions to rare beam - gas collisions , the fraction of which proved to be small after cuts , even in the case of peripheral pb+pb collisions . the centrality of the nuclear collisions is selected by use of information from a downstream calorimeter ( vcal ) , which measures the energy of the projectile spectator nucleons . the geometrical acceptance of the vcal calorimeter is adjusted in order to cover the projectile spectator region by the setting of the collimator ( coll ) . details of the na49 detector set - up and performance of tracking software are described in @xcite . the data used for the analysis consists of samples of p+p , c+c , si+si and pb+pb collisions at 158 @xmath0gev . for pb+pb interactions a minimum bias trigger was used allowing a study of the centrality dependence . the distribution of energy measured in the vcal for the minimum bias pb+pb events was divided into six centrality bins ( table [ data_sets ] ) which are numbered from 1 ( the most central ) to 6 ( the most peripheral ) . for each bin of centrality the range of the impact parameter @xmath64 , and the mean number of wounded nucleons @xmath65 were determined by use of the glauber model and the venus event generator @xcite . the fraction of the total inelastic cross section of nucleus+nucleus collisions ( @xmath66 ) corresponding to each data set was calculated directly by use of the distribution of energy measured in the vcal . in order to estimate the correlation between the energy deposited in the vcal and the impact parameter minimum bias venus events were processed through the geant detector simulation code , and the energy deposited in the vcal was simulated . the correlation between @xmath64 and @xmath67 was obtained from glauber model calculations using the spectator - participant model of a+a interactions . the values of @xmath66 , @xmath68 and @xmath64 presented in table [ data_sets ] are taken from @xcite for minimum bias pb+pb collisions and from @xcite for c+c and si+si interactions . the minimum bias pb+pb data consists of data taken during three different periods and at both magnetic field polarities . the most central pb+pb events correspond to 5@xmath69 of the total geometric cross section . since the minimum bias data provide only a small number of events in the most central pb+pb bin , additional central trigger runs were used . the aim of the event selection criteria is to reduce a possible contamination with non - target collisions . the primary vertex was reconstructed by fitting the intersection point of the measured particle trajectories . only events with a proper quality and position of the reconstructed vertex are accepted in this analysis . the vertex coordinate @xmath70 along the beam has to satisfy @xmath71 , where the nominal vertex position @xmath72 and cut parameter @xmath73 values are : -579.5 and 5.5 cm , -579.5 and 1.5 cm , -579.5 and 0.8 cm , -578.9 and 0.4 cm for p+p , c+c , si+si and pb+pb collisions , respectively . in order to reduce the contamination of particles from secondary interactions , weak decays and other sources of non - vertex tracks , several track cuts are applied . the accepted particles are required to have measured points in at least one of the vertex tpcs . a cut on the so - called track impact parameter , the distance between the reconstructed main vertex and the track in the target plane , is applied ( @xmath74 cm and @xmath75 cm ) to reduce the contribution of non - vertex particles . moreover , particles are accepted only when the potential number of points ( calculated on the basis of the geometry of the track ) in the detector exceeded 30 . the ratio of the number of points on a track to the potential number of points is required to be higher than 0.5 in order to avoid the counting of track segments instead of whole tracks . only forward rapidity tracks ( 4.0 @xmath76 5.5 , rapidity calculated assuming the pion mass for all particles ) with 0.005 @xmath77 1.5 gev / c are used in this analysis . the na49 detector provides a large ( but not complete ) acceptance in the forward hemisphere . two example plots of @xmath36 versus azimuthal angle @xmath78 ( see fig . [ setup ] for definition ) for positively charged particles ( for the standard polarity of the magnetic field ) are shown in fig [ angle ] . the solid lines represent a parametrization of the acceptance limits by the formula : @xmath79 where the values of parameters @xmath0 , @xmath80 and @xmath81 depend on the rapidity interval as given in table [ abc ] . these values apply to negatively charged particles as well , provided @xmath78 in eq . [ accept ] is replaced by @xmath82 . only particles within the analytical curves are used in this analysis . this well defined acceptance is essential for later comparison of the results with models and other experiments . the na49 detector is able to register particles produced in a significantly wider rapidity range covering almost the whole forward hemisphere . it would be interesting to determine the correlation measure @xmath1 in the mid - rapidity region . preliminary results for the rapidity range 2.9 - 4.0 were reported @xcite . however , the azimuthal acceptance in this region is more limited and systematic uncertainties affecting @xmath1 are not yet fully understood . the statistical error on @xmath1 was estimated as follows . the whole sample of events was divided into 30 subsamples . the value of @xmath1 was evaluated for each subsample and the dispersion ( @xmath83 ) of the results was then calculated . the statistical error of @xmath1 was taken to be equal to @xmath84 . the event and track selection criteria reduce the possible systematic bias of the measured @xmath1 values . in order to estimate the remaining systematic uncertainty , the values of cut parameters have been varied within a reasonable range and the systematic error has been estimated as a half of the difference between the highest and the lowest @xmath1 value . in addition , results obtained from the analysis of data taken at two different magnetic field polarities as well as from different running periods have been compared . event cuts are used to reject possible contamination of non - target interactions , however there is always a small fraction of remaining non - target events which can influence the @xmath1 values . the dependence of @xmath1 on the event selection cut @xmath73 is shown in fig . [ fipt_vz ] . the observed variation of @xmath1 with @xmath73 is small . the estimated systematic error is smaller than 0.85 mev / c for peripheral pb+pb collisions , 0.55 mev / c for si+si data and 0.5 mev / c for p+p events . the majority of tracks selected by the track selection criteria are main vertex tracks and the remaining fraction ( @xmath8510% ) originates predominantly from weak decays and secondary interactions with the material of the detector . in order to estimate the influence of this contamination on the measured value of @xmath1 , the impact parameter cut was varied ( fig . [ fipt_bxby ] ) . a small increase of @xmath1 with increasing impact parameter cut is observed and may be due to the increased contribution of non - vertex tracks from weak decays and secondary interactions . by use of the venus @xcite simulations these tracks were found to be correlated in @xmath36 thus increasing the measured value of @xmath1 . the estimated systematic error due to the contamination of non - vertex tracks is smaller than 1.6 mev / c for central pb+pb collisions , 0.75 mev / c for si+si data and 0.35 mev / c for p+p events . losses of tracks due to the reconstruction inefficiency and track selection cuts influence the measured @xmath1 values . in order to estimate this effect , the dependence of @xmath1 on the percentage of randomly rejected particles was calculated . these dependences for the most peripheral ( 6 ) , the most central ( 1 ) pb+pb collisions and for p+p interactions are shown in fig . [ random ] . within the considered kinematic region ( forward rapidity ) the tracking efficiency of our detector is higher than 95% . [ random ] implies that the bias due to tracking inefficiency is not higher than 0.5 - 1.0 mev / c . as an estimate of the systematic error on @xmath1 a maximal error resulting from the above study has been taken . the systematic error is about 1.6 mev / c for pb+pb collisions and 1.2 mev / c for p+p , c+c and si+si interactions . it has already been shown @xcite that the limited two track resolution influences the measured @xmath1 values . in order to estimate this contribution several samples of mixed events ( for different a+a collisions ) were produced . mixed events were constructed from original events , the multiplicities of mixed events being the same as in the case of real events but each particle in a mixed event taken at random from a different real event . the @xmath1 value calculated for the sample of mixed events was consistent with zero . in the second step the mixed events were processed by the na49 simulation software . the resulting simulated raw data were reconstructed and the @xmath1 measure calculated . the obtained @xmath1 values are negative as expected for the anti - correlation introduced by the losses due to the limited two track resolution . the additive two track resolution correction is calculated as the difference ( @xmath86 ) between the values of @xmath1 after and before this procedure . [ ttr ] presents this correction versus mean multiplicity . the lines correspond to an analytical parametrization of this dependence . the absolute values of the track resolution corrections are larger for heavier colliding systems where the density of tracks is relatively high . the absolute values of @xmath86 are also larger for positively charged particles than for negatively charged ones , which is mainly due to higher track density for positive particles caused by significantly larger number of protons than anti - protons . the @xmath86 values are negative indicating that @xmath1 measured with an ideal detector would be higher . for a given multiplicity , @xmath1 corrected for the limited two track resolution effect equals ` raw ' @xmath1 minus the corresponding @xmath86 . the results shown in this section refer to _ accepted _ particles , i.e. particles that are accepted by the detector and pass all kinematic cuts and track selection criteria . the data cover a broad range in @xmath36 ( @xmath87 gev / c ) . the rapidity of accepted particles is restricted to the interval 4.0 to 5.5 which corresponds to forward rapidities in the collision of equal mass nuclei ( at 158 @xmath0gev energy the center of mass rapidity equals 2.9 for fixed target geometry ) , where the azimuthal acceptance is large . the mean multiplicities of accepted particles , the dispersions @xmath88 of the multiplicity distributions , the mean inclusive transverse momenta , the dispersions @xmath89 of inclusive transverse momentum distributions and @xmath1 values for all data sets used in this analysis are given in table [ data ] . the @xmath1 values shown in this table have been calculated for all accepted charged particles as well as for the negatively and the positively charged particles separately . all values of @xmath1 were corrected for the two track resolution effect . [ meanpt ] shows the distributions of the mean ( per event ) transverse momentum @xmath90 for p+p , si+si and central pb+pb collisions . points correspond to data and the histograms to mixed events . the data are not corrected for any experimental effects . events with zero accepted particle multiplicity are not taken into account . the small difference between @xmath35 distributions for real and mixed events demonstrates that dynamical fluctuations are small . moreover , no distinct class of events with unusual fluctuations is observed . the width of the @xmath35 distribution strongly decreases with the colliding system size as expected from the increasing particle multiplicity . the fluctuation measure @xmath1 is more sensitive to small dynamical fluctuations . the measured values , corrected for two - track resolution , are plotted in fig . [ fipt ] versus mean number of wounded nucleons for all accepted charged particles and also for positively and negatively charged particles separately . the @xmath1 values are small ( when compared to @xmath89 ) for all investigated systems ( @xmath91 mev / c ) , but a significant centrality dependence is observed . the @xmath1 values increase with the system size up to the maximum value which is observed for the most peripheral pb+pb collisions , and then @xmath1 values decrease with increasing number of wounded nucleons . the @xmath1 value for si+si collisions seems to be lower than that for the most peripheral pb+pb ones although the number of wounded nucleons in both reactions is similar . this might suggest that @xmath92 does not fully determine the fluctuations in a+a collisions . the @xmath1 measure calculated for positively charged particles is always lower than that for the negatively charged ones . the @xmath1 value for all charged particles is always higher than that for either the negatively or the positively charged particles . two - particle correlation plots of the cumulant transverse momentum variable @xmath47 are presented in fig . [ 2d_plots_b ] for p+p , c+c , si+si and three centralities of pb+pb interactions ( note the different color scales ) . it is seen that the plots are not uniformly populated . in particular , significant long range correlations of about 40% ( the color scale varies from 0.75 to 1.6 ) are observed for the p+p data . this is rather unexpected when compared to the low @xmath1 value . these correlations are not seen when heavier colliding systems are studied . instead , short range correlations become visible as an enhancement of the point density in the region close to the diagonal . they are most prominent for central pb+pb collisions and are consistent with the effect of bose - einstein statistics . for the most peripheral collisions the pattern seen in the two - particle correlation plot is different from that in the remaining systems one observes an enhancement in the region close to @xmath93 and @xmath94 . in this section the results are compared with predictions of models and with the results of other experiments . fig . [ fipt ] shows that @xmath1 is a non - monotonic function of centrality with the maximum at approximately @xmath95 . such a behavior strongly resembles the dependence of the magnitude of collective flow - directed @xmath96 and elliptic @xmath97 - on @xmath98 @xcite . so , there is a natural suggestion that the @xmath30 fluctuations measured by @xmath1 may be caused by the collective flow . this suggestion was checked by performing a simple monte carlo analysis generating events with independent particles , following the measured inclusive @xmath30 distribution . the uniform azimuthal angle distribution of the events was modified by the collective elliptic flow but the particles remained independent from each other with respect to their transverse momenta . for such events the dynamical @xmath30 fluctuations vanish if the azimuthal angle acceptance is complete . however , the azimuthal angle acceptance of the na49 detector is not flat , and consequently the azimuthal anisotropy generates a finite value of @xmath1 , even though the @xmath30 of particles are independent of each other . it was found that an abnormally large value of @xmath99 leads to @xmath1 as large as 17 mev / c , but a realistic magnitude of @xmath100 , which depends on @xmath30 ( according to @xcite taken as @xmath101(gev / c ) ) , results in a value of @xmath1 which is consistent withzero ( @xmath102 mev / c ) . thus , one concludes that the effect of the azimuthal anisotropy caused by the collective flow combined with the incomplete azimuthal acceptance is not responsible for the observed dynamical @xmath30 fluctuations . in fig . [ fipt_hijing ] the dependence of @xmath1 on the mean number of wounded nucleons is directly compared to predictions of the hijing @xcite model ( default parameters were used ) for all charged particles , and for negatively charged and positively charged particles separately . the same kinematic cuts are applied as for the data . the black lines represent the results of the hijing simulations where the effect of the limited na49 acceptance ( @xmath30 versus azimuthal angle ) is taken into account . the gray lines refer to the hijing predictions for full azimuthal acceptance to demonstrate the effect of the limited acceptance of the detector . in contrast to the data , @xmath1 computed within the hijing model does not change when going from elementary to central pb+pb collisions because the hijing model represents an essentially independent superposition of n+n interactions . the effects of short range correlations ( bose - einstein and coulomb ) have not been incorporated in the hijing model . however it was estimated in the previous analysis @xcite that the combined effect of short range correlations produces @xmath1 values on the level of 5 mev / c for central pb+pb collisions . this effect strongly depends on multiplicity and becomes negligible for p+p interactions . for the hijing model @xmath1 values for positively charged particles are , as in the case of real data , lower than for negatively charged and for all charged particles . the fact that @xmath1 values for positively charged particles are always lower than for negatively charged ones has been found ( using hijing ) to be related to the limited acceptance and treating protons as pions ( assuming the pion mass for all produced particles ) . when @xmath1 was first introduced @xcite it was believed that its value would be non - zero for elementary interactions ( mainly due to the @xmath90 versus @xmath32 dependence ) and would vanish for heavier colliding systems as a result of equilibration . the present measurements do not confirm this expectation . although @xmath1 is close to zero for central pb+pb collisions , the maximum value is observed not for p+p data , but for colliding systems with @xmath103 40 . although the value of @xmath1 is small for p+p collisions , a significant structure appears in the two - particle correlation plot ( fig . [ 2d_plots_b ] ) . the first candidate for its origin is the dependence of @xmath90 on @xmath32 , observed for elementary interactions @xcite . [ pt_n ] compares @xmath104 versus @xmath32 for the hijing model and for real p+p data ( @xmath43 represents averaging over events with a given @xmath32 ) . the hijing model includes the na49 acceptance and all kinematic restrictions . it shows good agreement with the measurements . [ pp_2d ] presents the p+p two - particle correlation plots for data ( a ) , for the hijing model ( b ) and for a simple random generator , which reproduces the dependence of @xmath90 on @xmath32 observed in the data ( c ) . the accepted particle multiplicity distribution for the random generator ( c ) is the same as in the case of data ( all kinematic cuts and na49 geometric acceptance are included ) . both models ( b ) and ( c ) qualitatively reproduce the structure of the two - particle correlation plot observed in the data , however , the hijing model ( b ) shows additionally a small enhancement of the point density in the region of high @xmath47 . the @xmath1 value calculated for model ( c ) for all charged particles equals 1.2 @xmath105 0.2 @xmath105 1.8 mev / c and is consistent with @xmath1 for the p+p data ( 2.2 @xmath105 0.3 @xmath105 1.2 mev / c ) . nevertheless the small difference might indicate that , in agreement with the analysis presented in @xcite , there could be an additional source of correlations present in the data . however , due to the relatively high statistical and systematic errors the effect can not be estimated quantitatively . a similar analysis was carried out for negatively charged particles only , where the correlations caused by resonances and by charge conservation are expected to be smaller than for all charged particles . there , the experimental value of @xmath106 is 0.8 @xmath105 0.1 @xmath105 1.2 mev / c while the model ( c ) gives 0.6 @xmath105 0.2 @xmath105 0.3 mev / c . two - particle correlation plots for negatively charged particles only are very similar for data and for model ( c ) . thus , one concludes that the results on negative particles from p+p interactions are consistent with the conjecture @xcite that the particles are emitted independently but that their @xmath107 and @xmath30 distributions depend on event multiplicity . the last panel ( d ) of fig . [ pp_2d ] presents the result of a simple temperature fluctuation model ( the concept is described in @xcite ) , which assumes that the only source of fluctuations is event - by - event fluctuation of the inverse slope parameter ( @xmath108 ) of the transverse mass spectra . the model assumes a gaussian shaped rapidity distribution and an exponential shape of the transverse mass distribution with the mean inverse slope parameter @xmath109 = 152 mev adjusted to agree with the experimental p+p results . all kinematic cuts applied for the real data are also used in this model and the effect of the finite detector acceptance is taken into account . the mean multiplicity of all accepted particles is the same as in the data . the fluctuations of the inverse slope parameter lead to a saddle shaped structure in the two - particle correlation plots . panel d ) shows the result for fluctuations of @xmath108 on the level of about 10% ( the dispersion @xmath110 = 16 mev ) . because of the difference between the panels a ) and d ) one concludes that fluctuations of the inverse slope parameter are not the ( main ) source of correlations in p+p data . the hijing model has also been used to obtain a two - particle correlation plot for c+c collisions ( fig . [ cc_2d ] b ) , which appears to be similar to that observed for real events ( fig . [ 2d_plots_b ] ) . fig . [ cc_2d ] a ( the same as fig . [ pp_2d ] b ) presents p+p events simulated by the hijing model . the structure observed for p+p collisions vanishes for heavier systems due to the dilution effect from the higher number of uncorrelated particles ( resulting from different n+n interactions ) , whereas the @xmath1 measure is not affected by this dilution effect . in order to see how dynamical fluctuations influence two - particle correlation plots for central pb+pb data the above model with fluctuations of the inverse slope parameter was used again . the mean inverse slope parameter @xmath109 was set to 190 mev and the mean multiplicity of all accepted particles to 200 in order to compare the results with central pb+pb collisions . the inverse slope parameter varied from event to event with a gaussian shaped distribution of width @xmath110 . [ toy ] presents @xmath50 plots for different levels of the inverse slope parameter fluctuations . the fluctuations lead to a saddle shaped structure which is not visible in central pb+pb collisions . one can thus exclude significant @xmath108 fluctuations in central pb+pb collisions at top sps energy . in fig . [ fipt_deltat ] the predicted dependence of @xmath1 on @xmath108 fluctuations @xcite is plotted and compared to @xmath1 measured for the 5% most central pb+pb interactions . the experimental @xmath1 value contains both short and long range correlations . the solid line corresponds to the @xmath108-fluctuation model presented in @xcite which does not include short range correlations . the dashed line is the combination of this model with a contribution of short range correlations estimated experimentally @xcite as 5 mev / c , in agreement with theoretical arguments @xcite . one sees that the observed value of @xmath1 is already below the contribution of the bose - einstein correlations and that the inclusion of slope fluctuations makes the difference even larger . thus , one can conclude , in agreement with the previous results @xcite , that the data leave no space for significant @xmath108 fluctuations provided they are not canceled by other negative correlations . an increase of transverse momentum fluctuations was predicted @xcite to occur in a+a collisions which freeze out near a second order critical end point of the qcd phase diagram . based on calculations from @xcite and the numbers given in table [ data ] it can be estimated that such critical fluctuations alone should result in @xmath111 mev / c . this number is significantly larger than the maximum value of @xmath1 found in this analysis - for peripheral pb+pb interactions ( @xmath112 mev / c ) . note , however , that in this theoretical estimate the effect of the limited experimental acceptance was not taken into account . transverse momentum fluctuations in a+a collisions and elementary interactions were measured by several experiments in the sps energy range . a value of @xmath113 mev / c for charged hadrons was reported by the na22 experiment in @xmath114 and @xmath115 interactions @xcite at 250 gev in a rapidity acceptance similar to the one used in this paper . the value obtained by the na22 experiment is higher than the result presented in this paper for p+p interactions at 158 @xmath0gev ( @xmath116 mev / c ) . this can be caused by the differences in azimuthal acceptance , energy and types of interacting hadrons . the ceres experiment at the sps measured @xmath117 mev / c for charged hadrons in central pb+au collisions at 158 @xmath0gev @xcite . this measurement was performed close to mid - rapidity ( pseudo - rapidity @xmath118 ) in the @xmath30 range @xmath119 gev / c and with full azimuthal angle acceptance . a somewhat smaller value , @xmath120 mev / c , is found in this paper for the similar pb+pb reaction at forward rapidities . significant non - statistical fluctuations ( @xmath121 mev / c ) were measured for charged hadrons at mid - rapidity in central au+au collisions at @xmath122 = 130 gev by the star experiment @xcite . at the rhic energies @xmath30 fluctuations are expected to be larger than at the sps due to a significant contribution of correlated particles originating from ( mini-)jet fragmentation . a non - monotonic centrality dependence of @xmath30 fluctuations , with a maximum for semi - central au+au collisions , was reported by the phenix experiment at @xmath122 = 200 gev @xcite . this result is in qualitative agreement with the result presented in this paper for pb+pb collisions at 158 @xmath0gev . a centrality dependence of @xmath30 fluctuations was observed also by the star experiment at @xmath122 = 130 gev @xcite . it should be stressed that currently available event - by - event measurements were performed in a limited acceptance , the multiplicity of observed particles being below 20% of the total multiplicity . consequently , the sensitivity of these measurements , in particular to long range correlations is reduced . it is desirable to perform future event - by - event measurements in an extended acceptance . transverse momentum event - by - event fluctuations were studied for p+p , c+c , si+si and pb+pb collisions at 158 @xmath0gev . the analysis was limited to the forward rapidity region . three different characteristics were measured : the fluctuations of average transverse momentum ( @xmath90 ) of the event , the @xmath1 fluctuation measure , and transverse momentum two - particle correlations . all measured @xmath1 values are below 10 mev / c and are much smaller than the dispersion of the inclusive @xmath30 distribution . however , the correlations observed in p+p collisions are not simply more and more diluted when going to heavier colliding systems as could be expected if the created matter approaches a higher level of equilibrium with increasing system size . instead , a significant system size dependence of the @xmath1 measure is seen with a maximum for peripheral pb+pb collisions with @xmath103 40 . the two - particle correlation plot for p+p data shows a prominent structure which was found to be connected with the dependence of @xmath90 on @xmath32 . this structure disappears when going to heavier colliding systems . instead short range correlations become visible as an enhancement of the point density in the region close to the diagonal . this effect is strongest for the most central pb+pb interactions . no structure characteristic of event - by - event temperature fluctuations is observed . in future , a study of the energy dependence of transverse momentum fluctuations in the cern sps energy range is planned using the na49 pb+pb collision data taken at different beam energies . the aim is to search for possible anomalies connected with the onset of the deconfinement phase transition , which is indicated by features of pion and strangeness production at low sps energies @xcite . this work was supported by the director , office of energy rese arch , division of nuclear physics of the office of high energy and nuclear physics of the us department of energy ( de - aco3 - 76sfooo98 and de - fg02 - 91er40609 ) , the us national science foundation , the bundesministerium fr bildung und forschung , germany , the alexander von humboldt foundation , the uk engineering and physical sciences research council , the polish state committee for scientific research ( 2 p03b 130 23 , spb / cern / p-03 /dz 446/2002 - 2004 , 2 p03b 02418 , 2 p03b 04123 ) , the hungarian scientific research foundation ( t032648 , t14920 and t32293 ) , hungarian national science foundation , otka , ( f034707 ) , the ec marie curie foundation , and the polish - german foundation . each bin & @xmath65 & @xmath64 range [ fm]p+p & 570 000 & 0.9 & 2 & c+c & 33 000 & 0.153 & 14 & 0 - 2.0 si+si & 63 000 & 0.122 & 37 & 0 - 2.6 pb+pb(6 ) & 117 000 & 0.57 & 42 & 10.2 - pb+pb(5 ) & 59 000 & 0.10 & 88 & 9.1 - 10.2 pb+pb(4 ) & 68 000 & 0.10 & 134 & 7.4 - 9.1 pb+pb(3 ) & 68 000 & 0.11 & 204 & 5.3 - 7.4 pb+pb(2 ) & 45 000 & 0.075 & 281 & 3.4 - 5.3 pb+pb(1 ) & 180 000 & 0.05 & 352 & 0 - 3.4 @xmath123 $ ] & @xmath124 $ ] 3.9 - 4.1 & 0 & 0.3 & 6500 4.1 - 4.3 & 0 & 0.3 & 5500 4.3 - 4.5 & 0 & 0.25 & 4500 4.5 - 4.7 & 0 & 0.25 & 3500 4.7 - 4.9 & 0 & 0.2 & 2500 4.9 - 5.1 & 0.5 & 0.2 & 2500 5.1 - 5.3 & 1.0 & 0.1 & 2500 5.3 - 5.5 & 1.5 & 0.1 & 2500 @xmath125 & @xmath126 [ mev / c ] & @xmath89 [ mev / c ] & @xmath1 [ mev / c ] p+p ( all ) & 1.4 & 1.3 & 304 & 196 & 2.2 @xmath105 0.3 p+p ( - ) & 0.6 & 0.7 & 283 & 179 & 0.8 @xmath105 0.1 p+p ( + ) & 0.8 & 0.9 & 317 & 206 & -1.4 @xmath105 0.3	results are presented on event - by - event fluctuations in transverse momentum of charged particles , produced at forward rapidities in p+p , c+c , si+si and pb+pb collisions at 158 @xmath0gev . three different characteristics are discussed : the average transverse momentum of the event , the @xmath1 fluctuation measure and two - particle transverse momentum correlations . in the kinematic region explored , the dynamical fluctuations are found to be small . however , a significant system size dependence of @xmath1 is observed , with the largest value measured in peripheral pb+pb interactions . the data are compared with predictions of several models . = -2 cm * transverse momentum fluctuations * * in nuclear collisions at 158 @xmath0gev * * the na49 collaboration * * the na49 collaboration * t. anticic@xmath2 , b. baatar@xmath3,d . barna@xmath4 , j. bartke@xmath5 , m. behler@xmath6 , l. betev@xmath7 , h. biakowska@xmath8 , a. billmeier@xmath7 , c. blume@xmath9 , b. boimska@xmath8 , m. botje@xmath10 , j. bracinik@xmath11 , r. bramm@xmath7 , r. brun@xmath12 , p. buni@xmath13 , v. cerny@xmath11 , p. christakoglou@xmath14 , o. chvala@xmath15 , j.g . cramer@xmath16 , p. csat@xmath4 , n. darmenov@xmath17 , a. dimitrov@xmath17 , p. dinkelaker@xmath7 , v. eckardt@xmath18 , p. filip@xmath18 , d. flierl@xmath7,z . fodor@xmath4 , p. foka@xmath9 , p. freund@xmath18 , v. friese@xmath9 , j. gl@xmath4 , m. gadzicki@xmath7 , g. georgopoulos@xmath14 , e. gadysz@xmath5 , s. hegyi@xmath4 , c. hhne@xmath6 , k. kadija@xmath2 , a. karev@xmath18 , v.i . kolesnikov@xmath3 , t. kollegger@xmath7 , r. korus@xmath19 , m. kowalski@xmath5 , i. kraus@xmath9 , m. kreps@xmath11 , m. van leeuwen@xmath10 , p. lvai@xmath4 , l. litov@xmath17 , m. makariev@xmath17 , a.i . malakhov@xmath3 , c. markert@xmath9 , m. mateev@xmath17 , b.w . mayes@xmath20 , g.l . melkumov@xmath3 , c. meurer@xmath7 , a. mischke@xmath9 , m. mitrovski@xmath7 , j. molnr@xmath4 , st . mrwczyski@xmath19 , g. plla@xmath4 , a.d . panagiotou@xmath14 , d. panayotov@xmath17 , k. perl@xmath21 , a. petridis@xmath14 , m. pikna@xmath11 , l. pinsky@xmath20 , f. phlhofer@xmath6 , j.g . reid@xmath16 , r. renfordt@xmath7 , w. retyk@xmath21 , c. roland@xmath22 , g. roland@xmath22 , m. rybczyski@xmath19 , a. rybicki@xmath23 , a. sandoval@xmath9 , h. sann@xmath9 , n. schmitz@xmath18 , p. seyboth@xmath18 , f. siklr@xmath4 , b. sitar@xmath11 , e. skrzypczak@xmath21 , g. stefanek@xmath19 , r. stock@xmath7 , h. strbele@xmath7 , t. susa@xmath2 , i. szentptery@xmath4 , j. sziklai@xmath4 , t.a . trainor@xmath16 , d. varga@xmath4 , m. vassiliou@xmath14 , g.i . veres@xmath4 , g. vesztergombi@xmath4 , d. vrani@xmath9 , a. wetzler@xmath7 , z. wodarczyk@xmath19 i.k . yoo@xmath24 , j. zaranek@xmath7 , j. zimnyi@xmath4 @xmath10nikhef , amsterdam , netherlands . + @xmath14department of physics , university of athens , athens , greece . + @xmath11comenius university , bratislava , slovakia . + @xmath4kfki research institute for particle and nuclear physics , budapest , hungary . + @xmath24department of physics , pusan national university , busan , republic of korea . + @xmath22mit , cambridge , usa . + @xmath5institute of nuclear physics , cracow , poland . + @xmath9gesellschaft fr schwerionenforschung ( gsi ) , darmstadt , germany . + @xmath3joint institute for nuclear research , dubna , russia . + @xmath7fachbereich physik der universitt , frankfurt , germany . + @xmath12cern , geneva , switzerland . + @xmath20university of houston , houston , tx , usa . + @xmath19institute of physics wietokrzyska academy , kielce , poland . + @xmath6fachbereich physik der universitt , marburg , germany . + @xmath18max - planck - institut fr physik , munich , germany . + @xmath15institute of particle and nuclear physics , charles university , prague , czech republic . + @xmath16nuclear physics laboratory , university of washington , seattle , wa , usa . + @xmath17atomic physics department , sofia university st . kliment ohridski , sofia , bulgaria . + @xmath8institute for nuclear studies , warsaw , poland . + @xmath21institute for experimental physics , university of warsaw , warsaw , poland . + @xmath2rudjer boskovic institute , zagreb , croatia .
recent experimental data [ 14 ] on the production of strangeness in pp collisions , p + p @xmath0 p + @xmath1 + k@xmath2 [ 14 ] and p + p @xmath0 p + @xmath3 + k@xmath2 [ 4 ] , for energies of colliding protons in the region near thresholds of the final states p@xmath1k@xmath2 and p@xmath3k@xmath2 have represented experimental values of the cross sections @xmath4 and @xmath5 with high precision , where @xmath6 is an excess of energy that we define below . as has been obtained in ref.[4 ] the cross section for the @xmath3 hyperon , p + p @xmath0 p + @xmath3 + k@xmath2 , exceeds by a factor 28 the cross section for the production of @xmath1 hyperon , p + p @xmath0 p + @xmath1 + k@xmath2 , measured for the equivalent excess energies . these data give a possibility for testing of various theoretical approaches to mechanisms of a strangeness production from nucleons that is important for correct description of a strangeness production in heavy ion collisions . the parer is organized as follows . in section1 we calculate the effective lagrangians of the transitions p + p @xmath0 p + @xmath7 + k@xmath2 and p + p @xmath0 p + @xmath3 + k@xmath2 in the one pion exchange approximation and at leading order in momentum expansion in powers of the momenta of final state particles . in section2 we calculate the cross sections for the reactions p + p @xmath0 p + @xmath7 + k@xmath2 and p + p @xmath0 p + @xmath3 + k@xmath2 and compare the theoretical results with the experimental data . in the conclusion we discuss the obtained results and the polarization properties of the @xmath1 and @xmath3 hyperons . in our approach to the description of the reactions p + p @xmath0 p + @xmath1 + k@xmath2 and p + p @xmath0 p + @xmath3 + k@xmath2 , first , we suggest to investigate the transitions p + p @xmath0 p + @xmath1 + k@xmath2 and p + p @xmath0 p + @xmath3 + k@xmath2 , where the wave functions of the protons in the initial pp state are described by plane waves and all particles in the final states p@xmath1k@xmath2 and p@xmath3k@xmath2 are decoupled . these transitions we define by the effective lagrangians @xmath9 and @xmath10 . for the evaluation of these effective lagrangian we suggest to use a simplest one pion exchange approximation . the feynman diagrams defining the effective lagrangians @xmath9 and @xmath11 are depicted in fig.1 . the analytical expressions corresponding to these diagrams read @xmath12 \,\frac{\displaystyle g^2_{\pi nn}}{\displaystyle m^2_{\pi } - ( p_{\rm p } - p_1)^2}\,[\bar{u}(\vec{p}_{\lambda},\alpha_{\lambda})i\gamma^5\ , \frac{\displaystyle g_{p\lambda k^+}}{\displaystyle m_{\rm p } - \hat{p}_{\lambda } - \hat{p}_{k}}\,i\gamma^5 u(\vec{p}_2,\alpha_2)]\nonumber\\ \hspace{-0.5 in } & & - [ \bar{u}(\vec{p}_{\rm p},\alpha_{\rm p})i\gamma^5 u(\vec{p}_2,\alpha_2)]\,\frac{\displaystyle g^2_{\pi nn}}{\displaystyle m^2_{\pi } - ( p_{\rm p } - p_2)^2}\,[\bar{u}(\vec{p}_{\lambda},\alpha_{\lambda})i\gamma^5\ , \frac{\displaystyle g_{p\lambda k^+}}{\displaystyle m_{\rm p } - \hat{p}_{\lambda } - \hat{p}_{k}}\,i\gamma^5 u(\vec{p}_1,\alpha_1)],\end{aligned}\ ] ] where @xmath13 is the coupling constant of the @xmath14nn interaction , @xmath15 are bispinors of the protons for @xmath16 and the @xmath1hyperon for @xmath17 with polarizations @xmath18 . then , @xmath19 and @xmath20 are the masses of a @xmath21 meson and a proton . the amplitude of the transition p + p @xmath0 p + @xmath3 + k@xmath2 can obtained from eq.([label2.1 ] ) by a replacement @xmath22 and @xmath23 . we would like to accentuate that we are using the pseudoscalar couplings for the description of the @xmath21 and the k@xmath2 meson coupled to baryons that always fit data well [ 5 ] . in the center of mass frames of the colliding protons and the p@xmath1 system the amplitude eq.([label2.1 ] ) takes the form @xmath24\nonumber\\ \hspace{-0.5in}&&\times\,\frac{\displaystyle g^2_{\pi nn}}{\displaystyle m^2_{\pi } - \big(\sqrt{m^2_{\rm p } + ( \vec{q}_{\rm p\lambda } + \vec{p}_{\rm k}/2)^2 } - \sqrt{m^2_{\rm p } + \vec{p}^{\;2}}\,\big)^2 + ( \vec{p } + \vec{q}_{\rm p\lambda } + \vec{p}_{\rm k}/2)^2}\nonumber\\ \hspace{-0.5in}&&\times\,[\bar{u}(\vec{q}_{\rm p\lambda } -\vec{p}_{\rm k}/2,\alpha_{\lambda})i\gamma^5\,\nonumber\\ \hspace{-0.5in}&&\times\,\frac{\displaystyle g_{p\lambda k^+}}{\displaystyle m_{\rm p } - \gamma^0\big(\sqrt{m^2_{\lambda } + ( \vec{q}_{\rm p\lambda } -\vec{p}_{\rm k}/2)^2 } + \sqrt{m^2_{\rm k } + \vec{p}^{\;2}_{\rm k}}\,\big ) + \vec{\gamma}\cdot(\vec{q}_{\rm p\lambda } + \vec{p}_{\rm k}/2)}\,i\gamma^5 u(-\vec{p},\alpha_2)]\nonumber\\ \hspace{-0.5 in } & & -[\bar{u}(-\vec{q}_{\rm p\lambda } -\vec{p}_{\rm k}/2,\alpha_{\rm p})i\gamma^5u(- \vec{p},\alpha_2)]\nonumber\\ \hspace{-0.5in}&&\times\,\frac{\displaystyle g^2_{\pi nn}}{\displaystyle m^2_{\pi } - \big(\sqrt{m^2_{\rm p } + ( \vec{q}_{\rm p\lambda } + \vec{p}_{\rm k}/2)^2 } - \sqrt{m^2_{\rm p } + \vec{p}^{\;2}}\,\big)^2 + ( \vec{p } - \vec{q}_{\rm p\lambda } - \vec{p}_{\rm k}/2)^2}\nonumber\\ \hspace{-0.5in}&&\times\,[\bar{u}(\vec{q}_{\rm p\lambda } -\vec{p}_{\rm k}/2,\alpha_{\lambda})i\gamma^5\,\nonumber\\ \hspace{-0.5in}&&\times\,\frac{\displaystyle g_{p\lambda k^+}}{\displaystyle m_{\rm p } - \gamma^0\big(\sqrt{m^2_{\lambda } + ( \vec{q}_{\rm p\lambda } -\vec{p}_{\rm k}/2)^2 } + \sqrt{m^2_{\rm k } + \vec{p}^{\;2}_{\rm k}}\,\big ) + \vec{\gamma}\cdot(\vec{q}_{\rm p\lambda } + \vec{p}_{\rm k}/2)}\,i\gamma^5 u(\vec{p},\alpha_1)],\end{aligned}\ ] ] where @xmath25 is a relative momentum of the colliding protons , @xmath26 is a relative momentum of the p@xmath1 system , and @xmath27 is the momentum of the k@xmath2 meson . the reaction p + p @xmath0 p + @xmath8 + k@xmath2 is determined experimentally very close to threshold of the final state p@xmath1k@xmath2 ( or p@xmath3k@xmath2 ) . the minimum relative 3momentum of the initial protons is equal to @xmath28 , where we have used @xmath29 and @xmath30 , the masses of the @xmath1 hyperon and the k@xmath2 meson [ 6 ] p + @xmath3 + k@xmath2 the minimum relative momentum of the colliding protons amounts to @xmath31 at @xmath32 [ 6 ] . ] . due to this close vicinity to threshold the momentum of the k@xmath2 meson and the relative movement of the p@xmath1 system ( or p@xmath3 ) are smaller compared with all energy scales of the coupled particles . this allows to expand the matrix element eq.([label2.2 ] ) in powers of @xmath27 and @xmath26 by keep leading contributions : @xmath33\,[\bar{u}(\vec{q}_{\rm p\lambda } -\vec{p}_{\rm k}/2,\alpha_{\lambda } ) u(-\vec{p},\alpha_2)]\nonumber\\ \hspace{-0.5 in } & & - \bar{u}(-\vec{q}_{\rm p\lambda } -\vec{p}_{\rm k}/2,\alpha_{\rm p})i\gamma^5u(- \vec{p},\alpha_2)]\,[\bar{u}(\vec{q}_{\rm p\lambda } -\vec{p}_{\rm k}/2,\alpha_{\lambda } ) u(\vec{p},\alpha_1)]\}.\end{aligned}\ ] ] by introducing the effective coupling constant @xmath34 @xmath35 we can write down the effective lagrangian @xmath36 of the transition p + p @xmath0 p + @xmath1 + k@xmath2 . that reads @xmath37[\bar{\lambda}(x)p(x)],\end{aligned}\ ] ] where @xmath38 and @xmath39 are the operators of the interpolating proton and @xmath1hyperon fields , and @xmath40 is the operator of the interpolating k@xmath2meson field . by making the replacement @xmath41 in eq.([label2.5 ] ) we obtain the effective lagrangian @xmath42 of the transition p + p @xmath0 p + @xmath3 + k@xmath2 . it is convenient to represent the effective lagrangian in terms of the interactions describing the p@xmath1 system in the certain spin state . this can be carried out by means of a fierz transformation [ 7 ] . by performing a fierz transformation we recast the effective lagrangian @xmath9 into the form @xmath43[\bar{p^c}(x)p(x)]\nonumber\\ & & + [ \bar{p}(x)\lambda^c(x)][\bar{p^c}(x)\gamma^5p(x ) ] + [ \bar{p}(x)\gamma^{\mu}\lambda^c(x)][\bar{p^c}(x)\gamma_{\mu}\gamma^5p(x)]\}.\end{aligned}\ ] ] the first term and the last two in the lagrangian eq.([label2.6 ] ) describe the p@xmath1 system coupled in the spin singlet and triplet state , respectively . since near threshold the p@xmath1 ( or p@xmath3 ) system couples mainly in the spin singlet state , @xmath44 , we should leave only the first term . this yields the following effective lagrangian @xmath45[\bar{p^c}(x)p(x)].\end{aligned}\ ] ] the wave functions of the particles in the transition p + p @xmath0 p + @xmath1 + k@xmath2 are plane waves . in order to describe a physical reaction p + p @xmath0 p + @xmath1 + k@xmath2 we suggest to take into account interactions between particles both in the final and in the initial state . the contribution of the interaction in the p@xmath1system can be obtained by summing up an infinite series of one baryon loop diagrams with a point like @xmath47 coupling describing a low energy transition p + @xmath1 @xmath0 p + @xmath1 [ 7 ] . after the evaluation of momentum integrals and the renormalization of the wave functions of the proton and the @xmath1 hyperon [ 7 ] we can represent the contribution of this series in the phenomenological form in terms of the s wave scattering length @xmath48 and the effective range @xmath49 : @xmath50 this is well known watson form for the final state interaction [ 8 ] having been used by balewski _ _ [ 9 ] for the description of the final p@xmath1 interaction in the reaction p + p @xmath0 p + @xmath1 + k@xmath2 . below we use the numerical values of the s wave scattering length and the effective range , @xmath51 and @xmath52 , recommended by balewski _ we would like to emphasize that since finally the contribution of the final state interaction is expressed in terms of phenomenological parameters , the s wave scattering length and the effective range taken from experimental data , a knowledge of an explicit value of a coupling constant of a local @xmath47 interaction describing a low energy transition p + @xmath1 @xmath0 p + @xmath1 and defining vertices in the one baryon loop diagrams is not important [ 7 ] . for the analysis of elastic low energy p@xmath3 scattering we assume that @xmath53 and @xmath54 . below we show that this assumption does not contradict the experimental data [ 4 ] . unlike the p@xmath1 interaction in the final state in order to describe the interaction in the initial pp state we have to specify the coupling constant of the transition p + p @xmath0 p + p. since the relative momentum of the pp state is comeasurable with a mass of a proton , so that a coulomb repulsion between protons can be neglected . we suggest to describe the pp interaction in the one pion exchange approximation . as experimentally relative momenta of the pp state differ slightly from the threshold momentum , we can represent the pp interaction describing the transition p + p @xmath0 p + p in the following local form @xmath55[\bar{p^c}(x)p(x)].\end{aligned}\ ] ] this effective lagrangian describes the pp system coupled in the spin triplet state . in the factor @xmath56 the multiplier @xmath57 is caused by a fierz transformation of the one pion exchange interaction [ 7 ] . the phenomenological coupling constant @xmath58 is then defined by @xmath59 it is obtained from the one pion exchange diagram of the transition p + p @xmath0 p + p by averaging over all possible directions of a relative momentum of the final pp state . we would like to accentuate that since relative momenta of the incident protons are very close to threshold , the quantity @xmath58 is practically constant determined by @xmath60 at @xmath61 or @xmath62 for the reactions p + p @xmath0 p @xmath1 + k@xmath2 and p + p @xmath0 p @xmath3 + k@xmath2 , respectively . by summing up an infinite series of one proton loop diagrams the vertices of which are defined by the effective interaction eq.([label3.2 ] ) we arrive at the expression [ 7 ] @xmath63 \to \frac{[\bar{u^c}(-\vec{p},\alpha_2)u(\vec{p},\alpha_1)]}{\displaystyle 1 + \frac{c_{\rm pp}}{64\pi^2}\int\frac{d^4k}{\pi^2 i}\,{\rm tr}\bigg\{\frac{1}{m_{\rm p } - \hat{k}}\frac{1}{m_{\rm p } - \hat{k } - \hat{p}}\bigg\}},\end{aligned}\ ] ] where @xmath64 . after the subtraction of trivial @xmath65independent divergent contributions and the renormalization of the wave functions of the protons we obtain the contribution of the interaction of the protons in the initial state @xmath66 \to \frac{[\bar{u^c}(-\vec{p},\alpha_2)u(\vec{p},\alpha_1)]}{\displaystyle 1 + \frac{c_{\rm pp}}{64\pi^2}\int\frac{d^4k}{\pi^2 i}\,{\rm tr}\bigg\{\frac{1}{m_{\rm p } - \hat{k}}\frac{1}{m_{\rm p } - \hat{k } - \hat{p}}\bigg\}}\nonumber\\ & & \to \frac{[\bar{u^c}(-\vec{p},\alpha_2)u(\vec{p},\alpha_1)]}{\displaystyle 1 + \frac{c_{\rm pp}(\vec{p}^{\,2},\lambda)}{8\pi^2}\,\frac{\|\vec{p}\,\|^3}{\sqrt{\vec{p}^{\,2 } + m^2_{\rm p}}}\left[{\ell n}\left(\frac{\sqrt{\vec{p}^{\,2 } + m^2_{\rm p } } + \|\vec{p}\,\|}{\sqrt{\vec{p}^{\,2 } + m^2_{\rm p } } - @xmath67 is given by @xmath68}.\end{aligned}\ ] ] the appearance of the cut off @xmath1 is caused by non trivial @xmath65dependent logarithmically divergent contributions . the cut off @xmath1 restricts from above 3momenta of virtual proton fluctuations . since our approach is an effective one , so the dependence of the amplitude on the cut off seems to be usual [ 10,11 ] . the only thing one needs is to choose the value of the cut off by an appropriate physical way . in our case it is reasonable to have @xmath1 to be of order a mass of a resonance nearest to a nucleon , that is the @xmath69 resonance [ 6 ] . therefore , in our calculations we would set @xmath70 . thus , the amplitude of the reaction p + p @xmath0 p + @xmath1 + k@xmath2 near threshold of the final p@xmath1k@xmath2 state is defined by @xmath71}{\displaystyle 1 - \frac{1}{2}\,a_{\rm p\lambda}r_{\rm p\lambda}q^2_{\rm p\lambda } + i\,a_{\rm p\lambda}q_{\rm p\lambda}}\nonumber\\ \hspace{-0.5in}&&\times\ , \frac{[\bar{u^c}(-\vec{p},\alpha_2)u(\vec{p},\alpha_1 ) ] } { \displaystyle 1 + \frac{c_{\rm pp}(\vec{p}^{\,2},\lambda)}{8\pi^2}\,\frac{\|\vec{p}\,\|^3}{\sqrt{\vec{p}^{\,2 } + m^2_{\rm p}}}\left[{\ell n}\left(\frac{\sqrt{\vec{p}^{\,2 } + m^2_{\rm p } } + \|\vec{p}\,\|}{\sqrt{\vec{p}^{\,2 } + m^2_{\rm p } } - \hspace{-0.5in}&&\times\,\sqrt { \frac{m_{\rm pk^+}}{q_{\rm pk^+}}\,\frac{2\pi \alpha } { \displaystyle e^{\textstyle 2\pi \alpha m_{\rm pk^+}/q_{\rm pk^+ } } - 1}},\end{aligned}\ ] ] where the last factor depending of @xmath72 , the fine structure constant , and @xmath73 , a relative momentum of the pk@xmath2 system , takes into account the coulomb repulsion between the daughter proton and the k@xmath2 meson at low relative energies [ 9 ] ( see also [ 7 ] ) , @xmath74 is the reduced mass of the pk@xmath2 system . then , relative momenta of the pp system are very close to threshold , @xmath75 . hence , we can calculate the contribution of the interactions in the pp state numerically . this gives @xmath76 } = 0.308\,e^{\textstyle -i\,46.6 ^ 0}.\end{aligned}\ ] ] for the reaction p + p @xmath0 p + @xmath3 + k@xmath2 we get @xmath77 . by calculating numerically a part of the coupling constant @xmath78 given by eq.([label2.4 ] ) we reduce the amplitude of the reaction p + p @xmath0 p + @xmath1 + k@xmath2 to the following form @xmath79[\bar{u^c}(-\vec{p},\alpha_2 ) u(\vec{p},\alpha_1)]\nonumber\\ & & \times\,\sqrt{\frac{m_{\rm pk^+}}{q_{\rm pk^+}}\,\frac{2\pi \alpha } { \displaystyle e^{\textstyle 2\pi \alpha m_{\rm pk^+}/q_{\rm pk^+ } } - 1}}.\end{aligned}\ ] ] for the reaction p + p @xmath0 p + @xmath3 + k@xmath2 the numerical factor is equal to @xmath80 . the amplitude squared , averaged over polarizations of the initial protons and summed over polarizations of the final baryons amounts to @xmath81 for the reaction p + p @xmath0 p + @xmath3 + k@xmath2 the numerical factor acquires the value @xmath82 . the cross section for the reaction p + p @xmath0 p + @xmath1 + k@xmath2 reads @xmath83 where the cross section and the excess of energy @xmath84 are measured in ( nb ) and ( mev ) , respectively . the function @xmath85 related to the phase volume of the reaction is defined by @xmath86 and normalized to unity at @xmath87 and @xmath88 . the cross section for the reaction p + p @xmath0 p + @xmath3 + k@xmath2 can be evaluated in analogy with the reaction p + p @xmath0 p + @xmath1 + k@xmath2 and reads @xmath89 the function @xmath90 results from eq.([label3.12 ] ) via a replacement @xmath41 . in terms of the axial vector coupling constants @xmath91 and @xmath92 and @xmath93 the coupling constants @xmath94 and @xmath95 are defined by [ 12 ] @xmath96 the cross sections eqs.([label3.11 ] ) and ( [ label3.13 ] ) then read @xmath97 for the numerical analysis the functions @xmath98 and @xmath90 can be given in the more convenient form @xmath99 where we have denoted @xmath100 the function @xmath90 should be obtained from the function @xmath85 given by eq.([label3.16 ] ) via a simple replacement @xmath101 . the numerical values of these functions are tabulated in tables 1 and 2 , respectively . for numerical calculations of the cross sections we use @xmath102 and @xmath103 [ 13 ] . the numerical values of the cross sections for the excess of energy @xmath6 ranging over the region @xmath104 [ 1,2,4 ] are adduced in table 1 and 2 . we have developed a phenomenological approach to the description of the reactions p + p @xmath0 p + @xmath1 + k@xmath2 and p + p @xmath0 p + @xmath3 + k@xmath2 near thresholds of the final states p@xmath1k@xmath2 and p@xmath3k@xmath2 , respectively . the theoretical cross section for the reaction p + p @xmath0 p + @xmath1 + k@xmath2 agrees reasonably well with the experimental data . the numerical values of the cross sections are adduced in table 1 . we show that for the excess of energy @xmath6 ranging over the region @xmath104 the cross section is proportional to @xmath105 , @xmath106 . this fits well the experimental value @xmath107 [ 2 ] . the cross section for the reaction p + p @xmath0 p + @xmath1 + k@xmath2 has been also measured for the higher excess of energies [ 4 ] : @xmath108 , @xmath109 and @xmath110 . our theoretical predictions for these energies read : @xmath111 , @xmath112 and @xmath113 . for the calculation of these cross sections we have also taken into account the momentum dependence of the structure function @xmath114 . we would like to emphasize that within our approach the theoretical cross section for the reaction p + p @xmath0 p + @xmath1 + k@xmath2 fits well the experimental values just at @xmath115 . really , at @xmath116 the theoretical value of the cross section @xmath117 agrees well with the experimental one @xmath118 [ 3 ] . in average the accuracy of the agreement between the theoretical cross section for the reaction p + p @xmath0 p + @xmath1 + k@xmath2 and the experimental data [ 24 ] is about 11@xmath119 . however , we can not pass by the fact that the experimental value of the cross section measured at @xmath120 [ 3 ] : @xmath121 is by a factor 1.7 smaller compared with the theoretical one : @xmath122 . since the theoretical cross section in the excess of energy region @xmath123 is a smooth function of @xmath6 proportional to @xmath105 and fits reasonably well the experimental value of the cross section at @xmath115 , we argue that the result obtained at @xmath124 seems to be underestimated and demands to be remeasured . the cross section for the reaction p + p @xmath0 p + @xmath3 + k@xmath2 is also described well in our approach . the theoretical values of the cross section adduced in table 2 are in reasonable agreement with the experimental data for all excess of energies from the interval @xmath104 . in this energy region the theoretical cross section is proportional to @xmath105 . the average value @xmath125 fits well the experimental data @xmath126 [ 4 ] . at @xmath116 we predict @xmath127 that agrees well with the experimental value @xmath128 [ 3 ] . in our approach the enhancement of the cross section for the reaction p + p @xmath0 p + @xmath1 + k@xmath2 with respect to the cross section for the reaction p + p @xmath0 p + @xmath3 + k@xmath2 is completely a unitary symmetry effect . in fact , the coupling constant @xmath129 is smaller compared by a factor of 0.16 with the coupling constant @xmath94 . this is in agreement with the conclusion given by kaiser [ 11 ] . however , unlike the kaiser s approach to the description of the reactions p + p @xmath0 p + @xmath1 + k@xmath2 and p + p @xmath0 p + @xmath3 + k@xmath2 we point out the dominant role of the contribution of the one pion exchange and the strong interaction of the protons in the initial state . point of our approach to the description of the protons coupled in the initial state is a reduction of pp interaction to a local form via a phenomenological interaction eq.([label3.3 ] ) based on the one pion exchange . by having fixed the phenomenological coupling of a four proton interaction we have succeeded then in deriving the contribution of this interaction to the amplitudes of the reactions p + p @xmath0 p + @xmath1 + k@xmath2 and p + p @xmath0 p + @xmath3 + k@xmath2 via the summation of an infinite series of one proton loop diagrams . after the evaluation of these diagrams and the renormalization of the wave functions of the protons we have arrived at the expression that has turned out to be dependent on a cut off @xmath1 restricting from above 3momenta of virtual proton fluctuations . the appearance of a dependence on a cut off is a usual case in a phenomenological approach to the description of the reactions under consideration [ 10,11 ] . the main point is to fix this parameter in an appropriate way . the neglect of the contribution of baryon resonances to the amplitude of the pp interaction makes a hint that the cut off @xmath1 should be of order of the mass of a nearest resonance that is the @xmath69 resonance . that is why we have set @xmath70 . as has been discussed above this has led to the description of the cross sections for the reactions p + p @xmath0 p + @xmath1 + k@xmath2 and p + p @xmath0 p + @xmath3 + k@xmath2 with accuracy about 11@xmath119 . we would like to underscore that our approach to the description of the protons coupled in the initial state is rather similar ideologically and technically to that having been applied by achasov _ [ 14 ] to the analysis of the contribution of the scalar @xmath130 and @xmath131 mesons treated as four quark states [ 15 ] to the amplitudes of @xmath132 and @xmath14k elastic scattering in the energy region of order of @xmath133 . unlike other available theoretical approaches to the mechanism of @xmath1k@xmath2 and @xmath3k@xmath2 production in pp collisions [ 10,11,16 ] our mechanism does not demand the inclusion of exchanges of all mesons heavier than the @xmath21meson and baryon resonances @xmath134 , @xmath135 and so on . one can show that the summary contribution of the one meson exchanges of @xmath136 , @xmath137 and @xmath138 meson and the scalar isoscalar meson @xmath139 [ 1721 ] is of order of 10@xmath119 relative to the one pion exchange . in fact , the estimate of the summary contribution of the @xmath136 , @xmath139 , @xmath137 and @xmath138 meson exchanges relative to the one pion exchange reads @xmath140 where we have assumed that the @xmath136 is the eighth component of the octet of pseudoscalar mesons and the coupling constant of the @xmath141 interaction is equal to @xmath142 [ 17 ] with @xmath143 , the axial vector coupling constant [ 6 ] . then , we have set @xmath144 is the @xmath145 coupling constant [ 6 ] . the value 9@xmath119 can be reduced by including the contribution of the pseudoscalar @xmath146 meson . this confirms that with a good accuracy the one pion exchange dominates in pp reactions for @xmath1k@xmath2 and @xmath3k@xmath2 production at thresholds of the final p@xmath1k@xmath2 and p@xmath3k@xmath2 states . we do not take into account the contributions of baryon resonances @xmath134 , @xmath135 and so on [ 16 ] . nevertheless , the obtained agreement with the experimental data allows us to think that effectively the contributions of baryon resonances can be partly reproduced by the amplitude of the pp interaction in the initial state . in our approach the daugther proton and the @xmath1 hyperon as well as the daugther proton and the @xmath3 hyperon are in the spin singlet state . this implies that the direction of the spin of the @xmath1 and @xmath3 hyperons is strictly opposite to the direction of the spin of the daugther proton . thereby , according to our approach by measuring a polarization of the daughter proton one measures unambiguously a polarization of the @xmath1 and @xmath3 hyperons . of course , this is true only for the excess of energies very close to thresholds of the final states p@xmath1k@xmath2 and p@xmath3k@xmath2 . for the excess of energies when the contribution of the spin triplet state of the p@xmath1 and p@xmath3 system becomes perceptible the polarizations of the @xmath1 and @xmath3 hyperons are not so strictly determined . we are planning to carry out the analyse of the polarization properties of the @xmath1 and @xmath3 hyperons by taking into account the contribution of the spin triplet states of the p@xmath1 and p@xmath3 systems in our forthcoming publications p + @xmath1 + k@xmath2 and p + p @xmath0 p + @xmath3 + k@xmath2 with polarized colliding protons near thresholds of the final states p@xmath1k@xmath2 and p@xmath3k@xmath2 , respectively , has been carried out by rekalo _ [ 22 ] . ] . recent measurements of the polarization of the @xmath1 hyperon in the reaction p + p @xmath0 p + @xmath1 + k@xmath2 at the excess of energy @xmath147 [ 23 ] evidence an advantage of the k@xmath2meson exchange mechanism [ 24 ] with respect to the one pion exchange one . in our approach being valid for the excess of energies much less than @xmath147 the contribution of the k@xmath2meson exchange makes up about 0.1@xmath119 in comparison with the exchange by the @xmath21 meson . the work is supported in part by the scientific and technical programme of ministry of education of russian federation for fundamental researches in universities of russia . cross sections for the reaction p + p @xmath0 p + @xmath1 + k@xmath2 for the excess of energy ranging over the region @xmath104 . the experimental data are taken from ref.[2 ] . [ cols="^,^,^,^,^,^ " , ] + 9 j. t. balewski _ et al . _ , b * 388 * , 859 ( 1996 ) . j. t. balewski _ et al . _ , b * 420 * , 211 ( 1998 ) . r. bilger _ et al . _ , b * 420 * , 217 ( 1998 ) . s. sewerin _ et al . _ , cosy11 collaboration , nucl . a * 663 * , 473 ( 2000 ) ; phys . lett . * 83 * , 682 ( 1999 ) . l. r. thebaud and m. d. scadron , phys . d * 8 * , 2190 ( 1973 ) ; s. a. coon , m. d. scadron , and b. r. barrett , nucl . a * 249 * , 483 ( 1975 ) ; h. f. jones and m. d. scadron , phys . d * 11 * , 174 ( 1975 ) ; g. e. hite , r. j. jacob , and m. d. scadron , phys . d * 14 * , 1306 ( 1976 ) ; b. h. wilde , s. a. coon , and m. d. scadron , phys . d * 18 * , 4489 ( 1978 ) ; s. a. coon , m. d. scadron , p. c. mcnamee , b. r. barrett , d. w. e. blatt , and b. h. j. mckellar , nucl . phys . a * 317 * , 242 ( 1979 ) ; s. a. coon and m. d. scadron , phys.rev . c * 23 * , 1150 ( 1981 ) ; phys.rev.c * 26 * , 2402 ( 1982 ) ; m. d. scadron , ann . * 159 * , 184 ( 1985 ) ; s. a. coon and m. d. scadron , phys . c * 42 * , 2256 ( 1990 ) ; r. e. karlsen , w. h. ryan , and m. d. scadron , phys . rev . d * 43 * , 157 ( 1991 ) ; r. e. karlsen and m. d , scadron , phys . rev . d * 43 * , 1739 ( 1991 ) ; z. phys . c * 52 * , 325 ( 1991 ) ; phys . d * 45 * , 4113 ( 1992 ) ; m. d. scadron , phys . rev . d * 57 * , 5307 ( 1998 ) ; a. bramon , r. escribano , and m. d. scadron , eur . j. c 7 , 271 ( 1999 ) ; m. d. scadron , mod . a * 14 * , 1349 ( 1999 ) . d. e. groom _ et al . j. c * 15 * , 1 ( 2000 ) . a. n. ivanov , h. oberhummer , n. i. troitskaya , and m. faber , eur . j. a * 8 * , 223 ( 2000 ) . k. m. watson , phys . rev . * 88 * , 1163 ( 1952 ) ; m. l. goldberger , and k. m. watson , in _ collision theory _ , new york , john wiley @xmath148 sons , 1964 j. t. bawelski _ et al . _ , j. a * 2 * , 99 ( 1998 ) . a. sibirtsev , k. tsushima , and a. w. thomas , phys . lett . b * 421 * , 59 ( 1998 ) and references therein . n. kaiser , eur . j. a * 5 * , 105 ( 1999 ) . m. m. nagels _ et al . b * 147 * , 253 ( 1979 ) . m. anselmino and m. d. scadron , phys . b * 229 * , 117 ( 1989 ) ; m. d. scadron , z. phys . c * 54 * , 595 ( 1992 ) ; f. e. close and r. g. roberts , phys . b * 316 * , 165 ( 1993 ) ; x. song , p. k. kabir , and j. s. mccarthy , phys . d * 54 * , 2108 ( 1996 ) . n. n. achasov , s. a. devyanin , and g. n. shestakov , sov . phys . * 32 * , 566 ( 1980 ) ; phys . b * 96 * , 168 ( 1980 ) ; phys . b * 108 * , 134 ( 1982 ) ; z. phys . c * 16 * , 55 ( 1982 ) ; n. n. achasov , and g. n. shestakov , z. phys . c * 41 * , 309 ( 1988 ) ; n. n. achasov , and v. v. gubin , phys . d * 56 * , 4084 ( 1997 ) . r. l. jaffe , phys . d * 15 * , 267 , 281 ( 1977 ) ; r. l. jaffe and f. e. low , phys . rev . d * 19 * , 2105 ( 1979 ) . a. sibirtsev , phys . b * 359 * , 29 ( 1995 ) ; g. q. li and c. m. ko , nucl . phys . a * 594 * , 439 ( 1995 ) ; k. tsushima , a. sibirtsev , and a. w. thomas , phys . b * 390 * , 29 ( 1997 ) ; g. fldt and c. wilkin , z. phys . a * 357 * , 241 ( 1997 ) ; r. shyam , phys . rev . d * 60 * , 055213 ( 1999 ) ; a. sibirtsev , k. tsushima , w. cassing , and a. w. thomas , _ on the @xmath1 and @xmath3 ratio from proton proton collisions _ , nucl th/0004022 , may 2000 and references therein . a. n. ivanov , n. i. troitskaya , m. nagy , and m. k. volkov , phys . b * 235 * , 331 ( 1990 ) . a. n. ivanov , m. nagy , and m. d. scadron , phys . b * 273 * , 137 ( 1991 ) . a. n. ivanov , m. nagy , and n. i. troitskaya , mod . a * 7 * , 1997 ( 1992 ) ; m. d. scadron , yad . fiz . * 56 * , 245 ( 1993 ) ( phys . at . * 56 * , 1595 ( 1993 ) ) . e. van beveren and g. rupp , phys . b * 454 * , 165 ( 1999 ) ; a. v. anisovich , v. v. anisovich , d. v. bugg , and v. a. nikonov , phys . b * 456 * , 80 ( 1999 ) . l. babukhadia , ya.a . berdnikov , a.n . ivanov , and m.d . scadron , phys . d * 62 * , 037901 ( 2000 ) . m. p. rekalo , j. arvieux , and e. tomasi gustafsson , phys . c * 56 * , 2238 ( 1997 ) . f. balestra _ et al . _ , lett . * 83 * , 1534 ( 1999 ) ; nucl . a * 663 * , 477 ( 2000 ) . n. g. kelkar , and b. k. jain , _ hyperon nucleon interactions in the @xmath149 reaction _ , nucl th/9909038 august 2000 . fig1 . the one pion exchange diagrams describing the effective lagrangian of the low energy transition p + p @xmath0 p + @xmath1 + k@xmath2 .	the cross sections for the reactions of the strange production p + p @xmath0 p + @xmath1 + k@xmath2 and p + p @xmath0 p + @xmath3 + k@xmath2 near thresholds of the final states p@xmath1k@xmath2 and p@xmath3k@xmath2 are calculated in the effective lagrangian approach . our approach is based on the dominant contribution of the one pion exchange and strong interaction of the colliding protons in the initial state . the theoretical values of the cross sections agree reasonably well with the experimental data . the polarization properties of the @xmath1 and @xmath3 hyperons are discussed . _ state technical university of st . petersburg , department of nuclear physics , + polytechnicheskaya 29 , 195251 st . petersburg , russian federation _
the kugo - ojima criterion for color confinement @xcite is based on the assumption of an exact becchi - rouet - stora - tyupin ( brst ) @xcite invariance of yang - mills theory in the manifestly covariant gauge . in the kugo - ojima analysis , the color charge @xmath4 of the global color current @xmath5 plays the very important role in relation to the brst symmetry generator , i.e. , the brst charge ( operator ) @xmath6 . in order to correctly quantize the yang - mills theory under the gauge fixing condition , however , one must avoid the gribov copies in the faddeev - popov approach for yang - mills theory @xcite . an available lagrangian formalism for this purpose is the gribov - zwanziger theory @xcite . however , it is known that the restriction to the gribov region for avoiding gribov copies inevitably leads to a breaking of the brst symmetry @xcite . if the brst symmetry is broken , the precise meaning of the kugo - ojima confinement criterion becomes unclear . therefore , the kugo - ojima analysis can not be applied to the gribov - zwanziger theory . in view of this , it is not unreasonable that the kugo - ojima criterion can not be satisfied when the restriction to the gribov region is taken into account , as shown numerically in @xcite and analytically in @xcite . consequently , the ghost propagator is no longer enhanced , in harmony with the latest lattice data @xcite and other approaches @xcite . in spite of this fact , if we could find a modified brst transformation which leaves the gribov - zwanziger action invariant , then we could apply the kugo - ojima analysis to the gribov - zwanziger theory , which opens the path to searching for the modified color confinement criterion _ a la _ kugo and ojima . in this paper , we obtain such a modified brst symmetry of the gribov - zwanziger action . the modified brst transformation satisfies the nilpotency , but involves extra gribov - parameter dependent terms , which preserve lorentz covariance and color group structure , but lacks locality . it was shown @xcite that the partition function of the @xmath7-dimensional euclidean yang - mills theory restricted to the first gribov region for avoiding gribov copies can be written in the form : @xmath8 where @xmath9 is the yang - mills action , @xmath10 is the faddeev - popov operator @xmath11 and @xmath12(x)$ ] is the horizon function given by @xmath13 here the parameter @xmath14 called the gribov parameter is determined by solving a gap equation , commonly called the horizon condition : @xmath15 the action corresponding to the partition function ( [ ym1 ] ) contains the _ non - local _ horizon term : @xmath16 the kugo - ojima color confinement criterion was obtained in the framework of the brst quantization for the usual faddeev - popov approach , corresponding to the @xmath17 case of the above gribov - zwanziger formulation . in the conventional faddeev - popov approach , the @xmath7-dimensional euclidean yang - mills theory in the covariant gauge with a gauge fixing parameter @xmath18 is defined by @xmath19 [ d\mathscr{b}][d\mathscr{c}][d\bar { \mathscr{c } } ] \exp \ { -s_{ym}^{tot } \ } , \label{ym0}\ ] ] where @xmath20 and the dot and the cross are defined as @xmath21 using the structure constant of the gauge group @xmath22 . the landau gauge corresponds to @xmath23 . we assume that the euclidean result is obtained by the wick rotation of the minkowski one . later , it has been shown @xcite that the non - local action ( [ ym1 ] ) can be put in an equivalent _ local _ form by introducing a set of complex conjugate commuting variables and anticommuting ones , which is called the gribov - zwanziger ( gz ) action . the gz theory is renormalizable to all orders of perturbation theory . hence , the restriction to the ( first ) gribov region @xmath24 makes perfect sense at the quantum level , and finite results are obtained consistent with the renormalization group . in order to fix the notation , we reproduce gribov - zwanziger ( gz ) action . the non - local horizon term can be localized through the introduction of a suitable set of additional fields . first , we prepare a set of commuting bosonic fields , @xmath25 the gaussian integration leads to @xmath26 \exp \left\ { - \int d^dx \left [ \varphi_\mu^{ca } k^{ab } \varphi_\mu^{cb } + 2i \gamma^{1/2 } gf^{abc } \mathscr{a}_\mu^b \varphi_\mu^{ac } \right ] \right\ } \nonumber\\ = & [ ( \det k)^{-1/2}]^{({\rm dim}g)d } \exp \left\ { \int d^d x \int d^dy \gamma ( gf^{abc } \mathscr{a}_\mu^b(x ) ) ( k^{-1})^{ce}(x , y ) ( gf^{ade } \mathscr{a}_\mu^d(y ) ) \right\ } , \end{aligned}\ ] ] where we have defined the faddeev - popov ( fp ) operator by @xmath27 = : - \partial_\mu ( \partial_\mu \delta^{ab } + gf^{acb } \mathscr{a}_\mu^{c } ) .\ ] ] hence , the horizon term is written as @xmath28 \ } \nonumber\\ = : & \exp \left\ { \int d^dx \int d^dy \gamma ( gf^{abc } \mathscr{a}_\mu^b(x ) ) ( k^{-1})^{ce}(x , y ) ( gf^{ade } \mathscr{a}_\mu^d(y ) ) \right\ } \nonumber\\ = & ( \det k)^{({\rm dim}g)d/2 } \int [ d\varphi ] \exp \left\ { - \int d^dx \left [ \varphi_\mu^{ca } k^{ab } \varphi_\mu^{cb } + 2i \gamma^{1/2 } gf^{abc } \mathscr{a}_\mu^b \varphi_\mu^{ac } \right ] \right\ } .\end{aligned}\ ] ] second , we incorporate the factor @xmath29 into the action as an additional term . the procedure for realizing this goal is not unique . if we prepared a set of anticommuting independent grassmannian field , @xmath30 ( @xmath31 ; @xmath32 ) , then we would have @xmath33 [ d\bar{\omega } ] \exp \left\ { - \int d^dx \left [ \bar{\omega}_\mu^{a } k^{ab } \omega_\mu^{b } \right ] \right\ } , \end{aligned}\ ] ] to achieve the correct factor @xmath29 , however , the anticommuting fields must have extra degrees of freedom represented by @xmath34 ( @xmath31 ; @xmath35 ; @xmath36 ) : @xmath37 [ d\bar{\omega } ] \exp \left\ { - \int d^dx \left [ \bar{\omega}_\mu^{ca } k^{ab } \omega_\mu^{cb } \right ] \right\ } , \end{aligned}\ ] ] in this setting , however , there is an asymmetry between the two indices @xmath38 in @xmath34 . to treat them on equal footing , therefore , we introduce @xmath39 ( @xmath31 ; @xmath35 ) to yield @xmath40 [ d\bar{\omega } ] \exp \left\ { - \int d^dx \left [ \bar{\omega}_\mu^{ca } k^{ab } \omega_\mu^{cb } \right ] \right\ } , \label{omega - integral}\end{aligned}\ ] ] where @xmath41 this treatment is preferable to consider the brst transformation later . there the mismatch of the indices between @xmath42 and @xmath43 is not desirable . in order to give rise to the extra factor @xmath29 , we need to introduce independent commuting field @xmath44 . indeed , the horizon term can have another form : @xmath28 \ } = ( \det k)^{({\rm dim}g)d } \nonumber\\ \nonumber\\ & \times \int [ d\varphi ] [ d\phi ] \exp \left\ { - \int d^dx \left [ \varphi_\mu^{ca } k^{ab } \varphi_\mu^{cb } + \phi_\mu^{ca } k^{ab } \phi_\mu^{cb } + 2i \gamma^{1/2 } gf^{abc } \mathscr{a}_\mu^b \varphi_\mu^{ac } \right ] \right\ } , \end{aligned}\ ] ] now we can use ( [ omega - integral ] ) to rewrite the horizon term in the local form : @xmath45 \ } = \int [ d\varphi ] [ d\phi ] [ d\omega ] [ d\bar{\omega } ] \exp \left\ { - s_\gamma[\mathscr{a},\varphi,\phi,\omega,\bar{\omega } ] \right\ } , \ ] ] where @xmath46 . \label{s}\ ] ] it is more convenient to introduce the complex field @xmath47 and its complex conjugate @xmath48 by @xmath49 which implies the bilinear form ( note that the fp operator is hermitian . ) @xmath50 hence , the horizon term has the local form : @xmath45 \ } = \int [ d\xi ] [ d\bar{\xi } ] [ d\omega ] [ d\bar{\omega } ] \exp \left\ { - \tilde{s}_\gamma[\mathscr{a},\xi,\bar{\xi},\omega,\bar{\omega } ] \right\ } , \ ] ] where @xmath51 . \label{aux - action}\ ] ] thus , the localized action for the gribov - zwanziger theory reads @xmath52 + \tilde{s}_\gamma [ \mathscr{a},\xi,\bar{\xi},\omega,\bar{\omega } ] \nonumber\\ = & s_{\rm ym}[\mathscr{a } ] + s_{\rm gf+fp}[\mathscr{a},\mathscr{c},\bar{\mathscr{c}},\mathscr{b } ] + \tilde{s}_\gamma [ \mathscr{a},\xi,\bar{\xi},\omega,\bar{\omega } ] , \end{aligned}\ ] ] if @xmath17 , the factor resulting from the integration over @xmath53 and @xmath54 cancel out that coming from the integration over @xmath43 and @xmath55 . therefore , for @xmath17 , the introduction of @xmath56 is trivial , which is equivalent to setting @xmath57 , and the usual brst formulation of the yang - mills - faddeev - popov theory is reproduced . our main motivation is to find out a modified brst transformation @xmath58 such that @xmath58 leaves the action @xmath59 + \tilde{s}_\gamma [ \mathscr{a},\xi,\bar{\xi},\omega,\bar{\omega}]$ ] invariant , i.e. , @xmath60 + \tilde{s}_\gamma [ \mathscr{a},\xi,\bar{\xi},\omega,\bar{\omega } ] ) = 0 , \ ] ] and @xmath58 obeys the nilpotency , i.e. , @xmath61 such a transformation could be non - local . for @xmath62 , we require @xmath63 to guarantee @xmath64 ) = 0 $ ] , i.e. , @xmath65 \mathscr{c } .\ ] ] the requirement of the nilpotency @xmath66 uniquely determines the brst transformation for the ghost field @xmath67 : @xmath68 since @xmath69 .\ ] ] this transformation for the ghost field automatically satisfies nilpotency : @xmath70 for the antighost field @xmath71 , we modify the brst transformation as @xmath72 with @xmath73 being a functional to be determined later . then , the nilpotency for the antighost field @xmath74 determines the brst transformation for the auxiliary field @xmath75 : @xmath76 which satisfies the nilpotency @xmath77 in order to determine the functional @xmath73 , we examine the gf+fp term of the form : @xmath78 .\end{aligned}\ ] ] applying the modified brst transformation , we have @xmath79 \nonumber\\ = & -i ( \mbox{\boldmath $ \delta$ } ^\prime \bar{\mathscr{c } } ) \cdot \partial_\mu \mathscr{a}_\mu -i \frac{\alpha}{2 } ( \mbox{\boldmath $ \delta$ } ^\prime \bar{\mathscr{c } } ) \cdot \mathscr{b } + i \bar{\mathscr{c } } \cdot \partial_\mu ( \mbox{\boldmath $ \delta$ } ^\prime \mathscr{a}_\mu ) + i \frac{\alpha}{2 } \bar{\mathscr{c } } \cdot ( \mbox{\boldmath $ \delta$ } ^\prime \mathscr{b } ) \nonumber\\ = & ( \mathscr{b } -i f ) \cdot \partial_\mu \mathscr{a}_\mu + \frac{\alpha}{2 } ( \mathscr{b } -i f ) \cdot \mathscr{b } + i \bar{\mathscr{c } } \cdot \partial_\mu ( d_\mu[\mathscr{a } ] \mathscr{c } ) + i \frac{\alpha}{2 } \bar{\mathscr{c } } \cdot ( i \mbox{\boldmath $ \delta$ } ^\prime f ) .\end{aligned}\ ] ] in the landau gauge , @xmath23 , especially , the gf+fp term @xmath80 $ ] reduces to @xmath81 = \mathscr{b } \cdot \partial_\mu \mathscr{a}_\mu + i \bar{\mathscr{c } } \cdot \partial_\mu ( d_\mu[\mathscr{a } ] \mathscr{c } ) + i \partial_\mu f \cdot \mathscr{a}_\mu , \label{gffpl}\end{aligned}\ ] ] where integration by parts is used in the last term . for the auxiliary fields @xmath53 and @xmath55 , we assume the modified brst transformation of the form @xmath82 @xmath83 then , the nilpotency for @xmath53 and @xmath55 , i.e. , @xmath84 and @xmath85 yield the brst transformation for @xmath43 and @xmath54 : @xmath86 @xmath87 this construction guarantees the nilpotency for @xmath43 and @xmath54 , i.e. , @xmath88 and @xmath89 . in the landau gauge , the functionals @xmath90 and @xmath91 are simultaneously determined by requiring that additional part of the action @xmath92 with @xmath93 ( [ aux - action ] ) is written in the brst - exact form @xmath94 + \tilde{s}_\gamma [ \mathscr{a},\xi,\bar{\xi},\omega,\bar{\omega } ] \nonumber\\ = & \int d^dx \ { \mathscr{b}^a \partial_\mu \mathscr{a}_\mu^a - i \bar{\mathscr{c}}^a k^{ab } \mathscr{c}^b + \bar{\xi}_\mu^{ca } k^{ab } \xi_\mu^{cb } - \bar{\omega}_\mu^{ca } k^{ab } \omega_\mu^{cb } \nonumber\\ & \quad\quad\quad\quad + i \gamma^{1/2 } gf^{abc } \mathscr{a}_\mu^b \xi_\mu^{ac } + i \gamma^{1/2 } gf^{abc } \mathscr{a}_\mu^b \bar{\xi}_\mu^{ac } \ } \\ = & \int d^dx \ { -i \mbox{\boldmath $ \delta$ } ^\prime \left [ \bar{\mathscr{c}}^a ( \partial_\mu \mathscr{a}_\mu^a ) \right ] + \mbox{\boldmath $ \delta$ } ^\prime [ \bar{\omega}_\mu^{ca } ( -\partial_\rho d^{ab}_\rho[\mathscr{a } ] ) \xi_\mu^{cb } ] \ } , \end{aligned}\ ] ] since the brst invariance of @xmath95 , i.e. , @xmath96 , is guaranteed by the nilpotency ( @xmath97 ) of the modified brst transformation . in view of this , we examine an exact form @xmath98 \nonumber\\ = & ( \mbox{\boldmath $ \delta$ } ^\prime \bar{\omega } ) k \xi - \bar{\omega } k ( \mbox{\boldmath $ \delta$ } ^\prime \xi ) - \bar{\omega } ( \mbox{\boldmath $ \delta$ } ^\prime k ) \xi \nonumber\\ = & \bar{\xi } k \xi + h k \xi - \bar{\omega } k \omega - \bar{\omega } k g - \bar{\omega } ( \mbox{\boldmath $ \delta$ } ^\prime k ) \xi \nonumber\\ = & \bar{\xi } k \xi - \bar{\omega } k \omega + h k \xi - \bar{\omega } [ kg + ( \mbox{\boldmath $ \delta$ } ^\prime k ) \xi ] , \end{aligned}\ ] ] where we have used ( [ brst - g ] ) and ( [ brst - h ] ) . therefore , we obtain the relationship : @xmath99 in this way , the modified brst transformation for @xmath71 , @xmath53 and @xmath55 is determined . a solution of the equation ( [ pbrst - f ] ) for @xmath100 , i.e. , @xmath101 is given by the line integral along an oriented path @xmath102 starting at somewhere , e.g. , infinity for definiteness and ending at @xmath103 : @xmath104 for this solution to be a path independent , namely , the solution @xmath100 is determined uniquely at a spacetime point @xmath103 in question , the line integral along any closed path must be equal to zero , @xmath105 , which is equivalent to the rotation free ( or integrability ) condition of the vector field @xmath106 , i.e. , @xmath107 . indeed , this is satisfied , since @xmath108 is of the exact one - form : @xmath109 however , we adopt another form for @xmath73 : @xmath110 where @xmath111 is the @xmath7-dimensional laplacian with the inverse @xmath112 this form for @xmath113 is more suitable than the previous one ( [ f - path ] ) , since the path does not appear in the expression and it is similar to other functionals @xmath114 and @xmath115 , which are obtained by operating the inverse of the fp operator from the left and right of ( [ pbrst - g ] ) and ( [ pbrst - h ] ) , respectively . thus , we have found the modified brst transformation @xmath116\mathscr{c}(x))^{a } , \nonumber\\ \mbox{\boldmath $ \delta$ } ^\prime \mathscr{c}^{a}(x ) = & -\frac{g}{2 } ( \mathscr{c}(x ) \times \mathscr{c}(x))^{a } , \nonumber\\ \mbox{\boldmath $ \delta$ } ^\prime \bar{\mathscr{c}}^a(x ) = & i \mathscr{b}^a(x ) + f^a(x ) , \nonumber\\ \mbox{\boldmath $ \delta$ } ^\prime \mathscr{b}^{a}(x ) = & i \mbox{\boldmath $ \delta$ } ^\prime f^{a}(x ) , \nonumber\\ \mbox{\boldmath $ \delta$ } ^\prime \xi_\mu^{ab}(x ) = & \omega_\mu^{ab}(x ) + g_\mu^{ab}(x ) , \nonumber\\ \mbox{\boldmath $ \delta$ } ^\prime \omega_\mu^{ab}(x ) = & - \mbox{\boldmath $ \delta$ } ^\prime g_\mu^{ab}(x ) , \nonumber\\ \mbox{\boldmath $ \delta$ } ^\prime \bar{\omega}_\mu^{ab}(x ) = & \bar{\xi}_\mu^{ab}(x ) + h_\mu^{ab}(x ) , \nonumber\\ \mbox{\boldmath $ \delta$ } ^\prime \bar{\xi}_\mu^{ab}(x ) = & - \mbox{\boldmath $ \delta$ } ^\prime h_\mu^{ab}(x ) , \end{aligned}\ ] ] where(y ) ( k^{-1})^{ea}(y , x ) , \ ] ] provided that the integral of the total derivative vanishes . ] @xmath117 , \\ h_\mu^{ab}(x ) = & \int d^dy i \gamma^{1/2 } gf^{bce } \mathscr{a}_\mu^c(y ) ( k^{-1})^{ea}(y , x ) .\end{aligned}\ ] ] this modified brst transformation is nilpotent and leaves the gribov - zwanziger action invariant . @xmath118 \nonumber\\ = & s_{\rm ym}[\mathscr{a } ] + s_{\rm gf+fp}[\mathscr{a},\mathscr{c},\bar{\mathscr{c}},\mathscr{b } ] + \tilde{s}_\gamma [ \mathscr{a},\xi,\bar{\xi},\omega,\bar{\omega } ] \nonumber\\ = & s_{\rm ym}[\mathscr{a } ] + \int d^dx \ { -i \mbox{\boldmath $ \delta$ } ^\prime \left [ \bar{\mathscr{c}}^a ( \partial_\mu \mathscr{a}_\mu^a ) \right ] + \mbox{\boldmath $ \delta$ } ^\prime [ \bar{\omega}_\mu^{ca } ( -\partial_\rho d^{ab}_\rho[\mathscr{a } ] ) \xi_\mu^{cb } ] \ } .\end{aligned}\ ] ] among the fields , @xmath119 , only the fields @xmath120 appear in this construction . hence , the brst transformation for the fields @xmath121 do not affect the action . they are determined by the nilpotency requirement . in the limit @xmath122 , @xmath123 , @xmath124 @xmath116\mathscr{c}(x))^{a } , \nonumber\\ \mbox{\boldmath $ \delta$ } ^\prime \mathscr{c}^{a}(x ) = & -\frac{g}{2 } ( \mathscr{c}(x ) \times \mathscr{c}(x))^{a } , \nonumber\\ \mbox{\boldmath $ \delta$ } ^\prime \bar{\mathscr{c}}^a(x ) = & i \mathscr{b}^a(x ) , \nonumber\\ \mbox{\boldmath $ \delta$ } ^\prime \mathscr{b}^{a}(x ) = & 0 , \nonumber\\ \mbox{\boldmath $ \delta$ } ^\prime \xi_\mu^{ab}(x ) = & \omega_\mu^{ab}(x ) + g_\mu^{ab}(x ) , \nonumber\\ \mbox{\boldmath $ \delta$ } ^\prime \omega_\mu^{ab}(x ) = & - \mbox{\boldmath $ \delta$ } ^\prime g_\mu^{ab}(x ) , \nonumber\\ \mbox{\boldmath $ \delta$ } ^\prime \bar{\omega}_\mu^{ab}(x ) = & \bar{\xi}_\mu^{ab}(x ) , \nonumber\\ \mbox{\boldmath $ \delta$ } ^\prime \bar{\xi}_\mu^{ab}(x ) = & 0 .\end{aligned}\ ] ] note that @xmath125 does not vanish even in the limit @xmath122 and the modified brst transformation @xmath126 has the part @xmath125 involving the yang - mills field and the ghost field . even in the limit , therefore , the horizon term is not decoupled from the usual yang - mills - faddeev - popov theory . this issue is cured by redefining the auxiliary field @xmath127 , i.e. , shifting it by @xmath128 @xcite : @xmath129 then the modified brst transformation is simplified @xmath116\mathscr{c}(x))^{a } , \\ \mbox{\boldmath $ \delta$ } ^\prime \mathscr{c}^{a}(x ) = & -\frac{g}{2 } ( \mathscr{c}(x ) \times \mathscr{c}(x))^{a } , \\ \mbox{\boldmath $ \delta$ } ^\prime \bar{\mathscr{c}}^a(x ) = & i \mathscr{b}^a(x ) + f^a(x ) , \\ \mbox{\boldmath $ \delta$ } ^\prime \mathscr{b}^{a}(x ) = & i \mbox{\boldmath $ \delta$ } ^\prime f^{a}(x ) , \\ \mbox{\boldmath $ \delta$ } ^\prime \xi_\mu^{ab}(x ) = & \omega^\prime{}_\mu^{ab}(x ) , \\ \mbox{\boldmath $ \delta$ } ^\prime \omega^\prime{}_\mu^{ab}(x ) = & 0 , \\ \mbox{\boldmath $ \delta$ } ^\prime \bar{\omega}_\mu^{ab}(x ) = & \bar{\xi}_\mu^{ab}(x ) + h_\mu^{ab}(x ) , \\ \mbox{\boldmath $ \delta$ } ^\prime \bar{\xi}_\mu^{ab}(x ) = & - \mbox{\boldmath $ \delta$ } ^\prime h_\mu^{ab}(x ) , \end{aligned}\ ] ] where @xmath130 here note that the brst transformation is closed , although it is quite non - local , as can be seen as follows . @xmath131 .\end{aligned}\ ] ] it turns out that the nilpotency is satisfied for all the fields . in fact , the nilpotency reduces to that for @xmath132 : @xmath133 = 0 .\end{aligned}\ ] ] in the limit @xmath122 , the horizon term is decoupled from the usual yang - mills - faddeev - popov theory : @xmath116\mathscr{c}(x))^{a } , \nonumber\\ \mbox{\boldmath $ \delta$ } ^\prime \mathscr{c}^{a}(x ) = & -\frac{g}{2 } ( \mathscr{c}(x ) \times \mathscr{c}(x))^{a } , \nonumber\\ \mbox{\boldmath $ \delta$ } ^\prime \bar{\mathscr{c}}^a(x ) = & i \mathscr{b}^a(x ) , \nonumber\\ \mbox{\boldmath $ \delta$ } ^\prime \mathscr{b}^{a}(x ) = & 0 , \nonumber\\ \mbox{\boldmath $ \delta$ } ^\prime \xi_\mu^{ab}(x ) = & \omega^\prime{}_\mu^{ab}(x ) , \nonumber\\ \mbox{\boldmath $ \delta$ } ^\prime \omega^\prime{}_\mu^{ab}(x ) = & 0 , \nonumber\\ \mbox{\boldmath $ \delta$ } ^\prime \bar{\omega}_\mu^{ab}(x ) = & \bar{\xi}_\mu^{ab}(x ) , \nonumber\\ \mbox{\boldmath $ \delta$ } ^\prime \bar{\xi}_\mu^{ab}(x ) = & 0 .\end{aligned}\ ] ] for this choice of the field variables , the action changes @xcite @xmath94 + \tilde{s}_\gamma [ \mathscr{a},\xi,\bar{\xi},\omega^\prime,\bar{\omega } ] \nonumber\\ = & \int d^dx \ { -i \mbox{\boldmath $ \delta$ } ^\prime \left [ \bar{\mathscr{c}}^a ( \partial_\mu \mathscr{a}_\mu^a ) \right ] + \mbox{\boldmath $ \delta$ } ^\prime [ \bar{\omega}_\mu^{ca } ( -\partial_\rho d^{ab}_\rho[\mathscr{a } ] ) \xi_\mu^{cb } ] \ } \nonumber\\ = & \int d^dx \ { \mathscr{b}^a \partial_\mu \mathscr{a}_\mu^a + i \bar{\mathscr{c}}^a \partial_\mu ( d_\mu^{ab}[\mathscr{a } ] \mathscr{c}^b ) \nonumber\\ & \quad\quad - \bar{\xi}_\mu^{ca } \partial_\rho d^{ab}_\rho[\mathscr{a } ] \xi_\mu^{cb } + \bar{\omega}_\mu^{ca } \partial_\rho d^{ab}_\rho[\mathscr{a } ] \omega^\prime{}_\mu^{cb } - \bar{\omega}_\mu^{ca } \partial_\rho [ gf^{cde } ( d_\rho \mathscr{c})^{d } \xi_\mu^{ea } ] \nonumber\\ & \quad\quad + i \gamma^{1/2 } gf^{abc } \mathscr{a}_\mu^b \xi_\mu^{ac } + i \gamma^{1/2 } gf^{abc } \mathscr{a}_\mu^b \bar{\xi}_\mu^{ac } \ } .\end{aligned}\ ] ] the jacobian for the non - local change of variable from @xmath127 to @xmath134 is field - independent and hence the change of variable does not generate additional factor in the integration measure or the action . it is possible to further simplify the brst transformation by the redefinition of the field @xmath135 , @xmath136 . on the other hand , the integration measure may be complicated due to the jacobian associated with the non - local change of variables . consequently , such an option of the gribov - zwanziger theory would contain non - local terms again . in order to demonstrate the usefulness of the modified brst symmetry , we show that the horizon condition ( the gap equation determining the gribov parameter ) is consistent with the brst invariance of the vacuum . we define the generator @xmath137 , i.e. , the brst charge of the modified brst transformation by @xmath138_{\mp}\ ] ] where @xmath139_{\mp}$ ] denotes the commutator or anticommutator depending on the ghost number of @xmath140 being even or odd , respectively . then we have an slavnov - taylor identity : @xmath141_{+ } \| 0 \rangle \nonumber\\ = & \langle 0\| \mbox{\boldmath $ \delta$ } ^\prime [ gf^{abc}\mathscr{a}_\mu^a(x ) \bar{\omega}^{bc}_\mu(x ) ] \| 0 \rangle \nonumber\\ = & \langle 0\| [ gf^{abc } \mbox{\boldmath $ \delta$ } ^\prime \mathscr{a}_\mu^a(x ) \bar{\omega}^{bc}_\mu(x ) ] \| 0 \rangle + \langle 0\| [ gf^{abc}\mathscr{a}_\mu^a(x ) \mbox{\boldmath $ \delta$ } ^\prime \bar{\omega}^{bc}_\mu(x ) ] \| 0 \rangle \nonumber\\ = & \langle 0\| [ gf^{abc } ( d_\mu \mathscr{c})^a(x ) \bar{\omega}^{bc}_\mu(x ) ] \| 0 \rangle + \langle 0\| [ gf^{abc}\mathscr{a}_\mu^a(x ) \bar{\xi}^{bc}_\mu(x ) ] \| 0 \rangle \nonumber\\ & + \langle 0\| [ gf^{cab}\mathscr{a}_\mu^a(x ) \int d^d y i \gamma^{1/2 } gf^{cef } \mathscr{a}_\mu^{e}(y ) ( k^{-1})^{fb}(y , x ) ] \| 0 \rangle \nonumber\\ = & \langle 0\| [ gf^{abc}\mathscr{a}_\mu^a(x ) \bar{\xi}^{bc}_\mu(x ) ] \| 0 \rangle + i \gamma^{1/2 } \langle 0\| h(x ) \| 0 \rangle , \end{aligned}\ ] ] where we have used the fact that @xmath142 \| 0 \rangle $ ] vanishes due to the discrete symmetry of the gribov - zwanziger action , @xmath143 the original horizon condition in the non - local zwanziger theory @xcite is given by @xmath144 . therefore , the horizon condition in the gribov - zwanziger theory @xcite @xmath145 \| 0 \rangle = -i\gamma^{1/2 } ( n^2 - 1)d , \ ] ] follows from the brst invariance of the vacuum : @xmath146 which yields @xmath147_{+ } \| 0 \rangle = 0 .\ ] ] in other words , if the vacuum in the gribov - zwanziger theory obeys the subsidiary condition @xmath148 , the horizon condition is automatically satisfied . finally , it should be remarked that the gribov - zwanziger theory can be applied only to the landau gauge @xmath23 in the manifestly covariant gauge , which restricts the gauge field to the hypersurface @xmath149 , while for @xmath150 the gauge field is allowed to be anywhere in the configuration space of the field . the explicit form of the modified brst transformation obtained in this paper is meaningful only in the landau gauge . if one considers the other gauge fixing , e.g. , the coulomb gauge or the maximal abelian gauge in which the gribov copies are known to exist , the corresponding gribov - zwanziger theory has the different action depending on the gauge fixing , and hence the modified `` brst '' transformation obtained in the same sense as that in this paper may differ gauge to gauge . on the contrary , the usual brst transformation is independent of the gauge fixing condition to be chosen . in this sense , the modified brst transformation in this paper should be understood as a quantum symmetry for the respective theory , which is non - local and nilpotent . the modified confinement criterion remains to be clarified , once the nilpotent brst symmetry has been found . [ note added ] in preparing this paper , we have found that a paper by sorella @xcite was submitted to the archive . in this paper , he obtained a modified brst transformation leaving the gribov - zwanziger action invariant . as expected , the transformation is non - local in the sense that the inverse of the fp operator is involved in it . however , the modified brst transformations lack nilpotency . our brst transformation satisfies nilpotency by construction . the usefulness of the non - local brst transformation is demonstrated in his paper . the author would like to thank high energy physics theory group and theoretical hadron physics group in the university of tokyo , especially , prof . tetsuo hatsuda for kind hospitality extended to him on sabbatical leave . he would like to thank dr . david dudal and dr . nele vandersickel for sending valuable comments and questions on the first version of this paper . this work is financially supported by grant - in - aid for scientific research ( c ) 21540256 from japan society for the promotion of science ( jsps ) . d. dudal , j.a . gracey , s.p . sorella , n. vandersickel and h. verschelde , arxiv:0806.4348[hep - th ] , phys . d78 , 065047 ( 2008 ) . + d. dudal , s.p . sorella , n. vandersickel and h. verschelde , arxiv:0711.4496[hep - th ] , phys . rev . d77 , 071501(r ) ( 2008 ) . s. furui and h. nakajima , hep - lat/0609024 , brazilian journal of physics * 37 * , 186192 ( 2007 ) . + s. furui and h. nakajima , hep - lat/0503029 , few - body systems * 40 * , 101128 ( 2006 ) . + s. furui and h. nakajima , hep - lat/0305010 , phys . rev . d69 , 074505 ( 2004 ) . bogolubsky , e .- ilgenfritz and m. mller - preussker and a. sternbeck , arxiv:0901.0736[hep - lat ] . + a. cucchieri and t. mendes , arxiv:0804.2371[hep - lat ] , phys . rev . d78 , 094503 ( 2008 ) , + a. cucchieri and t. mendes , arxiv:0812.3261[hep - lat ] . + a. cucchieri and t. mendes , arxiv:0904.4033[hep - lat ] . + o. oliveira and p.j . silva , arxiv:0809.0258[hep - lat ] . + a. sternbeck and l. von smekal , arxiv:0811.4300[hep - lat ] . + a. sternbeck , l. von smekal , d.b . leinweber , a.g . williams , arxiv:0710.1982[hep - lat ] , pos lat2007 , 340 ( 2007 ) . boucaud , j.p . leroy , a. le yaouanc , j. micheli , o. pene and j. rodriguez - quintero , hep - ph/0803.2161 , jhep 06 , 099 ( 2008 ) . + ph . boucaud , j.p . leroy , a. le yaouanc , j. micheli , o. pene and j. rodriguez - quintero , arxiv:0801.2721[hep - ph ] , jhep 06 , 012 ( 2008 ) .	we have found a nilpotent quantum symmetry of yang - mills theory restricted to the gribov region . in fact , we give a set of transformations for the filds a la brst that leaves the gribov - zwanziger action invariant and obeys the nilpotency , although the usual brst symmetry is broken due to the non - vanishing gribov parameter representing the presence of the gribov horizon in the configuration space of the yang - mills field . we show that the horizon condition ( the gap equation determining the gribov parameter ) is consistent with the modified brst invariance of the vacuum . ' '' '' ' '' '' 1.0 cm * * 0.3 cm * the nilpotent `` brst '' symmetry * 0.3 cm * for the gribov - zwanziger theory * 0.3 cm 0.5 cm * kei - ichi kondo,@xmath0 * 0.5 cm _ @xmath1department of physics , university of tokyo , tokyo 113 - 0033 , japan _ _ @xmath2department of physics , chiba university , chiba 263 - 8522 , japan _ key words : brst symmetry , gribov - zwanziger , gribov horizon , gribov region , 0.5 cm pacs : 12.38.aw , 12.38.lg ' '' '' 0.1 cm @xmath3 on sabbatical leave of absence from chiba university . e - mail : [email protected] 0.5 cm
the amplified spontaneous emission ( ase ) noise from optical amplifiers , e.g. , erbium doped fiber amplifiers ( edfas ) , is one of the fundamental reasons for the bit - error - rate ( ber ) in an optical fiber communication ( ofc ) system . for an ofc system with non - negligible kerr nonlinearity , the ase impact evaluation is complicated due to the nonlinear interaction between signal and ase noise . in the case of signal power much stronger than the noise power , the noise - noise beating is relatively small so that the noise field in the fiber can be approximately described by linearized noise equation ( lne ) , which was proposed separately in ref . @xcite and refs . @xcite-@xcite . noise propagator is a matrix used to show noise field propagation in the fiber . in the case of linear perturbation , it is independent of the specific noise realizations , which makes it possible to calculate the moment - generating function ( mgf ) of the filtered photoelectric current at the receiver @xcite-@xcite . mgf method is an approach making use of mgf to evaluate the noise impact on the system performance . now , it is well known that this method is accurate for various linear ofc systems . for nonlinear ofc systems , the mgf method can also be applied , provided their noise fields obey lne . one can see this from doob s theorem , which means that , in a linearizable system driven by gaussian - distributed noise , each of the independent random variable keeps gaussian ( cf . p. 35 of ref . thus the mgf of the received photoelectric current can be calculated as those for linear ofc systems . the common form of lne @xcite-@xcite is based on the continuous wave ( cw ) assumption , i.e. , the noise - free signal in the lne is artificially simplified as a cw wave . as a result , a semi - analytical form of noise propagator and the noise power spectral density ( psd ) , so called parametric gain ( pg ) , can be obtained . the drawback of this simplification is that the noise - free signal in this lne neglects chromatic dispersion ( cd ) effect . as a result , the couplings between noise components ( in frequency domain ) can not be taken into account . a lne beyond cw , named accurate lne in this work , was first proposed and discussed in ref . dynamically taking into account the local cd and kerr nonlinearity along the fiber , this lne provides accurate noise information , with its computational cost being much higher than the cw approach . for example , given a noise - free signal obtained from nonlinear schr@xmath0dinger equation ( nlse ) , the computation required to update the accurate lne has cubic complexity in the number of fourier components @xcite . to reduce the computational complexity , covariance matrix method ( cmm ) was proposed in ref . @xcite , where the noise covariance matrix was obtained by processing large noise realizations . in refs . @xcite , the computational cost of cmm was further reduced by a deterministic approach using perturbation solution . since the raw covariance matrix obtained from nlse via monte carlo noise realizations @xcite or perturbation solution @xcite may contain nonlinear noise contribution , it needs to be separated from the nonlinearity - induced phase and timing jitter . thus , the obtained pdfs of the receiver voltage agrees well with monte carlo simulation @xcite . with the help of linear perturbation , the noise covariance matrix can also be solved from its ordinary differential equation ( ode ) proposed by @xcite . the covariance matrix obtained by solving such linear ode does not need jitter separation , although this ode is more complicated than the accurate lne @xcite . so far there is little comparison between the approaches of ref . @xcite and refs . @xcite . in this work , we simplify the cmm by showing that the noise propagator matrix can be obtained directly from the accurate lne . therefore , there is no nonlinearity - induced jitter . to effectively reduce the computational complexity in updating the lne , one can decompose the kerr effect related matrix into a symmetric and an antisymmetric matrices [ cf . the discussion after eq . ( [ hat_l_nu_mu ] ) in appendix a ] . making use of the fourth - order runge - kutta in the interaction picture ( rk4ip ) method @xcite , the accurate lne can be solved with large step size , as detailed in sec . [ noise_propagator ] . we evaluate the impacts of noise propagator on moment - generating function ( mgf ) and ber in sec . [ mgf_1 ] . the accuracy of this new approach depends on how far the linearized noise deviates from the actual noise . to numerically verify this new approach , we consider the bers in a 20-span dpsk system with nonlinear phase of @xmath1 @xcite in sec . [ n_verification ] . our ber calculations agree well with the published cmm results . in sec . [ n_verification ] , we also simulate the experiments of the multi - span dpsk systems discussed in ref . @xcite and show that , to fit the experimental data , one needs to take into account the nonlinearity induced phase difference between noise and noise - free signal , which will affect the signal - noise beating significantly . in an ofc system amplified by edfas , the noise propagator is a fundamental matrix that determines the noise impacts on mgf and ber . in this section , we show that the noise propagator matrix in a fiber of length @xmath2 can be obtained from the accurate lne given by eq . ( [ ns_edfa ] ) . for a multi - span ofc system , one needs to introduce an equivalent noise propagator which can be obtained from pg . the noise propagator in a fiber of length @xmath2 can be obtained by extending the rk4ip in refs . @xcite to the accurate lne given by eq . ( [ ns_edfa ] ) in appendix a , where the linear operator @xmath3 is associated with cd effect , whereas the nonlinear operator @xmath4 is caused by kerr nonlinearity . by introducing @xmath5 and @xmath6 , ( [ ns_edfa ] ) in the interaction picture ( ip ) has the form -4 mm @xmath7 -1 mm taking @xmath8 with step size @xmath9 and denoting @xmath10 , @xmath11 , @xmath12 , @xmath13 , and @xmath14 , one can use rk4ip @xcite to to solve eq . ( [ ns_edfa_ip ] ) with -5 mm @xmath15 \nonumber\\ & & \tilde{a}_n^i = e^{\hat{l}h/2}\tilde{a}_n \nonumber\\ & & k_1=\hat{n}_n^i\tilde{a}_n^i = e^{\hat{l}h/2}\hat{n}(z_n)\tilde{a}_n\equiv \hat{k}_1\tilde{a}_n \nonumber \\ & & k_2=\hat{n}^i_{n+1/2}[\tilde{a}_n^i+\frac{hk_1}{2}]=\hat{n}(z_n+h/2)[e^{\hat{l}h/2}+\frac{h\hat{k}_1}{2}]\tilde{a}_n\equiv \hat{k}_2\tilde{a}_n \nonumber \\ & & k_3=\hat{n}^i_{n+1/2}[\tilde{a}_n^i+\frac{hk_2}{2}]=\hat{n}(z_n+h/2)[e^{\hat{l}h/2}+\frac{h\hat{k}_2}{2}]\tilde{a}_n\equiv \hat{k}_3\tilde{a}_n \nonumber \\ & & k_4=\hat{n}_{n+1}^i[\tilde{a}_n^i+hk_3]=e^{-\hat{l}h/2}\hat{n}(z_{n+1})e^{\hat{l}h/2}[e^{\hat{l}h/2}+h\hat{k}_3]\tilde{a}_n \equiv \hat{k}_4\tilde{a}_n \label{h_rk4ip}\end{aligned}\ ] ] -3 mm or -3 mm @xmath16 + \frac{h}{6}\hat{n}(z_{n+1})e^{\hat{l}h/2}(e^{\hat{l}h/2}+h\hat{k}_3)\bigg)\tilde{a}_n , \label{h_rk4ip_2}\ ] ] which means the noise propagator for the fiber of length @xmath9 can be calculated as @xmath17 + \frac{h}{6}\hat{n}(z_{n+1})e^{\hat{l}h/2}\bigg[e^{\hat{l}h/2}+h\hat{k}_3\bigg ] \label{h_h}\ ] ] for the fiber of length @xmath2 , the noise propagator has the form @xmath18 note that the rk4ip used here is different from the rk4ip in ref . @xcite , where what to be solved was the noise - free signal ( a 1d matrix ) , while here what we want is the noise propagator ( a 2d matrix ) . the computational complexity for this 2d matrix is @xmath19 , due to that each @xmath20 ( @xmath21 ) in eq . ( [ h_rk4ip ] ) needs one dense matrix multiplication . here @xmath22 is the number of fourier components used for signal representation , as mentioned between eqs . ( [ hat_l_nu_mu ] ) and ( [ auto_noise_f_w ] ) . as discussed in appendix a , the ase from an edfa can be modeled as additive white gaussian noise ( awgn ) with its variance given by eq . ( [ auto_noise_f3 ] ) . given the ase injected at the input of a fiber and the noise propagator obtained from eqs . ( [ h_rk4ip ] ) , ( [ h_h ] ) , and ( [ p_noise_l ] ) , the noise psd ( or pg ) at the output of a fiber of length @xmath2 can be written as @xcite @xmath23 where @xmath24 is a unit matrix and eq . ( [ auto_noise_f3 ] ) has been used . in eq . ( [ pg_1 ] ) , @xmath25 is the transpose of @xmath26 . for a @xmath27-span system consisting of @xmath28 edfas ( cf . [ fig_link ] ) , its pg has the form @xmath29 where @xmath30 and @xmath31 is given by eqs . ( [ psd_ase ] ) and ( [ auto_noise_f3 ] ) . in eq . ( [ pg_k ] ) , the real symmetric matrix @xmath32 is positive definite . it can be factorized as @xmath33 where the equivalent noise propagator @xmath34 can be obtained either by using cholesky decomposition or symmetric ( square root ) decomposition @xcite . the latter yields @xmath35 . with the noise propagator matrix , one can evaluate the ber in the ofc system by calculating the mgf of the electrically filtered current @xmath36 expressed using karhunen - lo@xmath37ve series expansion ( klse ) . due to the noise in the ofc system , the received ( or filtered photoelectric ) current fluctuates around its expectation value . the mgf of such current is a useful form of its probability distribution . to get a simple form of mgf , the received current needs to be expressed using klse . for a nonlinear ofc system with its noise being linearizable , the klse form of the received current can be formulated as eq . ( [ y_receiv2 ] ) and the related formulas in appendix b. all the parameters for the nonlinear case are generalized from those for the linear case , which has been well discussed in refs . @xcite and other publications . in eqs . ( [ tilde_s_n ] ) and ( [ y_receiv2 ] ) , the dirac notation @xmath38 is used to represent the normalized noise ( in re - im form ) expressed by karhunen - lo@xmath37ve bases ( in re - im form ) . averaging @xmath39 ( @xmath40 ) with formula @xcite @xmath41=\!\ ! \int_{-\infty}^{\infty}\!\!\frac{dc}{\sqrt{2\pi\sigma^2}}\exp(-\frac { c^2}{2\sigma^2})\exp [ s(\tilde{\lambda } c^2 + 2c\tilde{b } ) ] = \frac { \exp[\frac{2\sigma^2 s^2 \tilde{b}^2}{1 - 2\sigma^2 s \tilde{\lambda } } ] } { \sqrt{1 - 2\sigma^2s\tilde{\lambda } } } , \label{gaus_int}\end{aligned}\ ] ] the mgf of the filtered current , denoted as @xmath42 here , can be written as @xmath43 \big ] = \exp [ si_{\mathrm{ss}}(t_{\mathrm{s } } ) ] \prod_{i=1}^{4m_n+2}\frac{\exp[\frac{2\sigma^2 s^2 \tilde{b}_i^2(t_{\mathrm{s}})}{1-s\beta_i } ] } { ( 1-s\beta_i)^{\xi}},\;\;\;\;(\beta_i=\!2\sigma^2\tilde{\lambda}_i ) \label{mgf}\end{aligned}\ ] ] where @xmath36 is the filtered photoelectric current at time @xmath44 . it consists of signal - signal beating ( @xmath45 ) , noise - noise beating ( @xmath46 ) , and signal - noise beating ( @xmath47 ) , which are detailed in eqs . ( [ y_d ] ) and ( [ y_receiv2 ] ) respectively . in eq . ( [ mgf ] ) , @xmath48 is the @xmath49th component of @xmath50 in ( [ y_receiv2 ] ) , while @xmath51 is the power of @xmath49th component of the noise in karhunen - lo@xmath37ve presentation . in eq . ( [ mgf ] ) , @xmath31 is given by eq . ( [ auto_noise_f3 ] ) . in this work , we take @xmath52 for polarized noise . with the help of eq . ( [ mgf ] ) as well as eqs . ( 7 ) and ( 8) in ref . @xcite , the ber can be calculated . for an optical system with ase power being much larger than other noise sources , the osnr with reference bandwidth @xmath53 ( 0.1 nm ) can be calculated as @xmath54 in eq . ( [ osnr ] ) , @xmath55 is the time - averaged ( noise free ) signal power , while @xmath56 is the noise power within @xmath53 . to obtain @xmath55 and @xmath56 , one needs to notice that the measurement bandwidth @xmath57 [ e.g. , the bandwidth of the transfer function of an optical spectrum analyzer ( osa ) ] may not be the same as @xmath53 . thus the @xmath55 in eq . ( [ osnr ] ) becomes the power of the signal filtered by @xmath57 , while @xmath56 becomes the ase filtered by @xmath57 and weighted by a factor @xmath58 @xcite . in a linear optical system , the ase noise along the fiber can be treated as awgn . thus , for the system of fig . [ fig_link ] , its osnr can be simply calculated as @xmath59b_m}\times\frac{b_m}{b_r},% ; \;\;\;\;\;\ ; \big{(}p_{ase}(b_m)=[gn_{in}+n_0(k+1)]b_m\big { ) } , \label{osnr_l}\ ] ] where @xmath60 is the ase power within @xmath57 and @xmath61 is given by eq . ( [ psd_ase ] ) . in eq.([osnr_l ] ) , the filter ( @xmath57 ) effect on the ase has been neglected . in a nonlinear optical system , the ase noise `` amplified by '' pg can not be treated as a white noise . similar to the noise - noise beating given in eq . ( [ y_receiv2 ] ) , the measured ase power will only relate with the self - beating terms of the noise , which yields @xmath62 here @xmath63 is the low - pass transfer function of the bandpass filter ( bandwidth @xmath57 ) . in eq . ( [ osnr_nl_1 ] ) , @xmath31 and @xmath64 are given by eq . ( [ auto_noise_f3 ] ) and eq . ( [ pg_k ] ) , respectively . in the case of traditional osa - based out - of - band osnr monitoring , the ase power can be interpolated using @xmath65}\sigma^2}{2 } , % \label{osnr_nl_2 } \label{ase_delta}\ ] ] where @xmath66 is the filter function centered at @xmath67 . when @xmath68 , eq . ( [ ase_delta ] ) returns to the ase power in eq . ( [ osnr_nl_1 ] ) , where @xmath69 . -0 mm cc + -2 mm to show that the new approach to get the noise propagator is numerically applicable , we will compare our rk4ip results with the cmm results given by ref . @xcite and with the experimental data given by ref . @xcite . both consider systems with @xmath7020 gb / s , using rz-50@xmath71 dpsk modulation . in the receiver , the optical filter is gaussian type , while the electric filter is the fifth - order bessel type . in the following calculations , we set @xmath72 by changing @xmath73 in eq . ( [ lmt_0 ] ) . this means , given the noise propagator matrix , the computational cost for ber is much higher than that in the linear case . for ber calculations , the length of the de bruijn sequence is @xmath74 @xcite . based on the relation between the rk4ip step and the fiber dispersion length ( @xmath75 ) or nonlinear length ( @xmath76 ) discussed in ref . @xcite as well as the detailed fiber parameters in the following discussion , we let the rk4ip step for the transmission fiber ( @xmath77 ) and that for the dcf fiber ( @xmath78 ) be related with @xmath79 . the value of @xmath77 and the required computational time will be detailed below . we consider the 20-span dpsk system discussed in ref . @xcite , where bers using the cmm and the ( improved ) cw approaches were plotted against the received osnr in the fig . 8 of ref . @xcite . in fact this system is basically the same as the one shown in fig . [ fig_link ] in appendix a , provided that one removes the pre- and postcompensating fibers and their amplifiers in the fig . 2 of ref . @xcite and removes the first edfa and @xmath80 in our fig . [ fig_link ] . thus the first term of @xmath64 [ in eq . ( [ pg_k ] ) ] needs to be ignored . like refs . @xcite , where osnr was calculated in the absence of pg , we obtain osnr from eq . ( [ osnr_l ] ) with @xmath81 and @xmath28 being replaced by @xmath27 . according to ref . @xcite , we change osnr by changing the @xmath82 in eq . ( [ psd_ase ] ) . as plotted in fig . [ fig_link ] , each span contains a transmission fiber followed by a dispersion - compensating fiber ( dcf ) . the transmission fiber is @xmath83100 km long with its cd parameter @xmath84 ps / nm / km . each span is fully compensated . the nonlinear phase accumulated in the fiber , defined as @xmath85 with @xmath86 being the time averaged signal power ( at the input of the fiber ) , is @xmath87 . the bandwidth of the optical ( electrical ) filter in the receiver is @xmath88 ( @xmath89 ) , respectively . to let our results be reproducible , we provide , as detailed as possible , other related parameters below . the dcf in each span is 8 km long with @xmath90 ps / nm / km . transmission fiber and dcf are assumed to have same fiber loss ( @xmath910.2 db / km ) and same nonlinear coefficient ( @xmath922.0 /w / km ) . the edfa in each span is used to compensate the total loss in the fiber of @xmath93 km . therefore the signal power at the input of each span ( @xmath86 ) keeps constant . ignoring the nonlinear phase contribution of dcf @xcite , we set @xmath86=0.7307 mw , obtained from @xmath94 with @xmath95 and @xmath96 km . the osnr is obtained using eq . ( [ osnr_l ] ) with @xmath97 . as shown in fig . [ fig_compare_paolo_2 ] , the curve using the proposed rk4ip approach ( dash - dotted ) agrees very well with the cmm curve ( solid ) given by ref . @xcite . in fig . [ fig_compare_paolo_2 ] , the curves using improved cw approach @xcite with cw power being transmitted peak power ( dashed ) and average power ( dotted ) are plotted for comparison . for each calculated point in fig . [ fig_compare_paolo_2 ] , the cpu time for the noise propagator calculation is @xmath981.8 hr with rk4ip step for transmission fiber ( @xmath99100 km long ) being @xmath1003.5 km . the cpu time for each ber calculation is @xmath98 0.5 hr . in fact , @xmath77 ranged within @xmath101 km yields almost the same curve . -0 mm cc + -2 mm -0 mm cc + -2 mm the optical system discussed in ref . @xcite can be modelled by fig . [ fig_link ] , except that each edfa should be replaced by an edfa followed by an optical filter ( gaussian ) @xmath102 with bandwidth @xmath1035 nm . also , the input noise @xmath80 needs to be filtered by an optical filter @xmath104 ( gaussian , @xmath1053 nm ) . as a result , in eq . ( [ pg_k ] ) , the noise propagator @xmath106 ( @xmath107 ) should be replaced with @xmath108 , @xmath109 with @xmath110 , and @xmath111 with @xmath112 . since we only consider the curves plotted in fig . 7 ( b ) of ref . @xcite , each fiber ( @xmath2 km long ) in fig . [ fig_link ] contains a smf ( 42 km long ) followed by a dcf ( 7 km long ) . in our calculation , all the related fiber parameters are same as those given in table 1 of ref . @xcite . in the receiver , the bandwidth of the optical filter is @xmath113 , while the bandwidth of the electrical filter is @xmath114 . we first consider the back - to - back case . similar to ref . @xcite , we modify the 20gb / s rz-50@xmath71 signal at the transmitter by comparing its calculated spectrum @xcite and its measured spectrum@xcite . due to that the input noise @xmath80 is filtered by @xmath104 , the osnr is calculated using eq . ( [ osnr_nl_1 ] ) with @xmath115 , yielding the back - to - back rk4ip curve shown in fig . [ fig_compare_coelho_1 ] . for the 5- , 10- , 25-span systems , their accumulated nonlinear phases are calculated according to eq . ( 48 ) of ref . @xcite . because of the spectral modification of the input signal , the optical power at the input of each span @xmath116 is smaller than @xmath117 , where @xmath118 is the energy per bit before the spectral modification . for example , to get nonlinear phase @xmath119 for the 25-span system , the fiber input power @xmath116 should be 1.516 mw , which means @xmath1200.127 mw or @xmath121 mw ( @xmath122 ) . different from the dpsk receiver shown in fig.[fig_link ] , where the delay is @xmath123 ps , the delay in the receiver of ref . @xcite was @xmath124=(24.84 ghz)@xmath125=40.26 ps . thus , the dpsk phase factors given in eq . ( [ r_dpsk ] ) should be modified as @xmath126 with @xmath127 , @xmath128 . in eq . ( [ r_dpsk_2 ] ) , @xmath129 is introduced as @xmath130 where @xmath131 , given by eq . ( [ d_phi_1 ] ) in appendix c , is the nonlinear phase difference between noise and noise - free signal . as shown in fig . [ fig_compare_coelho_1 ] , all rk4ip curves ( @xmath132 ) agree very well with the experiment results . the ase power is calculated using eq . ( [ ase_delta ] ) with @xmath133 . in eq . ( [ gm_shift ] ) , @xmath134 is a calibration constant that basically shifts the rk4ip curves in the osnr direction , while @xmath131 determines the slope of the rk4ip curves . to show this , we plot in fig . [ fig_compare_coelho_2 ] the rk4ip results for the 25-span system with @xmath135 and @xmath136 . also , we consider the rk4ip curves using eq . ( [ osnr_nl_1 ] ) to calculate ase power . our results for the 5-span , 10-span , and 25-span systems confirm that there is almost no difference between the curve using eq . ( [ ase_delta ] ) with @xmath132 and the curve using eq . ( [ osnr_nl_1 ] ) with @xmath137 . in eq . ( [ gm_shift ] ) , the calibration constant @xmath134 can be temporally considered as a fitting parameter . as mentioned above , it basically affects the ber vs osnr curve in the osnr direction and is related with the detailed osnr monitoring technique used in ref . @xcite . as there is few information about its osnr measurement , evaluation of @xmath134 is expected to be discussed elsewhere . in the above calculation , our cpu time for the noise propagator calculation is @xmath980.8 hr with rk4ip step for transmission fibers ( @xmath13840 km long ) being @xmath1006.0 km . the cpu time for each ber calculation is @xmath98 0.5 hr . the step size @xmath77 within @xmath139 km will result in almost the same curve . for linear ofc systems , mgf method is useful for one to evaluate the noise impacts on bers . this is true not only because it is computationally efficient for the cases with low ber ( e.g. , @xmath140 ) but also it can provide reliable information for the cases using coherent detection , which is now widely used in modern ofc systems . it is now well recognized that traditional gaussian fitting q - factor approximation is accurate for ook detection , while mgf method is accurate for various linear ofc systems . to extend the mgf method to a nonlinear ofc system , one needs to make sure its noise propagator varies within linear regime . this means the noise - noise interaction needs to be neglected and the noise field should follow lne . in cmm of refs . @xcite , the noise propagator was obtained from nlse . it may contain the nonlinearity - induced jitter , which is beyond the linear regime and should be removed . in this work we simplify the cmm by directly solving the accurate lne [ eq . ( [ ns_edfa_ip ] ) ] with rk4ip . like the cw approximations ( cf . refs . @xcite-@xcite and many others ) as well as the approach of ref . @xcite , where the noise propagation information obtained from the related linearized noise equation is automatically free from nonlinearity - induced jitter , the noise propagator obtained from accurate lne also varies within linear regime . to numerically verify this new approach , we consider a 20-span rz - dpsk system discussed in ref . the bers obtained using this new rk4ip agree well with those using cmm in ref . taking account the phase difference between noise and noise - free signal leads to quantitative matching between numerical evaluation and experimental result @xcite . the optical field @xmath141 in a fiber satisfies @xmath142 where @xmath143 is the fiber loss and @xmath144 relates to the cd parameter @xmath145 ( ps / nm / km ) with @xmath146=-@xmath147 ( c=3@xmath14810@xmath149 m / s ) . the slope parameter @xmath150 $ ] [ @xmath151 ( ps / nm@xmath152 km ) ] can be neglected if bit rate @xmath153 satisfies @xmath154 @xcite . introducing the transformation @xmath155 , eq . ( [ nlse ] ) can be reduced as @xmath156 in a @xmath27-span system amplified by @xmath28 edfas ( cf . [ fig_link ] ) , eq . ( [ nlse_2 ] ) can be modified as @xmath157 where @xmath158 is the ase forcing modeled as the complex awgn with correlation @xmath159 in eq . ( [ correlat_t ] ) , the fiber length in each span is assumed to be @xmath2 ( km ) long , according to wiener - kinchine theorem @xcite , @xmath160 in eq . ( [ correlat_t ] ) is the ase psd ( in one polarization direction ) at the output of the @xmath161th edfa . suppose each edfa has the same gain @xmath162 and spontaneous - emission parameter @xmath82 , we have @xcite @xmath163 decomposing optical field @xmath164 in the fiber into noise - free field @xmath165 and its perturbation @xmath166 [ i.e. , @xmath167 and assuming that @xmath168 ( so that the nonlinear terms of @xmath169 can be neglected ) , eq . ( [ nlse_ase ] ) can be decomposed as @xcite @xmath170 @xmath171 noise equation eq . ( [ nlse_ase2 ] ) differs from common equations in that the real and imaginary parts of the complex noise field need to be treated separately @xcite . denoting @xmath172 and the circulant matrices @xmath173_{lm}=\nu_{l - m}$ ] , @xmath174_{lm}=\mu_{l+m}$ ] with @xmath175 in frequency domain , eq . ( [ nlse_ase2 ] ) has the form @xmath176_{lm}a_m -j\gamma(z ) [ m_{\mu}]_{lm}a_m^ * % + \frac{g(z)}{2}a_k -jw_l , \label{ns_g2}\end{aligned}\ ] ] where @xmath177 is the fourier component of the forcing term @xmath158 in eq . ( [ nlse_ase ] ) . as indicated in ref . @xcite , since @xmath178 in ( [ hermit ] ) is real , @xmath179 . so @xmath180 in ( [ ns_g2 ] ) is hermitian , or , its real part @xmath181 is symmetric , while its imaginary part @xmath182 is anti - symmetric . also , as @xmath174_{km}=\mu_{k+m}$ ] , both the real ( @xmath183 ) and imaginary parts ( @xmath184 ) of @xmath185 are symmetric . the matrix form of eq . ( [ ns_g2 ] ) is @xmath186 introducing @xmath187 ( for @xmath188 ) and @xmath189 ( for @xmath190 ) , eq . ( [ ns_g_mtrx ] ) is equivalent to @xmath191 with @xmath192\delta_{ij}$ ] , @xmath193 , @xmath194 , @xmath195 , and @xmath196 . according to the discussion given below eq . ( [ ns_g2 ] ) , @xmath197 ( @xmath198 ) in eq . ( [ hat_l_nu_mu ] ) is antisymmetric ( symmetric ) , respectively . calculation of the kerr term @xmath199 according to eq . ( [ hat_l_nu_mu ] ) has the computational complexity much less than @xmath200 , where @xmath201 is the number of fourier components used for signal representation . in fact , the computational cost of this way is basically determined by the ffts in eq . ( [ hermit ] ) , which has the computational complexity of @xmath202 . in frequency domain , ase correlation relation ( [ correlat_t ] ) has its matrix form ( @xmath203 ) @xmath204 where eq . ( [ psd_ase ] ) has been used . in eq . ( [ auto_noise_f_w ] ) , @xmath205 and @xmath206 are given by eq . ( [ lmt_0 ] ) . ( [ auto_noise_f_w ] ) means that eq . ( [ ns_tilde ] ) can be equivalently replaced by @xmath207 with boundary condition @xcite @xmath208 with @xmath24 being a @xmath209 unit matrix and @xmath31 being the variance of the real or imaginary part of input ase . * appendix b : filtered photoelectric current expressed using klse * given a linear optical system , based on the discussions in ref . @xcite and the notations introduced in ref . @xcite , the filtered photoelectric current @xmath210 can be expressed in the form of @xmath211/2 $ ] . here @xmath212 ( @xmath213 ) represents the signal ( noise ) field at the input of the optical filter . dirac bra @xmath214 is the conjugate transpose ( or hermitian transpose ) of dirac ket @xmath215 [ @xmath216 . the dirac ket differs from usual complex vector in that the @xmath49th element of the latter is just the @xmath49th fourier coefficient of the ( signal or noise ) field , while the @xmath49th element of the former is the product of the @xmath49th fourier coefficient and its base function ( cf . ( 17 ) in ref . @xcite ) . according to refs . @xcite , the filtered current in a linear system can be formally expressed as @xmath217 with ( @xmath218 ; @xmath219 ) @xmath220/2 = \langle s^o(t_{\mathrm{s}})\|r_{ss}^d\|s^o(t_{\mathrm{s}})\rangle \nonumber \\ % ! gamma_{nn } y_{nn}(t_{\mathrm{s}})\!\!\!&\!\!\!=\!\!\!&\!\!\![\langle n^o(t_{\mathrm{s}}+t_b)\|r_{nn}\|n^o(t_{\mathrm{s}})\rangle+c.c.]/2 = \langle n^o\|r_{nn}^d\|n^o\rangle=\langle z\|\lambda\|z\rangle \nonumber \\ % gamma_{ns } y_{ns}(t_{\mathrm{s}})\!\!\!&\!\!\!=\!\!\!&\!\!\ ! [ \langle n^o(t_{\mathrm{s}}+t_b)\|r_{ns}\|s^o(t_{\mathrm{s}})\rangle+\langle n^o(t_{\mathrm{s}})\|r_{ns}\|s^o(t_{\mathrm{s}}+t_b)\rangle+c.c.]/2 \nonumber \\ \!\!\!&\!\!\!=\!\!\!&\!\!\ ! [ \langle n_{in}\|o^{\dagger}_{nn}r_{ns}^d\|s^o(t_{\mathrm{s}})\rangle+c.c]=[\langle z\|b^d(t_{\mathrm{s}})\rangle + c.c . ] \label{y_d}\end{aligned}\ ] ] where @xmath221 , @xmath222 , and @xmath223 in eqs . ( [ y_d])-([r_dpsk ] ) , @xmath224 represents the decoupled gaussian random variables with zero mean and real part and imaginary part variance of @xmath31 . the effects of the optical and electrical filters in the receiver are represented by matrices with their elements being @xmath225 , @xmath226 and @xmath227 . due to the optical and electrical filters , signal ( noise ) components outside @xmath228 ( @xmath229 ) can be neglected . here @xcite @xmath230 for a nonlinear optical system , to get the noise propagator from the accurate lne ( [ ns_tilde ] ) , one needs to separate complex numbers into their real and imaginary parts . denoting the re - im form of a complex matrix @xmath231 as @xmath232 [ where @xmath231 can be any complex matrix in eq . ( [ y_d ] ) ] and introducing @xmath233 with @xmath234 being the awgn from edfa , @xmath235 it is easy to generalize the noise related currents in eq . ( [ y_d ] ) as @xcite @xmath236 where @xmath34 can be obtained from eqs . ( [ h_rk4ip])-([p_noise_l ] ) and ( [ pg_k])-([pg_k_eq ] ) . * appendix c : nonlinearity induced phase difference between noise and noise - free signal * * i : the phase difference caused by @xmath237 * in this part , we assume that , in fig . [ fig_link ] , the external noise injected at the transmitter ( @xmath237 ) is much larger than the ase noise from the edfas , which is true for the experiments discussed in ref . thus , one can only consider the phase difference caused by @xmath237 and ignores the effect of @xmath238 ( @xmath239 ) . it is well known that , for the noise - free signal with its path average power being @xmath240 , its nonlinear phase accumulated at the fiber output is @xmath241 , where @xmath2 is the fiber length of each span , as denoted in fig . [ fig_link ] . due to the optical power fluctuation @xmath242 , the actual nonlinear phase becomes @xmath243 relative to the noise - free signal , the noise - induced phase change , @xmath244 , varies randomly . the average variance of such phase noise can be calculated as @xmath245 with the approximation of negligible @xmath160 ( @xmath246 ) , @xmath247 in eq . ( [ d_phi_2 ] ) is basically caused by @xmath237 in fig . [ fig_link ] . for the experiments in ref . @xcite , @xmath237 is filtered with bandwidth of @xmath248=3 nm . thus , we have @xmath249 . eq . ( [ d_phi_2 ] ) yields @xmath250 where @xmath251 is the input osnr . considering that @xmath252 , we have @xmath253 . this means @xmath254 rotates faster than @xmath255 and there is a phase difference between the actual optical field and the noise - free field . as part of the actual field , the noise field also has the same phase shift relative to the noise - free signal . note that this phase shift will not affect signal - signal and noise - noise beatings . but it will affect the signal - noise beating . in fact , when calculating the signal - noise beating , the noise and signal should be treated consistently . or , they should be considered within the same coordinate system . in this work , @xmath131 given by eq . ( [ d_phi_1 ] ) is approximated as the average of such phase shift . our numerical results plotted in figs . [ fig_compare_coelho_1 ] and [ fig_compare_coelho_2 ] confirm this approximation . for the experiments in ref . @xcite , we have @xmath256 . in this case , eq . ( [ d_phi_1 ] ) yields @xmath257 where @xmath258 and @xmath259 ( for @xmath260 ) have been used . obviously , @xmath131 in eq . ( [ d_phi_1d ] ) relates @xmath261 in eq.(23 ) of ref . @xcite with @xmath262 . also , @xmath134 in ref . @xcite now becomes @xmath263 in eq . ( [ gm_shift ] ) , while @xmath264 in ref . @xcite is named as @xmath129 in this work . * ii : the phase difference caused by @xmath237 and @xmath238 ( @xmath239 ) * for the @xmath27-span system of fig . [ fig_link ] with @xmath265 and @xmath27 not being large enough and with the ase from each edfa being filtered by @xmath102 and the ase injected at the transmitter being filtered by @xmath104 , eq . ( [ d_phi_2 ] ) can be generalized as @xmath266 \bar \phi_{nl}^2/\bar { p } , \label{d_phi_2c}\ ] ] where @xmath267 has been used . obviously , in the case of @xmath268 @xcite , eq . ( [ d_phi_2c ] ) yields eq . ( [ d_phi_1 ] ) . the authors acknowledge the financial support from canada research chair program . the first author sincerely thanks paolo serena and leonardo d. coelho for providing their detailed calculations in refs . the authors wish to thank the anonymous reviewers for their valuable comments and suggestions .	in a nonlinear optical fiber communication ( ofc ) system with signal power much stronger than noise power , the noise field in the fiber can be described by linearized noise equation ( lne ) . in this case , the noise impact on the system performance can be evaluated by moment - generating function ( mgf ) method . many published mgf calculations were based on the lne using continuous wave ( cw ) approximation , where the modulated signal needs to be artificially simplified as an unmodulated signal . results thus obtained should be treated carefully . more reliable results can be obtained by improving the cw - based lne with the accurate lne proposed by holzl@xmath0hner _ et al _ in ref . @xcite . in this work we show that , for the case of linearized noise amplified by edfas , its mgf can be calculated by obtaining the noise propagation information directly from the accurate lne . our results agree well with the experimental data of multi - span dpsk systems . 0.10 in 0.12 in r. holzl@xmath0hner , v. s. grigoryan , c. r. menyuk , and w. l. kath , `` accurate calculation of eye diagrams and bit error rates in optical transmission systems using linearization , '' j. lightwave technol . * 20 * , 389 - 400 ( 2002 ) . k. kikuchi , `` enhancement of optical - amplifier noise by nonlinear refractive index and group - velocity dispersion of optical fibers , '' ieee photon . . lett . * 5 * , 1079 - 1081 ( 1993 ) . a. carena , v. curri , r. gaudino , p. poggiolini , and s. benedetto , `` new analytical results on fiber parametric gain and its effects on ase noise , '' ieee photon . technol . lett . * 9 * , 535 - 537 ( 1997 ) . r. hui , m. osullivan , a. robinson , and m. taylor , `` modulation instability and its impact in multispan optical amplified imdd systems : theory and experiments , '' j. lightwave technol . * 15 * , 1071 - 1082 ( 1997 ) . p. serena , a. orlandini , and a. bononi , `` parametric - gain approach to the analysis of single - channel dpsk / dqpsk systems with nonlinear phase noise , '' j. lightwave technol . * 24 * , 2026 - 2037 ( 2006 ) . a. demir , `` nonlinear phase noise in optical - fiber - communication systems , '' j. lightwave technol . * 25 * , 2002 - 2032 ( 2007 ) . l. d. coelho , l. molle , d. gross , n. hanik , r. freund , c. caspar , e .- d . schmidt , and b. spinnler , `` modeling nonlinear phase noise in differentially phase - modulated optical communication systems , '' opt . express * 17 * , 3226 - 3241 ( 2009 ) . m. secondini , e. forestieri , c. r. menyuk , `` a combined regular - logarithmic perturbation method for signal - noise interaction in amplified optical systems , '' j. lightwave technol . * 27 * , 3358 - 3369 ( 2009 ) . r. holzl@xmath0hner , `` a covariance matrix method for the computation of bit errors in optical transmission systems , '' ph.d . thesis , university of maryland baltimore county . baltimore , maryland , usa ( 2003 ) . r. holzl@xmath0hner , c. r. menyuk and w. l. kath , `` efficient and accurate computation of eye diagrams and bit error rates in a single - channel crz system , '' ieee photon . technol . lett . * 14 * , 1079 - 1081 ( 2002 ) . r. holzl@xmath0hner , c. r. menyuk and w. l. kath , `` a covariance matrix method to compute bit error rates in a highly nonlinear dispersion - managed soliton system , '' ieee photon . . lett . * 15 * , 688 - 690 ( 2003 ) . r. holzl@xmath0hner and c. r. menyuk , `` use of multicanonical monte carlo simulations to obtain accurate bit error rates in optical communications systems , '' opt . lett . * 28 * , 1894 - 1896 ( 2003 ) . a. demir , `` non - monte carlo formulations and computational techniques for the stochastic nonlinear schrodinger equation , '' j. comput . phys . * 201 * , 148 - 171 ( 2004 ) . j. hult , `` a fourth - order runge - kutta in the interaction picture method for simulating supercontinuum generation in optical fibers , '' j. lightwave technol . * 25 * , 3770 - 3775 ( 2007 ) . z. zhang , l. chen , and x. bao , `` a fourth - order runge - kutta in the interaction picture method for numerically solving the coupled nonlinear schr@xmath0dinger equation , '' opt . express * 18 * , 8261 - 8276 ( 2010 ) . a.m. mathai and s.b . provost , _ quadratic forms in random variables : theory and applications _ ( marcel dekker , 1992 ) . e. forestieri , `` evaluating the error probability in lightwave systems with chromatic dispersion , arbitrary pulse shape and pre- and postdetection filtering , '' j. lightwave technol . * 18 * , 1493 - 1503 ( 2000 ) . z. zhang , l. chen , and x. bao , `` accurate ber evaluation for lumped dpsk and ook systems with pmd and pdl , '' opt . express * 15 * , 9418 - 9433 ( 2007 ) . d. gariepy and g. he , `` measuring osnr in wdm systems effects of resolution bandwidth and optical rejection ratio , '' white paper , exfo inc . ( 2009 ) . p. serena , a. bononi , j .- c . antona , and s. bigo , `` parametric gain in the strongly nonlinear regime and its impact on 10-gb / s nrz systems with forward - error correction , '' j. lightwave technol . * 23 * , 2352 - 2363 ( 2006 ) . a.bononi , p. serena , and a. orlandini , `` a unified design framework for single - channel dispersion - managed terrestrial systems , '' j. lightwave technol . * 26 * , 3617 - 3631 ( 2008 ) . e. ip and j. m. kahn , `` power spectra of return - to - zero optical signals , '' j. lightwave technol . * 24 * , 1610 - 1618 ( 2006 ) . a. papoulis , _ probability , random variables , stochastic processes _ ( mcgraw - hill , 1991 ) . e. forestieri and g. prati , `` exact analytical evaluation of second - order pmd impact on the outage probability for a compensated system , '' j. lightwave technol . * 22 * , 988 - 996 ( 2004 ) . e. forestieri and g. prati , `` exact analytical evaluation of second - order pmd impact on the outage probability for a compensated system , '' j. lightwave technol . * 22 * , 988 - 996 ( 2004 ) . z. zhang , l. chen , and x. bao , `` the noise propagator in an optical system using edfas and its effect on system performance : accurate evaluation based on linear perturbation , '' arxiv : physics.optics/1207.3362v1 .
global observables such as the multiplicity and the inclusive single particle transverse momentum ( ) and pseudorapidity ( @xmath1 ) distributions of hadrons have been valuable tools in studying heavy ion collisions . they represent the system at kinetic freeze - out , late in the evolution of the system when hadrons no longer interact with each other , and their momentum spectra do not change further . these final - state observables supply essential constraints on the possible evolutionary paths of the system that can help establish conditions in the early , hot and dense phase of the collision . a discussion on global observables can be found in @xcite . we summarize here results on the minimum - bias multiplicity , and @xmath1 spectra of negatively charged hadrons . the star experimental setup for the first rhic run is described in @xcite . the analysis is based on charged particle tracking in the star time projection chamber ( tpc ) . triggering was achieved using two hadronic calorimeters ( zdcs ) in the very forward region and an array of scintillator slats arranged in a barrel ( ctb ) around the tpc . particle production was studied through the yield of primary negative hadrons ( ) , _ i.e. _ , and including the products of strong and electromagnetic decays . low momentum ( @xmath2 ) particle identification was done via @xmath3 in the gas of the tpc . only negatively charged hadrons were studied in order to exclude effects due to participant nucleons . charged particle tracks reconstructed in the tpc were accepted if they fulfilled requirements on number of points on the track and on pointing accuracy to the event vertex . the measured raw distributions were corrected for acceptance , reconstruction efficiency , contamination due to interactions in material , misidentified non - hadrons , the products of weak decays , and track splitting and merging . the tracking efficiency was found by embedding simulated tracks into real events at the raw data level , reconstructing the full events , and comparing the simulated input to the reconstructed output . for this analysis , the acceptance for tracks within the fiducial volume having greater than 0.3 is found to be 95% . the tracking efficiency is found to be between 7095% , depending on the particle and the total multiplicity of the event . 75 mm figure [ fig : hminus ] shows the multiplicity distribution for minimum - bias au+au collisions with @xmath4 and @xmath5 . the data were normalized assuming a total hadronic inelastic cross section of 7.2 barn for au+au collisions at = 130 gev , derived from glauber model calculations @xcite . the systematic error on the vertical scale is estimated to be 10% and is dominated by uncertainties of the total hadronic cross - section and the shape at low @xmath6 . the systematic error of 6% on the horizontal scale is depicted by horizontal error bars on a few data points . the shape of the distribution is dominated by the collision geometry : large cross section at low multiplicity , corresponding to large impact parameter , followed by a region of slowly falling cross section over a wide range of multiplicity as the nuclei overlap , and a rapid decrease for near head - on collisions , where the shape is determined by fluctuations in the collision geometry and particle production process given the finite detector acceptance . the distribution for the 5% most central collisions ( 360 mbarn ) , defined via zdc coincidence , is shown as the shaded area in fig . [ fig : hminus ] . figure [ fig : pt ] , upper panel , shows the distribution for the 5% most central collisions at midrapidity ( @xmath7 ) . only statistical errors are shown ; the systematic errors are estimated to be below 6% . the data are fit by a power - law function of the form @xmath8 . as reference data sets , we also show the -distribution of negatively charged hadrons for central pb+pb collisions at = 17 gev from na49 @xcite and for minimum - bias collisions at = 200 gev from ua1 @xcite , fitted with the same function . the ua1 invariant cross - section @xmath9 reported in ref . @xcite was scaled by @xmath10 , where @xmath11 mb . the spectrum measured by star reflects a systematic increase in compared to that from both central a+a collisions at much lower energy ( na49 ) and collisions at a comparable energy ( ua1 ) . 80 mm the lower panel of fig . [ fig : pt ] shows the ratio of the star and ua1 -distributions . since the ua1 distribution is measured at = 200 gev , the invariant cross - section in each -bin is scaled by two factors for a quantitative comparison to the star data : _ ( i ) _ @xmath12 , the -dependent ratio of the yields in collisions at = 130 and 200 gev , and _ ( ii ) _ = 26 mb@xmath13 , the nuclear overlap integral @xcite for the 5% most central au+au collisions . see @xcite for further details . there are two simple predictions for the scaled ratio . the study of lower energy collisions has shown that the total pion yield due to soft ( low ) processes scales as the number of participants , i.e. , `` wounded '' nucleons ( ) in the collision , ( e.g. @xcite ) . the scaled ratio in this case is 0.164 assuming 172 participant pairs @xcite and a mean number of binary collisions ( ) of @xmath14 for the 5% most central au+au events . we assume @xmath15 mb for collisions at @xmath16 gev . if hadron production is due to hard ( high ) processes and there are no nuclear - specific effects ( see below ) , the hadron yield will scale as , which is proportional to the nuclear overlap integral . in this case the value of the ratio is unity . there are important nuclear effects which should alter the scaling from these simple predictions , including _ e.g. _ initial state multiple scattering @xcite , jet quenching @xcite , and radial flow @xcite . each of these exhibits characteristic features as a function of and system size . the scaled ratio shows a strong dependence , starting close to scaling at low and approaching scaling , not reaching the latter in the measured range even including the errors from the @xmath12 and scaling shown in gray . this behaviour is consistent with the presence of radial flow , as well as the onset of hard scattering contributions and initial state multiple scattering with rising . the shape of the distribution at higher is discussed in @xcite . the density at midrapidity for @xmath17 and @xmath7 is @xmath18 . extrapolation to @xmath19 = 0 yields @xmath20 . assuming an average of 172 participant pairs per central au+au collision , this corresponds to @xmath21 per participant nucleon pair per unit of pseudorapidity . this is a 35% increase over collisions at the same energy @xcite . comparison with central collisions at sps @xcite shows an increase in the yield per participant of @xmath22 49% at rhic . [ fig : etacent ] shows the centrality dependence of the @xmath1 distribution . one sees the expected rise in particle yield with collision centrality . preliminary analysis shows very little difference in the shape of the distribution at midrapidity with increasing centrality . using again a power - law fit to the spectra we obtain which is shown in fig . [ fig : meanptcent ] as a function of centrality . errors on the horizontal scale are systematic . we see an increase in of 15% from the most peripheral to the most central events . the values for the ua1 and na49 reference data sets are also shown . we have presented preliminary results on charged particle production at rhic . we find @xmath23 per participant in central collisions at = 130 gev increases by 35% relative to and 49% compared to nuclear collisions at = 17 gev . the distribution is harder than that of the reference systems for the region up to 2 , with the most central events having the highest . scaling of produced particle yield with number of participants shows a strong dependence on , with wounded nucleon scaling achieved only at the lowest measured . 99 k. j. eskola , these proceedings . see also j .- blaizot , nucl . a * 661 * , 3c ( 1999 ) . f. retire , these proceedings . ackermann _ et al . _ , nucl . a * 661 * , 681c ( 1999 ) . baltz , c. chasman , and s.n . white , nucl . a * 417 * , 1 ( 1998 ) . d. kharzeev and m. nardi , phys . b * 507 * 121 ( 2001 ) nucl - th/0012025 . h. appelshuser _ et al . _ , phys . lett . * 82 * , 2471 ( 1999 ) . c. albajar b * 355 * , 261 ( 1990 ) . dunlop , these proceedings . k. j. eskola , k. kajantie , and j. lindfors , nucl . b * 323 * , 37 ( 1989 ) . d. antreasyan , phys . d * 19 * , 764 ( 1979 ) . wang , phys . c * 58 * , 2321 ( 1998 ) . p. f. kolb , j. sollfrank , and u. heinz , phys . c * 62 * , 054909 ( 2000 ) . et al . _ , z. phys . c * 33 * , 1 ( 1986 ) .	negatively charged hadron ( ) production in au+au collisions at bnl - rhic is studied with the star experiment . results are presented on multiplicity , pseudorapidity and transverse momentum distributions at @xmath0 = 130 gev .
in the analytical part of our study , we analyze two ( weakly bound ) magnetic impurities in a three - dimensional @xmath0-wave superconductor . the corresponding hamiltonian @xmath1 can be decomposed into the kinetic energy , @xmath2 , the bcs mean field superconductivity , @xmath3 , and the impurity contribution , @xmath4 . if there were no impurities , the system could be modeled by the hamiltonian @xmath5 , where @xmath6 ( @xmath7 denotes the kinetic energy dispersion ) , and @xmath8 with a superconducting order parameter of constant value @xmath9 ( chosen to be positive ) . here , @xmath10 is the annihilation operator for an electron of three - dimensional momentum @xmath11 and spin @xmath12 , while @xmath13 creates a spin @xmath12 electron at position @xmath14 ( in the calculations , we use @xmath15 , @xmath16 ) . due to the impurities , however , the superconducting part of the hamiltonian is promoted from @xmath17 to @xmath18 the order parameter is set by the self - consistent equation @xmath19 , where @xmath20 is the microscopic attractive interaction between electrons in the superconductor , and where the expectation value is taken with respect to the full @xmath21 . the magnetic impurities , finally , are treated as classical spins @xmath22 residing at positions @xmath23 , and modeled by a purely magnetic @xcite and point - like scattering potential . their classical treatment implies that the we restrict our analysis to sufficiently large impurity spins . we address impurities polarized along the @xmath24 axis in spin space , thus covering both parallel and antiparallel impurity spin alignments . the corresponding impurity hamiltonian reads @xmath25 where @xmath26 is the @xmath24 component of the spin of impurity @xmath27 , and @xmath28 is the exchange coupling between this impurity and the electrons in the superconductor . this model comes with the debye frequency @xmath29 as a natural high - energy cutoff . if the renormalization of the superconducting gap is small , @xmath30 , one can approximate @xmath31 by evaluating the right - hand side of the gap equation for an unrenormalized gap , @xmath32 with @xmath33 . this is the approach we employ in the remainder of the analytical calculation . when @xmath34 becomes of order @xmath9 , this approximation ceases to be valid , and we make use of numerical simulations . the tight - binding hamiltonian @xmath35 is defined as @xmath36+\text{h.c.},\nonumber\end{aligned}\ ] ] where @xmath37 is the annihilation operator acting on an electron with spin @xmath12 at lattice site @xmath27 , and the first sum runs over neighboring sites @xmath27 and @xmath38 located in a two - dimensional square lattice of size @xmath39 with lattice constant @xmath40 . the chemical potential @xmath41 is taken from the bottom of the energy band , and the local order parameter @xmath42 is determined self - consistently in an iterative procedure for fixed value of the exchange coupling @xmath43 at the site @xmath27 starting from the uniform superconducting order parameter @xmath9 until convergence is reached . we begin with calculating the full imaginary time nambu green s function @xmath44 using the ` bare ' green s function @xmath45 , and where the nambu spinor is defined as @xmath46 . for equal positions @xmath47 , a fourier transformation from imaginary time to matsubara frequencies @xmath48 yields @xmath49 with @xmath50 being the density of states ( per spin ) at the fermi energy , and with the pauli matrices @xmath51 acting in nambu space ( @xmath52 ) , while we obtain @xmath53 at distances @xmath54 larger than the fermi wavelength ( @xmath55 is the fermi momentum ) . using the imaginary time - dependent dyson equation , the full green s functions can be expressed as @xmath56 where the @xmath57-matrix @xmath58^{-1}$ ] involves the @xmath59 -matrix @xmath60 with entries @xmath61 , and @xmath62 . when the impurities are far apart from each other , @xmath63 , the gap near a given impurity is predominantly renormalized by scattering processes off this impurity . @xcite we can then for instance focus on impurity @xmath64 . next to this impurity , and using @xmath65 , the dominant contribution to the gap renormalization at @xmath66 reads@xcite @xmath67 a gaussian high - energy cutoff at the scale @xmath29 regularizes the logarithmic uv divergence of this integral . for small @xmath68 , we can furthermore expand this expression to second order in @xmath69 . since the self - consistent equation yields a bare gap of @xmath70 , we find @xmath71 for a more strongly bound shiba state , @xmath72 , our analytical calculation hints at a very strong suppression of the gap at the positions of the impurities , @xmath73 , but the non - selfconsistent approach can not make any quantitative predictions in this regime . numerically , on the other hand , we can readily access the regime @xmath74 , see fig . [ two_impurity ] . for small values of @xmath75 , the gap under a ( single ) impurity is suppressed quadratically in @xmath75 in agreement with eq . ( [ eq : deltadeltasmall3 ] ) . interestingly , at a critical value @xmath76 corresponding to the phase transition , the superconducting gap under the impurity changes its sign and magnitude abruptly , giving rise to a local @xmath77 junction , see fig . [ two_impurity]a . also the energy of the bound state jumps from zero to a finite negative value . this feature results from self - consistency in determining @xmath78 and can not be captured analytically in the approach we used above . the stronger @xmath75 becomes , the smaller @xmath79 under the impurity gets , until the superconducting gap is eventualy totally suppressed . as a result , the internal s - s junction evolves into an s - n junction . after the phase transition , the modulus of the bound state energy increases , whereas the superconducting gap @xmath80 decreases as functions of @xmath75 . this means that the energy of the bound state lies outside the gap for @xmath81 . we refer to this state as an andreev bound state localized in an effective s - s junction . the transition is clearly visible in our numerical results plotted in fig . [ two_impurity]a . a key signature distinguishing shiba and andreev bound states is their degree of localization . this is demonstrated in fig . [ two_impurity_numerical ] , which shows the numerically obtained bound state wave functions . indeed , the shiba states are more localized under the impurity , see fig . [ two_impurity_numerical]a . in contrast , the andreev bound states are delocalized directly under the impurity as their energies are higher than the superconducting gap in this region , see fig . [ two_impurity_numerical]b . this difference in the characteristics of the bound state wavefunctions can be checked experimentally with stm techniques . we note that the filling of the shiba state does not necessarily coincide with the shiba to andreev transition , see again fig . [ two_impurity]a . as a consequence , there exists a finite range of exchange couplings with a filled but localized shiba state , which delocalizes at the shiba to andreev transition . our results agree qualitatively with earlier numerical studies of a single impurity , @xcite which , however , do not address the shiba to andreev state transition . directly under the impurity ( red dots ) for a square lattice of size @xmath82 with ( a ) a single impurity ( b ) two identical impurities ( @xmath83 ) as function of exchange @xmath84 found numerically . @xmath79 gets suppressed with increasing @xmath75 and an effective s - s junction builds up . after the phase transition at the critical value @xmath85 , @xmath79 changes its sign giving rise to a @xmath77-junction . sub - gap shiba states ( green dots ) evolve into supra - gap andreev states ( brown dots ) at values of @xmath86 where the bound state energy coincides with @xmath79 . ( b ) for two impurities , separated by a distance @xmath87 , the hybridization between the two bound states lifts their degeneracy . there are thus two phase transitions , one at @xmath85 and one at @xmath88 , where these states become occupied , accompanied by jumps in bound state energies and , hence , in @xmath79 and @xmath89 . there are also two distinct shiba to andreev state transitions . the gap between the impurities ( blue dots ) is suppressed for small @xmath75 , when the two bound states have finite overlap , but restores at larger @xmath75 , see also fig . [ cross_section ] . for corresponding wavefunction plots , see fig . [ two_impurity_numerical ] . the chosen parameters are @xmath90 and @xmath91 . ] of the bound states in the case of a single impurity ( a , c ) and of two identical impurities ( b , d ) , @xmath92 found in the tight - binding model on a lattice of size @xmath93 by self - consistent numerical diagonalization . for two impurities , separated by a distance @xmath87 , the hybridization between the two bound states lifts their degeneracy . before the phase transition with @xmath94 ( a , b ) , the shiba bound state is well localized directly under the corresponding impurity . the andreev bound state occurring after the phase transition at stronger values of the exchange coupling , @xmath95 ( c , d ) , is delocalized over the entire region where the superconducting gap is suppressed , with a maximum wavefunction amplitude between the impurities , see ( d).the chosen parameters are @xmath90 and @xmath91 . ] for two impurities located on the @xmath96 axis for fixed @xmath97 and varying @xmath75 : @xmath98 ( yellow dotted line ) , @xmath99 ( green dotted dashed line ) , @xmath100 ( red dashed line ) , @xmath101 ( blue solid line ) . the gap under the first impurity is increasingly suppressed with increasing exchange coupling : after the phase transition @xmath102 , @xmath77-junctions ( negative gap value ) under the impurity arise accompanied by restoration of the gap between the two impurities . the gap under the second impurity is suppressed the most when @xmath103 since then the two bound states hybridize strongly . for wavefunction plots , see fig . [ two_impurity_numerical ] . all parameters are the same as in fig . [ two_impurity]b . ] to obtain further insight into the sub - gap shiba to supra - gap andreev state transition , we study the case of a single impurity with a classical spin @xmath104 at position @xmath105 . we use the simple gap renormalization @xmath106 to mimic a suppression of the gap from @xmath9 to @xmath107 within a region of volume @xmath108 around the impurity ( note that we require @xmath109 since there can not be a bound state with energy larger than the gap if the latter is locally enhanced ) . inverting the hamiltonian,@xcite we obtain an equation for the wave function at the impurity site , @xmath110 , in terms of the bound state energy @xmath111 , @xmath112 where @xmath113 , and @xmath114 corresponds to the parallel and antiparallel orientation of the bound state spin as compared to the impurity spin , respectively . this equation can be rewritten as @xmath115\psi ( 0)=0\ , , \label{an_h}\end{aligned}\ ] ] where @xmath116 , and @xmath117 . we find that bound state energies @xmath111 within the bulk gap @xmath9 satisfying eq . are given by @xmath118 where @xmath119 is the eigenvalue of @xmath120 , and @xmath121 . plugging this expression for @xmath111 back into eq . , we find that a necessary condition is @xmath122 . because consistency with bcs theory requires @xmath123 , and since we are primarily interested in the regime where @xmath124 is of order unity , we find that @xmath125 . therefore , @xmath119 and @xmath126 must be of the same sign , and there are two solutions with opposite spin like for the uniform case.@xcite defining as above the bound state as a shiba state when its energy is within the renormalized gap , @xmath127 , and as an andreev state when its energy is between the renormalized gap and the bulk gap , @xmath128 , the critical value of the exchange coupling for the transition from a shiba to an andreev state reads @xmath129 , @xmath130 ( as in the numerics ) , and for @xmath131 , @xmath132 , while otherwise the fraction under the square root has to be inverted . when the impurities are further apart than the fermi wavelength , @xmath133 , we can expand the gap renormalization in orders of @xmath134 . to second order , we find @xmath135 where @xmath136^{-1}$ ] is the @xmath57-matrix for scattering off impurity @xmath137 . the terms @xmath138 give rise to the single impurity renormalizations of the order parameter , which we denote as @xmath139 , while the other corrections to the bare green s function result in inter - impurity gap renormalizations @xmath140 . these different contributions are depicted in fig . [ fig : gap_renorm ] . the inter - impurity renormalizations have both terms that depend on the relative orientation of the two spins ( parallel for @xmath141 , antiparallel for @xmath142 ) , as well as orientation - independent contributions . all of these are of the order @xmath143 . most interesting is the orientation - dependent contribution evaluated at the site of one of the impurities . as an example , @xmath144 reads at zero temperature @xmath145\nonumber\\ & -2(1-\alpha_1 ^ 4)(1+\alpha_2 ^ 2)\omega^2\cos(2k_fr_{12})\bigr\},\label{eq : renorm_12}\end{aligned}\ ] ] with @xmath146 . the upper ( lower ) sign applies for parallel ( antiparallel ) impurity spin alignment . besides addressing the gap renormalization at the sites of the impurities , it is also interesting to analyze the scaling of the different contributions to the gap renormalization as a function of the inter - impurity distance in their middle , that is at @xmath147 . the single impurity contributions are found to scale as @xmath148 where @xmath149 is the superconducting coherence length . this scaling has a simple geometrical interpretation : to leading order in @xmath150 , the anomalous green s function ( which determines the gap ) is renormalized by the electron traveling from @xmath151 to impurity @xmath27 at @xmath152 , scattering there , and coming back to @xmath151 . since a trip to , and back from , the impurity involves a propagator proportional to @xmath153 , see eq . , the gap renormalization deriving from the single impurity scattering processes scales as @xmath143 . the total distance covered during these processes , precisely equal to @xmath154 , determines the argument of the exponential . the leading order inter - impurity terms , on the other hand , are found to scale as @xmath155 since they stem from an electron traveling first to impurity one , then to impurity two , and then coming back . this involves three trips , and a total distance of @xmath156 . the inter - impurity renormalization @xmath140 is thus suppressed by an additional power of @xmath157 in between the impurities as compared to its value at one of the impurities , see eq . . consequently , the inter - impurity gap renormalization @xmath140 can be modeled as a function with well - defined peaks close to the two impurities , as shown in fig . [ fig : gap_renorm ] . we also confirm numerically that the renormalizations of the superconducting gap is the largest directly under the impurity , while , depending on parameters , @xmath158 could even get _ larger _ than its initial value around the impurities , see fig . [ cross_section ] . we finally note that the regime @xmath159 is qualitatively similar to the regime @xmath160 . this is illustrated by fig . [ two_impurity]b , which is not far from this point . with increasing exchange strengths , a similar transition as before between shiba and andreev states can also be observed in the case of two impurities , see fig . [ two_impurity]b . when the two impurities are close and with comparable exchange strengths , the bound states overlap and their degeneracy gets lifted by hybridization , @xcite which in turn gives rise to two separate quantum phase transitions in @xmath161 , one at @xmath85 and one at @xmath88.@xcite similarly , we find now two distinct shiba to andreev state transitions around these values of exchange strengths , see fig . [ two_impurity]b . some illustrative plots of the corresponding wavefunctions are given in fig . [ two_impurity_numerical ] . the spectroscopic signatures of the shiba to andreev transition are now even more drastic than for the single impurity case . as shown in fig . [ two_impurity_numerical]b , the bound state wavefunction of two strongly localized , but hybridized shiba states is strongly suppressed between the impurities . in stark contrast , the wavefunction of the andreev bound state spreads between the impurities , and is even maximal between the impurities for the chosen aprameters , see fig . [ two_impurity_numerical]d . , in the presence of two spin - impurities localized at positions @xmath162 and @xmath163 along the @xmath164 axis connecting the two impurities . the sketch shows the envelopes of three different contributions to the gap renormalization ( see text).,title="fig : " ] + in the case of small impurity distances @xmath165 , where @xmath166 . keeping all orders of @xmath167 in eq . , we find that the gap renormalization is in this limit given by @xmath168 where @xmath169 is given in eq . . quite naturally , the gap renormalization resulting from two very close classical impurities with @xmath69 and @xmath170 is thus equal to the renormalization of a single impurity with @xmath171 . provided that @xmath172 and @xmath173 have the same sign , the superconducting gap is reduced the least if the two spins are aligned antiparallel for short distances . as a final remark , we note that our non - selfconsistent analytical approach is also able to tackle the regime of intermediate impurity distances @xmath159 . in this regime , we find some lengthy equations that simply connect the limits of small and large impurity distances in a smooth fashion . as an illustration , we refer the reader to the numerical plots in fig . [ two_impurity]b . these are not far from the regime @xmath159 , and even treat the gap self - consistently . analyzing two classical spins in a superconductor , we obtained the gap renormalization analytically for weakly and numerically for strongly coupled bound states . in addition , we found a transition for shiba to andreev states which is accompanied by two subsequent quantum phase transitions as the exchange coupling is varied . all these predictions lead to dramatic spectral changes . in order to observe these effects by stm tecniques,@xcite it would be preferable to control experimentally the exchange interaction parametrized by @xmath174 . this can be realized for example by considering a 2d electron gas ( with immersed magnetic impurities ) deposited on top of a bulk superconductor such that the sc gap is now a proximity gap . we propose to bring an stm tip on top of a magnetic impurity in order to locally control the impurity density of states without affecting its spin and therefore to continuously monitor the parameter @xmath174 . a number of additional directions can be envisioned in future work . instead of investigating the physics associated with large impurity spins , for which a classical treatment is appropriate , one could generalize our work the limit of small impurity spins , for which kondo physics becomes relevant at low temperatures . with two close - by impurities , this could for instance lead to interesting multi - stage screening effects . another promising generalization is offered by the use of more exotic superconductors , such as @xmath175-wave materials , where impurities induce virtual bound states @xcite . we thank l. glazman and s. gangadharaiah for valuable discussions . this work has been supported by the swiss nf , nccr qsit , by the dfg through grk 1621 and sfb 1143 , and by the french agence nationale de la recherche through the anr contracts dymesys and mistral .	we study the renormalization of the gap of an @xmath0-wave superconductor in the presence of two magnetic impurities . for weakly bound shiba states , we analytically calculate the part of the gap renormalization that is sensitive to the relative orientation of the two impurity spins . for impurities with a strong exchange coupling to the conduction electrons , we solve the gap equation self - consistently by numerics and find that the sub - gap shiba state turns into a supra - gap andreev state when the local gap parameter changes sign under the impurities . the study of magnetic impurities in superconductors has a long history . important effects include the renormalization ( reduction ) of the gap in the host superconductor by the impurities,@xcite and the induced yu - shiba - rusinov , or shiba , bound states.@xcite in recent years , renewed interest in chains of magnetic impurities in superconductors was furthermore driven by their potential of hosting majorana zero modes . @xcite using scanning tunneling microscopy , the shiba states induced by individual magnetic impurities on the surface of a superconductor have been shown to be strongly localized at the impurity sites.@xcite this can be understood as a consequence of their sub - gap energy . when this energy is below the fermi level , the shiba state is occupied by a single electron , whose spin direction is dictated by the impurity spin.@xcite shiba states are to be contrasted with andreev bound states in extended superconductor - normal - superconductor ( s - n - s ) heterostructures , or in heterostructures composed of superconductors with different gaps ( s - s-s ) . andreev bound states can be viewed as standing waves of electrons and holes inside the gapless region that result from andreev reflections on the s - n - s and s - s-s interfaces.@xcite in this work , we address a scenario intermediate between single impurities and impurity chains , and specifically study the renormalization of the superconducting order parameter in the presence of two magnetic impurities . we quantify the inter - impurity scattering analytically and discuss how the shiba states are affected by the gap renormalization . in particular , we observe that sub - gap shiba states can transmute into supra - gap andreev states when the impurities are strongly bound to the superconductor , which we show by a simple analytical model and by self - consistent numerics . the outline of the paper is as follows . in the next section , we introduce our model hamiltonian for the magnetic impurity in an @xmath0-wave superconductor . in sec . [ sec : t_matrix ] , we summarize briefly our analytical approach based on the t - matrix formalism . in sec . [ sec : single_imp ] , we apply it to a single classical impurity . in sec . [ sec : shiba_andreev ] , we present a detailed study of the shiba to andreev transition as a function of the impurity strength using both analytical and numerical methods . in sec . [ sec : two_imp ] , we consider the two - impurity case , and show the presence of two subsequent quantum phase transitions as a function of the exchange coupling . finally , in sec . [ sec : concl ] , we present a summary of our results and some perspectives .
study of the heated under dc electric field nonequilibrium carriers in graphene is stimulated by attempts of realization of efficient field - effect transistor , see @xcite and refs . therein . essential heating also takes place for the cleaning of graphene by strong pulses of applied current @xcite and in recent measurements , @xcite conducted in electric fields @xmath01 kv / cm . in addition , excited by ultrafast interband pumping electron - hole pairs are actively studied @xcite . in connection with such experiments theoretical studies of energy relaxation , both analytical @xcite and numerical @xcite , were conducted as well as studies of relaxation dynamics after ultrafast photoexcitation @xcite . recently , investigation of carriers heating under dc electric field applied to an intrinsic graphene has been given in @xcite . main peculiarities of the processes of heating in graphene ( gapless and massless semiconductor , which is described by the weyl - wallace model @xcite ) are related with an interplay between quasi - elastic energy relaxation on acoustic phonons and generation - recombination processes , that are effective nearby the cross point of energy spectrum . in a gated graphene , when monopolar regime of transport is realized , contribution of generation - recombination processes is suppressed . effect of the gate voltage on the processes of carriers heating and on current - voltage characteristics of a gated graphene ( i.e. , a study of the model of field transistor located on graphene sheet of large dimensions ) have not been considered so far . in this paper a heating of carriers in doped graphene is analyzed theoretically , for the case when difference between the densities of electrons and holes is controlled by a gate voltage . present treatment is limited by the region of energies below the energy of optical phonon ( @xmath1ev ) , when it is necessary to take into account the joule heating produced by dc current , induced by electric field @xmath2 , as well as following processes of relaxation : a ) elastic relaxation of momentum due to elastic scattering on structural disorder @xcite , which is the most effective mechanism of scattering so that anisotropy of carriers distributions is small ; b ) quasi - elastic relaxation of energy on acoustic phonons ; and c ) generation - recombination processes for interband transitions caused by thermal radiation ( relaxation of energy and concentrations were analyzed for photoexcitation in @xcite ) . here we consider the regime when the interparticle coulomb scattering dominates over the processes of relaxation of the energies and the concentrations , so that the carriers are described by quasiequilibrium distributions with effective temperature and nonequilibrium concentrations of electrons and holes . these parameters are defined from the balance equations of the concentrations and the total energy , where the areal charge density of graphene sheet , given by the difference of the electron and the hole densities , is controlled by gate voltage . besides determination of the average energies of carriers and their concentrations , nonlinear current - voltage characteristics are also studied below . the analysis performed below is organized as follows . the basic equations , describing heating of carriers in gated graphene are evaluated in sec . results of calculations , including the parameters of quasiequilibrium distributions , nonequilibrium concentrations and current - voltage characteristics as a function of applied field , gate voltage , and temperature are presented in sec . discussion of the assumptions used and concluding remarks are given in sec . iv . in appendix the collision integrals used in present treatment are specified . our description of the heating of bipolar plasma in graphene under dc electric field is based on quasiclassical kinetic equations that take into account the scattering mechanisms listed above . here we assume that @xmath3 , where @xmath4 , @xmath5 , and @xmath6 ( @xmath7 ) are the relaxation frequencies of momentum , intercarrier scattering , and energy ( concentration ) , respectively . due to dominance of the relaxation of momentum , the weak anisotropic contributions to the electron ( @xmath8 ) and the hole ( @xmath9 ) distributions are given as @xcite @xmath10 here @xmath11 ( @xmath12 ) corresponds the upper ( lower ) sign and the frequency of momentum relaxation @xmath13 is expressed by means of the characteristic velocity @xmath14 ( that defines efficiency of scattering ) , and of the truncation factor for long - range scattering @xcite @xmath15 . here the correlation length of static disorder , @xmath16 , and the modified bessel function of the first kind , @xmath17 , are used . the nonequilibrium isotropic distributions @xmath18 are defined by kinetic equations @xmath19 where an overline stands for the averaging over the angle in @xmath20-plane . the collision integrals in the right hand side of eq . ( [ 2 ] ) are given in appendix . obtained from eq . ( [ 2 ] ) distributions @xmath18 define concentrations of electrons , @xmath21 , and holes , @xmath22 , as follows @xmath23 where the factor @xmath24 takes into account degeneration over the spin and the valleys . the current density , @xmath25 , is given as @xmath26 where @xmath27 is the velocity of particle with momentum @xmath28 , written using the characteristic velocity @xmath29cm / s . the sheet charge is defined by the difference of concentrations @xmath30 that is controlled by the gate voltage @xmath31 applied to the back gate , placed at a distance @xmath32 from the graphene sheet . considering such structure as plane capacitor , we obtain relation between @xmath33 and @xmath31 as follows : @xmath34 , where @xmath35 is dielectric constant of sio@xmath36 substrate . because of predominance of the intercarrier scattering in eq . ( [ 2 ] ) , the symmetric distributions @xmath18 must satisfy the following conditions @xmath37 which nullify the factor in integrand of the intercarrier collision integral ( [ a3 ] ) . hence @xmath38 imposes the quasiequlibrium distributions @xmath39^{-1 } , \label{6}\ ] ] where @xmath40 is the effective temperature of carriers , and @xmath41 are the electrochemical potentials of carriers . these three parameters are related by requirement for the area charge density to be constant and equal to @xmath42 , which gives the electroneutrality condition in the form @xmath43 further , due to conservation of concentrations of electrons and holes under the intercarrier scattering processes [ see eqs . ( [ a4 ] ) ] we obtain the concentration balance equation written through the radiative collision integral ( [ a2 ] ) @xmath44=0 . \label{8}\ ] ] here @xmath45 is the planck distribution function of the equilibrium thermal radiation with temperature @xmath46 and we have introduced the characteristic thermic momentum @xmath47 . point out , to obtain the concentration balance equation ( [ 8 ] ) we have used that : @xmath48 is not modified by scattering on acoustic phonons ( because of smallness of the velocity of sound in comparison with @xmath49 ) and @xmath50 is independent of @xmath51 . in a similar way , by summing eq . ( [ 2 ] ) over @xmath28 and @xmath51 with the weight @xmath52 , we obtain the energy balance equation : @xmath53 . here the energy losses term , @xmath54 , is written as follows : @xmath55 while the joule heating term , @xmath56 , is given as @xmath57 using the integration by parts , one can rewrite eq . ( [ 10 ] ) as @xmath58 , where the nonlinear conductivity @xmath59 is introduced by the relation @xmath60 . under substitution eq . ( 1 ) into ( 4 ) we obtain the conductivity @xmath61 , \label{11}\ ] ] where @xmath62 and @xmath63 is the linear response conductivity for the short - range scattering in intrinsic graphene . thus , the parameters of the distributions eq . ( [ 6 ] ) are defined by transcendent equations ( [ 7])-([11 ] ) . the nonequilibrium concentrations @xmath48 and the current density @xmath25 are determined by eq . ( [ 3 ] ) and eq . ( [ 4 ] ) , respectively . below we discuss the solutions of eqs . ( [ 7])-([11 ] ) as well as the concentrations and the current - voltage characteristics versus the field @xmath64 , the gate voltage @xmath31 , and the temperature @xmath46 . calculations are performed at temperature interval @xmath6577 - 300 k for the graphene on sio@xmath36 substrate of width @xmath66 cm and for the correlation length @xmath6710 nm . the quasiequilibrium distributions given by eq . ( [ 6 ] ) are determined from the balance equations , ( [ 7])-([10 ] ) , where it is convenient to use the dimensionless momentum , @xmath68 , such that @xmath69^{-1}$ ] . then the charge neutrality condition ( [ 7 ] ) obtains the form @xmath70 the concentration balance equation ( [ 8 ] ) is given as @xmath71 and it is not dependent explicitly on the field @xmath64 or gate voltage @xmath31 . multiplying the energy balance equation @xmath53 by @xmath72 we obtain its dimensionless form @xmath73 . where from eqs . ( [ 9 ] ) and ( [ 10 ] ) it follow the dimensionless energy losses @xmath74 and the dimensionless joule heating @xmath75 . \label{15}\end{aligned}\ ] ] here the characteristic momentum @xmath76 is introduced as @xmath77 and @xmath78 . for heavily doped case electrons are degenerated and holes are nondegenerated ( or vice versa ) . then the electron quasi - fermi energy , @xmath79 , is given from eq . ( [ 12 ] ) as @xmath80 which is weakly dependent of temperature . further , equation that defines @xmath40 ( e.g. , as function of @xmath64 , for given @xmath31 and @xmath46 ) is obtained from eq . ( [ 15 ] ) as @xmath81 . \label{17}\ ] ] point out , the condition of heavily doping ( in particular , @xmath82 ) we can rewrite as @xmath83 . ( a ) and maximal distribution of electrons @xmath84 ( b ) and holes @xmath85 ( c ) versus electric field for @xmath6577 k ( 1 ) , 150 k ( 2 ) , and 300 k ( 3 ) . solid , dashed , and dotted curves correspond to @xmath860 v , 1 v , and 3 v , respectively . ] ( a ) and maximal distribution of electrons @xmath84 ( b ) versus electric field for t = 77 k ( 1 ) , 150 k ( 2 ) , and 300 k ( 3 ) . solid and dashed curves correspond to @xmath8610 v and 20 v , respectively . ] in figs . 1a , 1b , and 1c we plot the effective temperature @xmath87 and the maximal distributions , @xmath84 and @xmath85 , for different doping levels at @xmath88=0 v , 1 v , and 3 v [ notice , in fig . 1(c ) the solid curves , pertinent to @xmath31=0 v , are omitted as they coincide with the solid curves of fig . it is seen from fig . 1a that , for given @xmath88 , the relative increase of the effective temperature of carriers , @xmath87 , with growing @xmath64 becomes smaller for larger @xmath46 . 1b shows that the decrease of @xmath84 with growing @xmath64 becomes , at given @xmath46 , slower for larger @xmath88 . from figs . 1c and 1b it is seen that @xmath85 , for given @xmath88 , quickly decreases as @xmath46 grows . for heavily doped case , at @xmath8910 v and 20 v , in figs . 2a and 2b we plot @xmath87 and @xmath84 as functions of @xmath64 . for these gate voltages , the hole concentrations are small , so that @xmath90 and we do not plot the dependencies @xmath91 here . the dependencies @xmath92 and @xmath84 on @xmath46 , @xmath88 , and @xmath64 now are similar to those in the low - doping region . ) versus electric field at @xmath6577 k ( 1 ) , 150 k ( 2 ) , and 300 k ( 3 ) . the solid , the dash , the dot , the dash - dot , and the dash - dot - dot curves correspond to @xmath860 v , 1 v , 3 v , 10 v , and 20 v , respectively . ] using solutions of the balance eqs . ( [ 12])-([15 ] ) in eq . ( [ 3 ] ) we plot in fig . 3a dimensionless concentrations of electrons , @xmath93 , and in fig . 3b of holes , @xmath94 . here @xmath95 and @xmath96 are equilibrium concentration of electrons and holes for @xmath97 and these concentrations are strongly dependent on @xmath31 and @xmath46 . @xcite for small concentrations ( solid and dashed curves in fig . 3a ) of electrons ( the main carriers ) their concentration is increased on tens of percents for growing @xmath64 . however , for @xmath98v the concentration of main carriers becomes very weakly growing function of @xmath64 even at @xmath99k . at the same time , the density of the minor carriers ( holes ) according to fig . 3b can be enlarged by tens of times , especially at low temperatures . , versus electric field for : @xmath65300 k ( a ) , 150 k ( b ) and 77 k ( c ) . the solid , the dash , the dot , the dash - dot , and the dash - dot - dot curves correspond to @xmath86 0 v , 1 v , 3 v , 10 v , and 20 v , respectively . ] in fig . 4 , we plot the normalized conductivity @xmath100 versus field @xmath64 for different levels of doping , dependent on gate voltage varied between @xmath101 and @xmath8920 v and on different temperatures . point out @xmath102 for each curve in figs . 4 a - c has specific value and @xmath103 increases essentially when @xmath88 grows . @xcite the field dependency of the conductivity is determined by a competition between the first and the second terms in eq . . indeed , as @xmath104 is a decreasing function of @xmath64 the integral ( second ) contribution , for @xmath105 or / and @xmath106 , is either much slower decreasing function of @xmath64 or even an increasing function of @xmath64 . in particular , for @xmath106 and @xmath107 , the second contribution is the integral of the product of @xmath108 ( that is @xmath109 for @xmath110 and drops rapidly for @xmath111 ) with the other factor that is a large and growing function of @xmath112 , @xmath113 , in actual region of @xmath112 , @xmath114 . so for low temperatures the contribution @xmath104 ( see fig . 4 b and c ) is essential and leads to the appearance of a minimum in the field dependence of @xmath59 for small concentrations , for @xmath115 3v . with the growth of electric field or the concentration ( for @xmath1163 v ) the main contribution to the conductivity comes from the integral term in eq . ( [ 11 ] ) and now @xmath59 is monotinically growing function of @xmath64 , cf . 4 a , b , and c. indeed , pertinent current - voltage characteristics show transition from a sublinear dependence to a superlinear one , as the concentration or / and the temperature grow . for @xmath1173 v ( and at @xmath65300 k for any @xmath31 , or concentration ) a superlinear current - voltage characteristic is realized . in the present work , we investigate the effect of doping on carrier heating in a gated graphene based on the balance equation approach . the effective temperature and concentrations of carriers are studied as functions of the thermostat temperature , the gate voltage , and the applied dc field . pertinent current - voltage characteristics show the transition from the regime of heating for intrinsic bipolar conductivity to the regime of heating for the conductivity of strongly degenerated carriers . in the latter case the heating is well described by simple analytical formulas , see ( 16 ) . next , we list and discuss the assumptions used in the calculations performed . here we have examined the heating of carriers in the low energy region only , when the optical phonon emission is not essential and the radiative - induced direct interband transitions are assumed to be the main generation - recombination mechanism . point out that for much larger electric fields ( for @xmath1180.1 kv / cm , see @xcite ) present approach will not be valid as the optical phonon contribution is not taken into account , however , the influence of optical phonons is smaller for a higher level of the doping ( gate voltage ) . we have also restricted ourselves by the study of limiting case when the intercarrier coulomb scattering is dominating . as shown in @xcite for the intrinsic graphene case , this is an adequate approach . possible contribution of other generation - recombination mechanisms ( note , that the auger - processes are forbidden due to the symmetry of electron - hole states @xcite ) require an additional investigation . further , we have assumed that the heat removal is sufficiently effective ( this point becomes insignificant in the case of short electric pulses when the thermostat is not overheated ) . at last , to describe the momentum relaxation we take into account only the statical disorder scattering using the phenomenological model of ref . 10 ( which is in good agreement with the experimental data ; the microscopic mechanisms of scattering are still unclear @xcite ) . the listed assumptions should not essentially modify the peculiarities of the heating mechanisms , in comparison with a study that will treat relaxation processes in more details ; in addition , here many limitations are imposed also because of the lack of data on graphene . to conclude , essential heating of carriers is shown in present study . it will define different parameters of possible graphene - based devices . in addition , investigation of hot carriers ( even within the considered limited region of parameters ) gives important information about mechanisms of relaxation and recombination in graphene . therefore , further experimental and theoretical ( including numerical modeling ) study of a heating in graphene is very opportunely now . below we present the collision integrals @xmath119 , @xmath120 and @xmath38 ( deduced in @xcite and @xcite , respectively ) that have been used to obtain the balance equations . for quasi - elastic scattering on acoustic phonons the energy relaxation is described by the fokker - planck nonlinear differential form @xmath121 \right\ } . \label{a1}\ ] ] here @xmath122 and it is introduced the frequency of quasi - elastic relaxation @xmath123 that contains the characteristic velocity @xmath124 , where for typical parameters of graphene we have @xmath125 cm / s at @xmath99k . the thermal - radiation - induced interband transitions are being depicted by the collision integrals @xmath126 , \label{a2}\ ] ] that are equal for electrons and holes . in eq . ( [ a2 ] ) it is used the frequency of radiative relaxation @xmath127 , where the characteristic velocity @xmath128cm / s for graphene imbedded between sio@xmath36 substrate and cover layer . as auger - processes are forbidden because of the electron - hole symmetry of the bands in graphene @xcite , the intercarrier collision integral receives the form @xmath129 , \nonumber \label{a3}\end{aligned}\ ] ] where it describes the transitions due to carrier - carrier scattering from the initial states ( @xmath130 ) to the states ( @xmath131 ) . here @xmath132 determines transference of the energy for scattering and @xmath133 corresponds to electron - electron or hole - hole transitions whereas the channel @xmath134 describes scattering of electrons on holes . also in eq . ( [ a3 ] ) it is carried out average over the angle and as a result the probability of transition @xmath135 is dependent from @xmath136 , @xmath137 and from the transference of the energy @xmath138 . deduction of the concentration ( [ 8 ] ) and the energy ( [ 9 ] ) balance equations , needed , along with eq . ( [ 7 ] ) , to obtain the parameters of quasiequilibrium distributions eq . ( [ 6 ] ) , is based on the property of conservation as of the concentration so of the energy density for interparticle scattering as @xmath139 these conditions straightforwardly follow from eq . ( [ a3 ] ) if take into account the symmetry of the probability of transitions under the interchange of scattering channels @xmath140 and @xmath141 . i. meric , m. y. han , a. f. yang , b. ozyilmaz , p. kim , and k. l. shepard , nature nanotech . * 3 * , 654 ( 2008 ) ; a. barreiro , m. lazzeri , j. moser , f. mauri , and a. bachtold , phys . lett . * 103 * , 076601 ( 2009 ) . j. m. dawlaty , s. shivaraman , m. chandrashekhar , f. rana , and m. g. spencer appl . . lett . * 92 * , 042116 ( 2008 ) . d. sun , z .- k . wu , c. divin , x. li , c. berger , w. a. de heer , p. n.first , and t. b. norris , phys * 101 * , 157402 ( 2008 ) . p. a. george , j. strait , j. dawlaty , s. shivaraman , mvs . chandrashekhar , f. rana , and m. g. spencer , nanoletters * 8 * , 4248 ( 2008 ) . r. w. newson , j. dean , b. schmidt , and h. m. van driel , opt . 17 , 2326 ( 2009 ) . e. h. hwang , b.y .- k . hu , and s. das sarma , phys . b * 76 * , 115434 ( 2007 ) ; w .- k . tse , s. das sarma , phys . rev . b * 79 * , 235406 ( 2009 ) ; r. bistritzer and a. h. macdonald , phys . rev . lett . * 102 * , 206410 ( 2009 ) . s. butscher , f. milde , m. hirtschulz , e. malic , and a. knorr , appl . * 91 * , 203103 ( 2007 ) ; f. rana , p. a. george , j. h. strait , j. dawlaty , s. shivaraman , mvs chandrashekhar , and m. g. spencer , phys . b * 79 * , 115447 ( 2009 ) ; p.n . romanets and f.t . vasko , phys . rev . b * 81 * , 085421 ( 2010 ) . for @xmath890 , 1 , 3 , 10 , and 20 v one obtains @xmath142 , @xmath143 , @xmath144 , @xmath145 , and @xmath146 @xmath147 at @xmath65300 k. the correspondent hole concentrations are @xmath148 , @xmath149 , @xmath150 , @xmath151 , and @xmath152 @xmath147 , i.e. , the number of holes is negligible at @xmath11710 v. at @xmath65150 and 77 k , essentially lower @xmath95 , @xmath96 are obtaned for @xmath153 v ; for @xmath154 v @xmath95 tends to an independent from @xmath46 value @xmath155 and @xmath96 tends to exponentially small value @xmath156 , where @xmath157 is a coefficient . for @xmath890 , 1 , 3 , 10 , and 20 v one obtains @xmath1581.91 , 1.97 , 2.34 , 5.07 , and 10.79 at @xmath65300 k ( fig . 4a ) , @xmath1581.15 , 1.26 , 1.73 , 4.25v , and 9.57 at @xmath65150 k ( fig . 4b ) , and @xmath1581.035 , 1.19 , 1.65 , 4.02 , and 9.24 at @xmath6577 k ( fig . 4c ) . l.a . ponomarenko , r. yang , t.m . mohiuddin , m.i . katsnelson , k.s . novoselov , s.v . morozov , a.a . zhukov , f. schedin , e.w . hill , and a. k. geim , phys . lett . * 102 * , 206603 ( 2009 ) ; s. adam , p.w . brouwer , and s. das sarma , phys . b * 79 * , 201404(r ) ( 2009 ) .	hot carriers in a doped graphene under dc electric field is described taking into account the intraband energy relaxation due to acoustic phonon scattering and the interband generation - recombination transitions caused by thermal radiation . the consideration is performed for the case when the intercarrier scattering effectively establishes the quasiequilibrium electron - hole distributions , with effective temperature and concentrations of carriers . the concentration and energy balance equations are solved taking into account an interplay between weak energy relaxation and generation - recombination processes . the nonlinear conductivity is calculated for the momentum relaxation caused by the elastic scattering . the current - voltage characteristics , and the transition between bipolar and monopolar regimes of conductivity are obtained and analyzed , for different temperatures and gate voltages .
the problem of finding a highly connected subset of vertices in a large graph arises in a number of applications across science and engineering . within social network analysis , a highly connected subset of nodes is interpreted as a community @xcite . many approaches to data clustering and dimensionality reduction construct a ` similarity graph ' over the data points . a highly connected subgraph corresponds to a cluster of similar data points @xcite . a closely related problem arises in the analysis of matrix data , e.g. in microarray data analysis . in this context , researchers are often interested in a submatrix whose entries have an average value larger ( or lower ) than the rest @xcite . such an anomalous submatrix is interpreted as evidence of association between gene expression levels and phenotypes ( e.g. medical conditions ) . if we consider the graph adjacency matrix , a highly connected subset of vertices corresponds indeed to a principal submatrix with average value larger than the background . the special case of finding a completely connected subset of vertices ( a clique ) in a graph has been intensely studied within theoretical computer science . assuming p@xmath6np , the largest clique in a graph can not be found in polynomial time . even a very rough approximation to its size is hard to find @xcite . in particular , it is hard to detect the presence of a clique of size @xmath7 in a graph with @xmath8 vertices . such hardness results motivated the study of random instances . in particular , the so - called ` planted clique ' or ` hidden clique problem ' @xcite requires to find a clique of size @xmath9 that is added ( planted ) in a random graph with edge density @xmath10 . more precisely , for a subset of vertices @xmath11 $ ] , all edges @xmath12 , with @xmath13 are present . all other edges are present independently with probability @xmath10 . such a clique can be found reliably by exhaustive search as soon as @xmath14 @xcite . however , despite many efforts , no algorithm is known that achieves this goal for @xmath15 @xcite . in other words , the problem of finding cliques of size @xmath16 is solvable , but possibly hard . proving that indeed it is computationally hard to find cliques in this regime is an outstanding problem in theoretical computer science . for general polynomial algorithms , it is known since @xcite that a clique of size @xmath17 can be found in time @xmath18 for any @xmath19 fixed . hence , if we allow any time complexity polynomial in @xmath8 , then the question is whether the planted clique can be found for @xmath20 . a more stringent computational constraint requires that the clique is found in nearly - linear time , i.e. in time of order @xmath21 . note that the number of bits required to encode an instance of the problem is of order @xmath22 , so @xmath23 is a logarithmic multiple of the time required to read an instance . dekel , gurel - gurevittch and peres @xcite developed a linear - time algorithm ( i.e. with complexity @xmath24 ) that finds the hidden clique with high probability , provided @xmath25 . in @xcite it was proved that , if @xmath26 , then there exists a message passing algorithm that finds with high probability the clique with @xmath27 operations . the same paper provided evidence that a certain class of ` local ' algorithms fails at the same threshold . among other motivations , the present paper generalizes and supports the existence of a fundamental threshold for local algorithms at least in the sparse graph setting . in the present paper , we consider the problem of finding a highly connected subset of vertices in a _ sparse graph _ , i.e. in a graph with _ bounded average degree_. in this case , the hidden set size must scale linearly with @xmath8 to obtain a non - trivial behavior . somewhat surprisingly , we find that the phase transition ` at @xmath28 ' leaves a trace also in the sparse regime . more precisely , we consider a random graph generated as follows . we select a subset of vertices @xmath29 of size @xmath30 , uniformly at random given its size . we connect any two vertices in the set independently with probability @xmath31 . any other edge is added independently with probability @xmath32 , @xmath33 . the problem distribution is therefore parametrized by @xmath34 and we will be therefore interested in the limit @xmath35 , with @xmath36 fixed . a more intuitive parametrization is obtained by replacing @xmath37 with the average degrees for vertices @xmath38 , and @xmath39 denoted , respectively , by @xmath4 , @xmath5 our main rigorous result is a sharp phase transition in the following double asymptotics : * first @xmath35 . this corresponds to considering large graphs . * then @xmath40 and @xmath41 . this corresponds to focusing on small hidden sets , but still linear in @xmath8 . the requirement @xmath41 is a necessary consequence of @xmath40 : it can be shown that otherwise the hidden set can not possibly be detected . our main rigorous result ( theorem [ thm : main ] ) establishes that , in the above double asymptotics , a phase transition takes place for local algorithms at @xmath42 namely , we consider the problem of testing whether a vertex @xmath43 is in @xmath29 or not . we say that such a test is reliable if , in the above limit , the fraction of incorrectly estimated vertices vanishes in expectation . then : * for @xmath44 , a local algorithm can estimate reliably @xmath29 in time of the order of the number of edges . this is achieved for instance , by the belief propagation algorithm . * for @xmath45 , no local algorithm can reliably reconstruct @xmath29 . analogously to the classical hidden clique problem , there is a large gap between what can be achieved by local algorithms , and optimal estimation with unbounded computational resources . proposition [ propo : exhaustive ] estabilishes that exhaustive search will find @xmath29 in exponential time , as soon as @xmath46 for some positive @xmath47 ( in the same double limit ) . note that , in both cases , a small fraction of the vertices in @xmath29 remains undetected because of the graph sparsity , for any @xmath48 . in particular , the number of vertices of degree @xmath49 is linear in @xmath8 , and such nodes can not be identified . let us finally mention the degree of a vertex @xmath43 is a poisson with mean @xmath4 if @xmath38 and mean @xmath5 if @xmath39 . hence , the degree standard deviation ( outside @xmath29 ) is @xmath50 . therefore , the ratio @xmath51 is the difference in mean degree divided by the standard deviation , and has the natural interpretation of a ` signal - to - noise ratio . ' while our rigorous analysis focuses on the limit @xmath40 and @xmath41 ( after @xmath35 ) , we will use the cavity method from spin glass theory to investigate the model behavior for _ arbitrary _ @xmath52 ( in the @xmath35 limit ) or , equivalently , arbitrary @xmath36 . we will use two approaches to obtain concrete predictions from the cavity method : * for general @xmath37 ( bounded degree ) , we derive the cavity predictions for local quantities , as well as for the free energy density . we show that this indeed coincide ( up to a shift ) with the mutual information per variable between the hidden set @xmath29 and the observed graph @xmath53 . + we use the ` population dynamics ' ( or ` sampled density evolution ' ) algorithm @xcite to solve numerically the cavity equations . * we then consider the limit of large @xmath37 ( large degree ) for arbitrary @xmath54 . in order to obtain a non - trivial limit , the signal - to - noise ratio @xmath55 $ ] is kept fixed in this limit , together with @xmath56 $ ] . + the cavity equations simplify in this limit ( the cavity field distributions become gaussian ) , and we can derive an exact phase diagram , without recourse to intensive numerical methods , cf . figure [ fig : phasediaglargedeg ] . this two approaches are complementary in that the large - degree asymptotics yields closed - form expressions . the qualitative features of the resulting phase diagram should remain unchanged at moderately small values of @xmath37 . our population dynamics analysis confirms this . as already mentioned , one of the motivations for the present work was to better understand the computational phase transition discovered in @xcite for the classical hidden clique problem . for background edge density @xmath10 , this takes place when the size of the hidden clique is @xmath57 . this phase transition can indeed be formally recovered as a dense limit of the results presented in this paper . more precisely , the phase transition for hidden cliques @xcite is captured by eq . ( [ eq : basiccondition ] ) , once we rewrite the latter in terms @xmath58 , the variance of the degrees of nodes @xmath39 . in the sparse regime , the degree is approximately poisson distributed , and hence @xmath59 . therefore eq . ( [ eq : basiccondition ] ) can be rewritten as @xmath60 for the classical ( dense ) hidden clique problem , we have @xmath61 , @xmath62 and @xmath63 , and hence we recover the condition @xmath64 . from a different perspective , the present work offers a statistical mechanics interpretation of the phase transitions in the hidden clique problem . namely , the latter can be formally recovered as the @xmath40 limit of the phase diagram in figure [ fig : phasediaglargedeg ] below . in particular , the computational phase transition at @xmath65 corresponds to a dynamical phase transition ( a spinodal point ) in the underlying statistical mechanics model . the rest of the paper is organized as follows . in the next section we define formally our model and some related notations . section [ sec : cavity ] derives the phase diagram using the cavity method . in particular , we show that the model undergoes two phase transitions as the signal - to - noise ratio increases ( for @xmath66 small enough ) : a static phase transition and a dynamic one , the two phase transitions are well separated . section [ sec : rigorous ] presents rigorous bounds on the behavior of local algorithms and exhaustive search , that match the above phase transitions for small @xmath66 . this section is self - contained and the interested reader can move directly to it , after the model definition ( some useful , but elementary results are presented in section [ sec : treeinterpretation ] ) . proofs are deferred to the appendix . finally , section [ sec : related ] positions our results in the context of recent literature . several research communities have been working on closely related problems : statistical physics , theoretical computer science , machine learning , statistics , information theory . we tried to write a paper that could be accessible to researchers with different backgrounds , both in terms of tools and of language . we apologize for any redundancy that might have followed from this approach . we use @xmath67= \{1,\dots,\ell\}$ ] to denote the set of first @xmath68 integers , and @xmath69 to denote the size ( cardinality ) of set @xmath70 . ) for a set @xmath71 , we write @xmath72 to indicate that @xmath12 runs over all unordered pairs of distinct elements in @xmath71 . for instance , for a symmetric function @xmath73 , we have @xmath74 } f(i , j ) \equiv \prod_{1\le i < j\le n } f(i , j)\ , . \ ] ] if instead @xmath75 is a set of edges over the vertex set @xmath71 ( unordered pairs with elements in @xmath71 ) we write @xmath76 to denote elements of @xmath75 . we use @xmath77 to denote the gaussian distribution with mean @xmath78 and variance @xmath79 . other classical probability distributions are denoted in a way that should be self - explanatory ( bernoulli@xmath80 , poisson@xmath81 , and so on ) . we consider a random graph @xmath82 with vertex set @xmath83 \equiv \{1,\dots , n\}$ ] and random edges generated as follows . a set @xmath84 is chosen at random . introducing the indicator variables @xmath85 we let @xmath86 independently with @xmath87 in particular @xmath88 is a binomial random variable , and is tightly concentrated around its mean @xmath89 . edges are independent given @xmath29 , with the following probability for @xmath90 distinct : @xmath91 we let @xmath92 denote the vector identifying @xmath29 . by using bayes theorem , the conditional distribution of @xmath93 given @xmath53 is easily written @xmath94}\big(\frac{\kappa}{1-\kappa}\big)^{x_i } \prod_{(i , j)\subseteq [ n ] } \big(\frac{1-a / n}{1-b / n}\big)^{x_ix_j}\prod_{(i , j)\in e } \rho_n^{x_ix_j}\ , , \label{eq : exactposterior } \ ] ] where @xmath95 . we next replace the last probability distribution with one that is equivalent as @xmath35 , and slightly more convenient for the cavity calculations of the next section . ( these simplifications will not be used to prove the rigorous bounds in section [ sec : rigorous ] . ) we first note that , as @xmath35 , we have @xmath96 with @xmath97 next , letting @xmath98 , we can rewrite the second product in eq . ( [ eq : exactposterior ] ) as @xmath74 } & \big(\frac{1-a / n}{1-b / n}\big)^{x_ix_j } = \big(\frac{1-a / n}{1-b / n}\big)^{\binom{\|x\|}{2 } } = c\ , \big(\frac{1-a / n}{1-b / n}\big)^{(2\kappa n-1)(\|x\|-\kappa n)/2 } \cdot \big(\frac{1-a / n}{1-b / n}\big)^{\frac{1}{2}(\|x\|-\kappa n)^2}\label{eq : simplificationglobal } \ ] ] with @xmath99 a constant independent of @xmath93 . notice that @xmath100 is tightly concentrated around @xmath101 . in particular @xmath102 with high probability , and therefore the last term in eq . ( [ eq : simplificationglobal ] ) is of order @xmath103 . we will neglect it , thus obtaining @xmath74 } & \big(\frac{1-a / n}{1-b / n}\big)^{x_ix_j } \approx c'\ , \big(\frac{1-a / n}{1-b / n}\big)^{\kappa n(\|x\|-\kappa n ) } \approx c '' \ , e^{-\kappa(a - b)\|x\|}\ , . \ ] ] the error incurred by neglecting the last term in eq . ( [ eq : simplificationglobal ] ) can be corrected by considering the following approximate conditional distribution of @xmath93 given the graph @xmath53 @xmath104 where @xmath105 is the indicator function on condition @xmath70 and @xmath106 note that we multiplied @xmath107 by the indicator function @xmath108 . this can be interpreted as replacing the i.i.d . bernoulli distribution ( [ eq : kappadef ] ) with the uniform distribution over @xmath29 with @xmath109 , which is immaterial as long as local properties of @xmath110 are considered . in the following , we shall compare different reconstruction methods . any such method corresponds to a function @xmath111 of vertex @xmath43 and graph @xmath53 , with the interpretation @xmath112 we characterize such a test through its rescaled success probability @xmath113 note that a trivial test ( assigning @xmath111 at random independently of @xmath53 ) achieves @xmath114 , while a perfect test has @xmath115 . we shall often omit the arguments @xmath116 , @xmath117 from @xmath118 in the following . we note in passing that the optimal estimator with respect to the metric ( [ eq : psucct ] ) is the maximum - likelihood estimator @xmath119 namely , for any other estimator @xmath116 , we have @xmath120 ( see , for instance , the textbook @xcite for a proof of this fact ) . the resulting success probability coincides with the total variation distance between the conditional distribution of @xmath53 given the two hypotheses @xmath38 and @xmath121 . recall that , given two probability measures @xmath122 on the same finite space @xmath123 , their total variation distance is defined as @xmath124 . then we have @xmath125 in this section we use the cavity method to derive an exact phase diagram of the model . it is convenient to introduce the following signal - to - noise - ratio parameter : @xmath126 using the fact that the degree outside @xmath29 is poisson with mean @xmath127 , and inside is poisson with mean @xmath128 , we also have @xmath129 we will therefore think in terms of the three independent parameters : @xmath54 ( the relative size of @xmath88 ) ; @xmath130 ( the average degree in the background ) ; @xmath131 ( the signal - to - noise ratio ) . we generically find two solutions of the cavity recursion , that possibly coincide depending on the parameters values . this correspond to two distinct phases of the statistical mechanics models , and also have a useful algorithmic interpretation , which will be spelled out in detail in section [ sec : treeinterpretation ] . initializing the recursion with the ` exact solution ' of the reconstruction problem ( ` plus ' initialization ) , we converges to a ferromagnetic fixed point . this provides an upper bound on the performance of any reconstruction algorithm . initializing the recursion with a completely oblivious initialization ( ` free ' initialization ) , we converge to a paramagnetic fixed point . this also corresponds to the performance of the best possible local algorithm ( see next section for a formal definition ) . a very similar qualitative picture is found in other inference problems on random graphs , one early example being the analysis of sparse graph codes @xcite . an important simplification is that we do not expect replica - symmetry breaking in these models @xcite . depending on the model parameters , we encounter two types of behaviors as @xmath131 increases . * for large @xmath54 or small @xmath130 , the two fixed points mentioned above coincide for all @xmath131 and no phase transition takes place . * for small @xmath54 and large @xmath130 , two phase transitions take place : a static phase transition at @xmath132 and a dynamic phase transition at a larger value @xmath133 . in addition , a spinodal point occurs at @xmath134 . + for @xmath135 the two fixed point above coincide , and yield bad reconstruction . for @xmath136 they coincide and yield good reconstruction . in the intermediate phase @xmath137 , the two fixed points do not coincide . the relevant fixed point for bayes - optimal reconstruction corresponds to the one of smaller free energy , and the transition between the two takes place at @xmath138 . the reader might consult fig . [ fig : phasediaglargedeg ] for an illustration . also , a very similar phase diagram was obtained in the related problem of sparse principal component analysis in @xcite . fixing @xmath43 , let @xmath139 ( respectively @xmath140 ) be the law of @xmath53 subject to @xmath29 containing ( respectively not containing ) vertex @xmath43 . consider the random variable @xmath141 the likelihood ratio test ( maximizing @xmath142 ) amounts to choosing . this is achieved by setting @xmath143 . ] @xmath144 as @xmath35 , the distribution of @xmath145 under @xmath139 converges to the law of a certain random variable @xmath146 , and the distribution of @xmath145 under @xmath140 converges instead to @xmath147 . the cavity method allows to write fixed point equations for these limit distributions . we omit details of the derivation since they are straightforward given the model ( [ eq : approxmodel ] ) , and since it is sufficient here to consider the replica - symmetric version of the method . general derivations can be found in ( * ? ? ? * chapter 14 ) . a closely related calculation is carried out in @xcite , which studies a more general random graph model , the so - called stochastic block model . the distribution of @xmath146 , @xmath147 are fixed point of the following recursion ( the symbol @xmath148 means that the distributions of quantities on the two sides are equal ) @xmath149 here @xmath150 are independent copies of @xmath151 . further @xmath152 , @xmath153 , @xmath154 , @xmath155 , are independent poisson random variables , independent of the @xmath156 . finally , @xmath157 and the function @xmath158 is given by @xmath159 ( recall that @xmath160 , cf . ( [ eq : rhodef ] ) . ) the cavity method predicts that the asymptotic distribution of @xmath145 ( conditional to @xmath38 or @xmath121 ) is a fixed point of eqs . ( [ eq : generalcavity1 ] ) , ( [ eq : generalcavity2 ] ) . in order to find the fixed points , we iterate these distributional equations with two types of initial conditions ( that correspond , respectively , to the poor reconstruction and good reconstruction phases ) @xmath161 we refer to section [ sec : treeinterpretation ] for the interpretation and monotonicity properties of these conditions : in particular it can be proved that @xmath151 converge in distribution if initialized in this manner . we implemented eqs . ( [ eq : generalcavity1 ] ) , ( [ eq : generalcavity2 ] ) numerically using the ` population dynamics ' method in the sampled density evolution . this was done using the method of @xcite . ] of @xcite ( also known as ` sampled density evolution ' @xcite ) . ( left ) , @xmath162 ( right ) and @xmath163 ( corresponding to average degree outside the set @xmath29 , @xmath164 ) . red curves correspond @xmath165 ( i.e. free boundary / initial conditions ) , and provide to the optimal performance of local algorithms . blue curves yield @xmath166 ( i.e. plus boundary / initial conditions ) and yield an upper bound on the performance of any algorithm . the continuous black line at @xmath167 coincides with the phase transition of bayes - optimal estimation . these curves were computed by averaging over @xmath168 runs of the population dynamics algorithm with @xmath169 samples and @xmath170 iterations.,title="fig : " ] ( left ) , @xmath162 ( right ) and @xmath163 ( corresponding to average degree outside the set @xmath29 , @xmath164 ) . red curves correspond @xmath165 ( i.e. free boundary / initial conditions ) , and provide to the optimal performance of local algorithms . blue curves yield @xmath166 ( i.e. plus boundary / initial conditions ) and yield an upper bound on the performance of any algorithm . the continuous black line at @xmath167 coincides with the phase transition of bayes - optimal estimation . these curves were computed by averaging over @xmath168 runs of the population dynamics algorithm with @xmath169 samples and @xmath170 iterations.,title="fig : " ] ( -385,23)@xmath171 ( -365,23)@xmath138 ( -345,23)@xmath172 in figure [ fig : sparsepsucc ] , we plot the predicted behavior of @xmath142 for @xmath163 and two different values of the clique size : @xmath173 . the success probability is predicted to be ( for @xmath35 ) @xmath174 we denote by @xmath175 and @xmath166 the predictions obtained with the two initializations above . as anticipated two behaviors can be observed . for @xmath54 sufficiently large , the curves @xmath165 and @xmath166 coincide for all @xmath131 . when this happens , this is also the success probability of the optimal likelihood ratio test @xmath176 , and the latter can be effectively approximated using a local algorithm ( e.g. belief propagation ) , see section [ sec : treeinterpretation ] . for @xmath54 small the two curves remain distinct in an intermediate interval of values : @xmath177 . in this regime , the asymptotic behavior of the bayes - optimal test is captured by the fixed point that yields the lowest free energy . it is convenient to define the rescaled free energy density as follows ( assuming that the limit exists ) @xmath178 the reason for this choice of the additive constants is that the resulting free energy is also equal to the asymptotic mutual information between the hidden set @xmath29 and the observed graph @xmath53 @xmath179 this quantity has therefore an immediate interpretation and several useful properties . the replica symmetric cavity method ( equivalently , bethe - peierls approximation ) predicts @xmath180 where the supremum is over all probability distributions @xmath181 over the real line satisfying the following symmetry property ( see section [ sec : treeinterpretation ] for further clarification on this property ) : @xmath182 the functional @xmath183 is defined as follows @xmath184 here expectation is taken with respect to the following independent random variables : * @xmath185 that are i.i.d . random variables with distribution @xmath186 ; * @xmath187 that are i.i.d . random variables with distribution @xmath188 ; * @xmath189 with joint distribution @xmath190 , @xmath191 , @xmath192 , where @xmath193 . * @xmath194 with the following mixture distribution . with probability @xmath54 : @xmath195 , @xmath196 . with probability @xmath197 : @xmath195 , @xmath198 . . the horizontal line corresponds to the maximal mutual information @xmath199 . , the vertical line at @xmath167 corresponds to the phase transition of the bayes - optimal estimator . this curve was computed by averaging over @xmath168 runs of the population dynamics algorithm with @xmath169 samples and @xmath170 iterations.,title="fig : " ] ( -145,23)@xmath138 let @xmath200 and @xmath201 the distributions of the fixed points obtained with plus and free initial conditions . in figure [ fig : sparsefree ] we plot the minimum of the corresponding bethe free energies @xmath202 and @xmath203 for @xmath163 , @xmath204 ( as obtained by the population dynamics algorithm ) . this is the cavity prediction for the free energy density @xmath205 . the value of @xmath131 for which @xmath206 corresponds to the phase transition point @xmath138 between paramagnetic and ferromagnetic phases from the reconstruction point of view , this is the phase transition for bayes - optimal estimation : @xmath207 notice from figure [ fig : sparsefree ] that as expected @xmath208 is monotone increasing in the signal - to - noise ratio @xmath131 , with @xmath209 as @xmath210 , and @xmath211 as @xmath212 ( here @xmath213 is the entropy of a bernoulli random variable with mean @xmath54 ) . also , the curve fig . [ fig : sparsefree ] presents some ` wiggles ' at large @xmath54 that are due to the limited numerical accuracy of the population dynamics algorithm . in the previous section we solved numerically the distributional equations ( [ eq : generalcavity1 ] ) , ( [ eq : generalcavity2 ] ) . this approach is somewhat laborious and its accuracy is limited . asymptotic expansions provide complementary analytical insights into the solution of these equations . with @xmath131 and @xmath54 fixed . here @xmath214 . left frame : success probability for free boundary condition ( equivalently , local algorithms , red curve ) , and plus boundary condition ( equivalently , general upper bound , blue curve ) . center frame : free energy ( equivalently , mutual information per vertex ) with same boundary conditions . right frame : zoom of the free energy curves.,title="fig : " ] with @xmath131 and @xmath54 fixed . here left frame : success probability for free boundary condition ( equivalently , local algorithms , red curve ) , and plus boundary condition ( equivalently , general upper bound , blue curve ) . center frame : free energy ( equivalently , mutual information per vertex ) with same boundary conditions . right frame : zoom of the free energy curves.,title="fig : " ] with @xmath131 and @xmath54 fixed . here left frame : success probability for free boundary condition ( equivalently , local algorithms , red curve ) , and plus boundary condition ( equivalently , general upper bound , blue curve ) . center frame : free energy ( equivalently , mutual information per vertex ) with same boundary conditions . right frame : zoom of the free energy curves.,title="fig : " ] ( -450,17)@xmath171 ( -423,17)@xmath138 ( -380,17)@xmath172 ( -292,17)@xmath171 ( -270,17)@xmath138 ( -233,17)@xmath172 ( -120,17)@xmath171 ( -80,17)@xmath138 here we consider @xmath215 with @xmath54 fixed , and @xmath216 converging to a limit . in particular , the signal - to - noise ratio @xmath131 is also a constant . let us emphasize once more that these limits are taken _ @xmath35 and hence the graph is still sparse . with @xmath131 and @xmath54 fixed . here we plot the ( shifted ) free energy function @xmath217 for @xmath214 and @xmath218 . comparing with figure [ fig : imu ] we see that @xmath219 , @xmath220 , @xmath221 . ] in this limit , the fixed points of eqs . ( [ eq : generalcavity1 ] ) , ( [ eq : generalcavity2 ] ) take the form @xmath222 further @xmath78 satisfies the fixed point equation @xmath223 where the function @xmath224 is defined by @xmath225 with expectation being taken with respect to @xmath226 . in other words , the distributional equations ( [ eq : generalcavity1 ] ) , ( [ eq : generalcavity2 ] ) reduced to a single nonlinear equation for the scalar @xmath78 . large @xmath78 correspond to accurate recovery . more formally , we expect the distributional solutions of eqs . ( [ eq : generalcavity1 ] ) , ( [ eq : generalcavity2 ] ) to converge to solutions of eqs . ( [ normalapprox1 ] ) to ( [ eq : cavitylargedegreelast ] ) . we do not provide a ` physicists ' derivation of this statement since this follows heuristically does not prove rigorously that the fixed points of eqs . ( [ eq : generalcavity1 ] ) , ( [ eq : generalcavity2 ] ) converge to fixed points of eqs . ( [ normalapprox1 ] ) to ( [ eq : cavitylargedegreelast ] ) . a complete proof would require controlling the convergence rate to fixed points . however in heuristic statistical physics derivation this is typically not done . also , the proof of lemma [ lemma : gaussianapprox ] follows the same strategy that would be employed in a heuristic derivation . ] from lemma [ lemma : gaussianapprox ] . the latter establishes that , iterating the cavity equations eqs . ( [ eq : generalcavity1 ] ) , ( [ eq : generalcavity2 ] ) any fixed number of times @xmath227 is equivalent ( in the large - degree limit ) to iterating eqs . ( [ normalapprox1 ] ) to ( [ eq : cavitylargedegreelast ] ) . , with @xmath228 ( relative size of the hidden set ) and @xmath229 ( signal - to - noise ratio ) fixed . the three curves are , from top to bottom @xmath230 , @xmath231 and @xmath232 . ] the free energy ( [ eq : freeenergy ] ) becomes a function of @xmath78 ( we still denote it by @xmath183 with a slight abuse of notation ) : @xmath233 where expectation is with respect to independent random variables @xmath234 and @xmath226 . its local minima are solutions of eq . ( [ eq : cavitylargedegree ] ) . equation ( [ eq : cavitylargedegree ] ) can be easily solved numerically , yielding the phase diagram in figure [ fig : phasediaglargedeg ] . as before , we obtain phase transitions @xmath235 as long as @xmath54 is below a critical point @xmath236 . the critical point location is @xmath237 the free energy @xmath238 has two local minima @xmath239 for @xmath236 , @xmath240 , and one local minimum otherwise . the local minimum @xmath241 is the global minimum for @xmath242 , while @xmath243 is the global minimum for @xmath244 . we refer to figures [ fig : imu ] and [ fig : psimu ] for illustration . of particular interest is the case of small hidden subsets , i.e. the limit @xmath40 ( note that @xmath88 is still linear in @xmath8 ) . for small @xmath54 we have @xmath245 . hence the solutions ( [ eq : cavitylargedegree ] ) that stay bounded converges to the solution of @xmath246 this equation has two solutions for @xmath247 and no solution for @xmath248 . this implies that @xmath249 which is the result announced in eq . ( [ eq : basiccondition ] ) . it is also easy to see that @xmath250 as @xmath40 . the distributional equations ( [ eq : generalcavity1 ] ) and ( [ eq : generalcavity2 ] ) define a sequence of probability distributions indexed by @xmath251 . more precisely , for every @xmath227 the recursion defines the probability distributions @xmath252 ( the distribution of @xmath253 ) and @xmath254 ( the distribution of @xmath255 ) . when specialized to the free / plus initial conditions ( cf . ( [ eq : freebc ] ) , ( [ eq : freebc ] ) ) , these probability distributions have a simple and useful interpretation that we will now explain . define @xmath256 to be the ball of radius @xmath227 centered at @xmath257 , in graph @xmath53 . namely , this is the subset of vertices of @xmath53 whose distance from @xmath43 is at most @xmath227 . by a slight abuse of notation , this will also denote the subgraph induced in @xmath53 by those vertices . the following remarks are straightforward . * free boundary condition . * consider the optimal test @xmath258 among those that only use local information . in other words , @xmath258 is the optimal test that is a function of @xmath256 . this is again a likelihood ratio test . concretely , we can define the log - likelihood ratio @xmath259 then the optimal test takes the form @xmath260)$ ] ( if we are interested in maximizing @xmath261 ) or @xmath262 ( if we are interested in minimizing the expected number of incorrectly assigned vertices ) . fixing the depth parameter @xmath227 , the distribution of @xmath263 converges ( as @xmath264 ) to @xmath265 for @xmath38 , and to @xmath266 for @xmath121 . mathematically , for any fixed @xmath43 @xmath267 in particular , for any fixed @xmath227 , the success probability @xmath268\big ) + \p_1^{(t),\fr}\big(\xi\ge \log[\kappa/(1-\kappa)]\big)-1\ , \label{eq : psuccfree } \ ] ] is the maximum asymptotic success probability achieved by any test that is @xmath227-local ( in the sense of being a function of depth-@xmath227 neighborhoods ) . it follows immediately from the definition that @xmath269 is monotone increasing in @xmath227 . its @xmath270 limit @xmath165 is the maximum success probability achieved by any local algorithm . this quantity was computed through population dynamics in the previous section , see figure [ fig : sparsepsucc ] . * plus boundary condition . * let @xmath271 be the complement of @xmath256 , i.e. the set of vertices of @xmath53 that have distance at least @xmath227 from @xmath43 . then @xmath272 has the interpretation of being the log - likelihood ratio , when information is revealed about the labels of vertices in @xmath273 . namely , if we define @xmath274 then we have @xmath275 in particular , @xmath166 is an upper bound on the performance of any estimator . in the previous section we computed this quantity numerically through population dynamics . let us finally comment on the relation ( [ eq : symmetry ] ) between @xmath186 and @xmath188 . this is an elementary consequence of bayes formula : consequences of this relation have been useful in statistical physics under the name of ` nishimori property ' @xcite . it is also known in coding theory as ` symmetry condition ' @xcite . consider the general setting of two random variables @xmath276 , with @xmath277 , @xmath278 , and let @xmath279 $ ] . then for any interval @xmath70 ( with non - zero probability ) , applying bayes formula , @xmath280 which is the claimed property . in the previous section we relied on the non - rigorous cavity method from spin glass theory to derive the phase diagram . most notably we used numerical methods , and formal large - degree asymptotics to study the distributional equations ( [ eq : generalcavity1 ] ) , ( [ eq : generalcavity2 ] ) . here we will establish rigorously some key implications of the phase diagram , namely : * by exhaustive search over all subsets of @xmath9 vertices in @xmath53 , we can estimate @xmath29 accurately for any @xmath281 and @xmath54 small . * local algorithms succeed in reconstructing accurately @xmath29 if @xmath248 , and fail for @xmath247 ( assuming large degrees and @xmath54 small ) . given a set of vertices @xmath282 $ ] , we let @xmath283 denote the number of edges with both endpoints in @xmath284 . exhaustive search maximizes this quantity among all the sets that have the ` right size . ' namely , it outputs @xmath285}\big\{\ , e(r)\ , : \|r\| = \lfloor\kappa n \rfloor\big\}\ , . \ ] ] ( if multiple maximizers exist , one of them is selected arbitrarily . ) we can also define a test function @xmath258 by letting @xmath286 for @xmath287 and @xmath288 otherwise . note that , for @xmath289 growing with @xmath117 , this algorithm is non - polynomial and hence can not be used in practice . it provides however a useful benchmark .. we have the following result showing that exhaustive search reconstructs @xmath29 accurately , for any constant @xmath131 and @xmath54 small . we refer to section [ sec : proofexhaustive ] for a proof . [ propo : exhaustive ] let @xmath290 be the asymptotic success probability of exhaustive search and assume @xmath291 . then @xmath292 in particular , we have the following large degree asymptotics as @xmath293 with @xmath294 fixed @xmath295 and @xmath296 as @xmath40 for any @xmath281 fixed . we next give a formal definition of @xmath227-local algorithms . let @xmath297 is the space of unlabeled rooted graphs , i.e. the space of graphs with one distinguished vertex ( see for instance @xcite for more details ) . formally , an estimator @xmath258 for the hidden set problem is a function @xmath298 . since the pair @xmath299 is indeed a graph with one distinguished vertex ( and the vertices labels clearly do not matter ) , we can view @xmath116 as a function on @xmath297 : @xmath300 the following definition formalizes the discussion in section [ sec : treeinterpretation ] ( where the definition of @xmath256 is also given ) . the key fact about this definition is that @xmath227 ( the ` locality radius ' ) is kept fixed , while the graph size can be arbitrarily large . given a non - negative integer @xmath227 , we say that a test @xmath116 is _ @xmath227-local _ if there exists a function @xmath301 such that , for all @xmath302 , @xmath303 we say that a test is local , if it is @xmath227-local for some fixed @xmath227 . we denote by @xmath304 and @xmath305 the sets of @xmath227-local and local tests . the next lemma is a well - known fact that we nevertheless state explicitly to formalize some of the remarks of section [ sec : treeinterpretation ] . recall that @xmath261 denotes the success probability of test @xmath116 , as per eq . ( [ eq : psucct ] ) , and let @xmath269 be defined as in eq . ( [ eq : psuccfree ] ) , with @xmath265 , @xmath266 , the laws of random variables @xmath306 , @xmath307 . we have @xmath308 in particular @xmath309 further , the maximal local success probability @xmath269 can be achieved using belief propagation with respect to the graphical model ( [ eq : approxmodel ] ) in @xmath310 time . we will therefore valuate the fundamental limits of local algorithms by analyzing the quantity @xmath165 . the following theorem establishes a phase transition for this quantity at @xmath311 . [ thm : main ] consider the hidden set problem with parameters @xmath36 , and let @xmath312 . then : * if @xmath247 , then all local algorithms have success probability uniformly bounded away from one . in particular , letting @xmath313 to be the smallest positive solution of @xmath314 , we have @xmath315 * if @xmath248 , then local algorithms can have success probability arbitrarily close to one . in particular , considering the large degree asymptotics @xmath215 with @xmath316 fixed @xmath317 we have @xmath318 as a useful technical tool in proving part @xmath319 of this theorem , we establish a normal approximation result in the spirit of eqs . ( [ normalapprox1 ] ) , ( [ normalapprox2 ] ) . in order to state this result , we recall the definition of wasserstein distance of order @xmath320 , @xmath321 between two probability measures @xmath78 , @xmath322 on @xmath323 , with finite second moment @xmath324 , @xmath325 . namely , denoting by @xmath326 the family of couplings if it is a probability distribution on @xmath327 such that @xmath328 and @xmath329 for all @xmath70 . ] of @xmath78 and @xmath322 , we have @xmath330 given a sequence of probability measures @xmath331 with finite second moment , we write @xmath332 if @xmath333 . [ lemma : gaussianapprox ] for @xmath334 , let @xmath335 be the random variables defined by the distributional recursion ( [ eq : generalcavity1 ] ) , ( [ eq : generalcavity2 ] ) , with initial condition ( [ eq : freebc ] ) , and denote by @xmath265 , @xmath266 the corresponding laws . further let @xmath336 be defined recursively by letting @xmath337 and @xmath338 then , considering the limit @xmath215 with @xmath54 fixed and @xmath339 , we have @xmath340 the proof of this lemma is presented in section [ sec : proofgaussian ] . as mentioned in the introduction , the problem of identifying a highly connected subgraph in an otherwise random graph has been studied across multiple communities . within statistical theory , arias - castro and verzelen @xcite established necessary and sufficient conditions for distinguishing a purely random graph , from one with a hidden community . with the scaling adopted in our paper , this ` hypothesis testing ' problem requires to distinguish between the following two hypotheses : @xmath341 note that this problem is trivial in the present regime and can be solved for instance by counting the number of edges in @xmath53 . the sparse graph regime studied in the present paper was also recently considered in a series of papers that analyzes community detection problems using ideas from statistical physics @xcite . the focus of these papers is on a setting whereby the graph @xmath53 contains @xmath342 non - overlapping communities , each of equal size @xmath343 . using our notation , vertices within the same community are connected with probability @xmath31 and vertices belonging to different communities are connected with probability @xmath32 . interestingly , the results of @xcite point at a similar phenomenon as the one studied here for @xmath344 . namely , for a range of parameters the community structure can be identified by exhaustive search , but low complexity algorithms appear to fail . let us mention that the very same phase transition structure arises in other inference problem , for instance in decoding sparse graph error correcting codes , or solving planted constraint satisfaction problems @xcite . a unified formalism for all of these problems is adopted in @xcite . all of these problems present a regime of model parameters whereby a large gap separates the optimal estimation accuracy , from the optimal accuracy achieved by known polynomial time algorithms . establishing that such a gap can not be closed under standard complexity - theoretic assumptions is an outstanding challenge . ( see @xcite for partial evidence in this direction albeit in a different regime . ) one can nevertheless gain useful insight by studying classes of algorithms with increasing sophistication . local algorithms : : are a natural starting point for sparse graph problems . the problem of finding a large independent set in a sparse random graph is closely related to the one studied here . indeed an independent set can be viewed as a subset of vertices that is ` less - connected ' than the background ( indeed is a subset of vertices such that the induced subgraph has no edge ) . + the largest independent set in a uniformly random regular graph with @xmath8 vertices of degree @xmath345 has typical size @xmath346 where , for large bounded degree @xmath345 , @xmath347 . hatami , lovsz and szegedy @xcite conjectured that local algorithms can find independent sets of almost maximum size up to sublinear terms in @xmath8 . gamarnik and sudan @xcite recently disproved this conjectured and demonstrated a constant multiplicative gap for local algorithms . roughly speaking , for large degrees no local algorithm can produce an independent set of size larger than @xmath348 of the optimum . this factor of @xmath348 was later improved by rahman and virag @xcite to @xmath349 . this gap is analogous to the gap in estimation error established in the present paper . we refer to @xcite for a broader review of this line of work . + as mentioned before , belief propagation ( when run for an arbitrary _ fixed _ number of iterations ) is a special type of local algorithm . further it is basically optimal ( among local algorithms ) for bayes estimation on locally tree like graphs . the gap between belief propagation decoding and optimal decoding is well studied in the context of coding @xcite . spectral algorithms . : : let @xmath350 be the adjacency matrix of the graph @xmath351 ( for simplicity we set @xmath352 for @xmath38 , and @xmath353 for @xmath121 ) . we then have @xmath354 this suggests that the principal eigenvector of @xmath355 should be localized on the set @xmath29 . indeed this approach succeeds in the dense case ( degree of order @xmath117 ) , allowing to reconstruct @xmath29 with high probability @xcite . + in the sparse graph setting considered here , the approach fails because the operator norm @xmath356 is unbounded as @xmath35 . concretely , the sparse graph @xmath351 has large eigenvalues of order @xmath357 localized on the vertices of largest degree . this point was already discussed in several related problems @xcite . several techniques have been proposed to address this problem , the crudest one being to remove high - degree vertices . + we do not expect spectral techniques to overcome the limitations of local algorithms in the present problem , even in their advanced forms that take into account degree heterogeneity . evidence for this claim is provided by studying the dense graph case , in which degree heterogeneity does not pose problems . in that case spectral techniques are known to fail for @xmath358 @xcite , and hence are strictly inferior to ( local ) message passing algorithms that succeed in the present paper correspond to @xmath359 in ( * ? ? ? * ; * ? ? ? ] for any @xmath248 . semidefinite relaxations . : : convex relaxations provide a natural class of polynomial time algorithms that are more powerful than spectral approaches . feige and krauthgamer @xcite studied the lovsz - schrijver hierarchy of semidefinite programming ( sdp ) relaxations for the hidden clique problem . in that setting , each round of the hierarchy yields a constant factor improvement in clique size , at the price of increasing complexity . it would be interesting to extend their analysis to the sparse regime . it is unclear whether sdp hierarchies are more powerful than simple local algorithms in this case . let us finally mention that the probability measure ( [ eq : approxmodel ] ) can be interpreted as the boltzmann distribution for a system of @xmath30 particles on the graph @xmath53 , with fugacity @xmath360 , and interacting attractively ( for @xmath361 ) . statistical mechanics analogies were previously exploited in @xcite . ( see also @xcite for the general community detection problem . ) i am grateful to yash deshpande for carefully reading this manuscript and providing valuable feedback . this work was partially supported by the nsf grants ccf-1319979 and dms-1106627 , and the grant afosr fa9550 - 13 - 1 - 0036 . for the sake of simplicity , we shall assume a slightly modified model whereby the hidden set @xmath29 is uniformly random with size @xmath362 , with @xmath363 . recall that , under the independent model ( [ eq : kappadef ] ) @xmath364 and hence is tightly concentrated around its mean @xmath289 . hence , the result the independent model follows by a simple conditioning argument . let @xmath365 . by exchangeability of the graph vertices , we have @xmath366 where the last inequality follows since , without loss of generality , @xmath367 . setting @xmath368 , we will prove that for any @xmath369 there exists @xmath370 such that @xmath371 the claim the follows by using the inequality ( [ eq : proofpropo - simple ] ) together with the fact that @xmath372 . for two sets @xmath70 , @xmath373 $ ] , we let @xmath374 the number of edges @xmath76 such that @xmath375 , but @xmath376 . in order to prove eq . ( [ eq : basicproofpropo ] ) note that , for @xmath377 @xmath378 to see this notice that , by definition , if @xmath379 then @xmath380 . this mean that there must exists at least one set @xmath381 $ ] satisfying the following conditions : * @xmath382 . * @xmath383 . * @xmath384 . indeed @xmath385 is such a set . this immediately implies eq . ( [ eq : tobeexplained ] ) by noticing that @xmath386 and @xmath387 . by a union bound ( setting @xmath388 ) : @xmath389 in the last inequality we used union bound and the fact that edges contributing to @xmath390 and @xmath391 are independent . using chernoff bound on the tail of binomial random variables ( with @xmath392 the kullback - leibler divergence between two bernoulli random variables ) , we get @xmath393\cap \naturals } \prob\big(\binom(m;a / n)\le j\big ) \,\,\prob\big(\binom(m;b / n)\ge j\big)\\ & \le ( m+1)\binom{k}{\ell}\binom{n - k}{k-\ell}\ , \exp\big\{-m\ , \min_{j\in [ bm / n , am / n ] } \big[d(j / m\|\| a / n)+d(j / m\|\| b / n)\big]\big\ } , . \label{eq : boundwkl } \ ] ] here , the first inequality follows because both probabilities are increasing for @xmath394 and decreasing for @xmath395 . we further note that , @xmath396 and therefore , for @xmath397 $ ] , @xmath398 this implies that , for @xmath399 , we have @xmath400 } \big[d(x\|\|p_1)+d(x\|\|p_2)\big]&\ge \frac{1}{2}\big(\frac{1}{p_2}+1\big)\ , \min_{x\in [ p_1,p_2 ] } \big[(x - p_1)^2+(x - p_2)^2\big]\\ & \ge \frac{1}{4}\big(\frac{1}{p_2}+1\big)\ , ( p_1-p_2)^2\ , . \ ] ] we substitute the last inequality in eq . ( [ eq : boundwkl ] ) , together with the bounds @xmath401 $ ] @xmath402 we let @xmath403 whence @xmath404 we therefore get @xmath405 for @xmath406 , the argument in parenthesis is smaller than @xmath407 and therefore @xmath408 summing over @xmath409 , we get @xmath410 which implies the claim ( [ eq : basicproofpropo ] ) , after eventually adjusting @xmath411 , since @xmath412 . recall that convergence in @xmath415 distance is equivalent to weak convergence , plus convergence of the first two moments ( * ? ? ? * theorem 6.9 ) . we will prove by the following by induction over @xmath227 : these claims obviously hold for @xmath423 . next assuming that they hold up to iteration @xmath227 , we need to prove them for iteration @xmath424 . for the sake of brevity , we will only present this calculation for @xmath425 , since the derivation for @xmath426 is completely analogous . let us start by considering eq . ( [ eq : limitexpxi0 ] ) . first notice that the absolute value of right - hand side of eq . ( [ eq : generalcavity1 ] ) is upped bounded by @xmath427 and hence @xmath428 follows from the induction hypothesis i@xmath429 and the fact that @xmath430 are poisson . next to prove eq . ( [ eq : limitexpxi0 ] ) , we take expectation of eq . ( [ eq : generalcavity1 ] ) , and let , for simplicity , @xmath431 : @xmath432 where the last equality follows from bounded convergence , since , for all @xmath433 , @xmath434 . note that the laws of @xmath253 and @xmath255 satisfy the symmetry property ( [ eq : symmetry ] ) . hence , for any measurable function @xmath435 such that the expectations below make sense , we have @xmath436 in particular applying this identity to @xmath437 and @xmath438 ^ 2 $ ] , we get @xmath439 substituting in eq . ( [ eq : exiexpansion ] ) , and expressing @xmath440 in terms of @xmath441 we get @xmath442 where @xmath443 denotes a quantity vanishing as @xmath444 . the last equality follows from induction hypothesis iii@xmath429 and the fact that @xmath445 is bounded continuous , with @xmath226 . this yields the desired claim ( [ eq : limitexpxi0 ] ) after comparing with eq . ( [ eq : mutrecursion ] ) . consider next eq . ( [ eq : limitvarxi0 ] ) . the upper bound on the right - hand side of eq . ( [ eq : generalcavity1 ] ) given by eq . ( [ eq : upperboundrhs ] ) immediately imply that @xmath446 . in order to estabilish eq . ( [ eq : limitvarxi0 ] ) , we recall an elementary formula for the variance of a poisson sum . if @xmath447 is a poisson random variable and @xmath448 are i.i.d . with finite second moment , then @xmath449 applying this to eq . ( [ eq : generalcavity1 ] ) , and expanding for large @xmath130 thanks to the bounded convergence theorem , we get @xmath450 ^ 2\right\}+\kappa b \ , \e\left\{\big[\log\left(1+(\rho-1)\frac{e^{\xi_1^{(t ) } } } { 1+e^{\xi_1^{(t)}}}\right)\big]^2\right\}\\ & = ( 1-\kappa ) \frac{(a - b)^2}{b } \e\left\{\left(\frac{e^{\xi_0^{(t)}}}{1+e^{\xi_0^{(t)}}}\right)^2\right\ } + \kappa\ , \frac{(a - b)^2}{b}\e\left\{\left(\frac{e^{\xi_1^{(t)}}}{1+e^{\xi_1^{(t)}}}\right)^2\right\ } + o(b^{-1/2})\\ & = \kappa \frac{(a - b)^2}{b}\e\left(\frac{e^{\xi_1^{(t)}}}{1+e^{\xi_1^{(t)}}}\right ) + o(b^{-1/2})\ , , \ ] ] where the last equality follows by applying again eq . ( [ eq : secondidentity ] ) . by using the induction hypothesis iii@xmath319 and the fact that @xmath451 is bounded lipschitz , @xmath452 which is eq . ( [ eq : mutrecursion ] ) . we finally consider eq . ( [ eq : limitweakp0 ] ) . by subtracting the mean , we can rewrite eq . ( [ eq : generalcavity1 ] ) as @xmath453 where @xmath454 , @xmath455 . note that @xmath456 , @xmath457 have zero mean and , by the calculation above , they have variance @xmath458 . denoting the right hand side by @xmath459 : @xmath460 because ( for instance ) @xmath461 is a sum of order @xmath462 independent random variables with zero mean and variance of order @xmath463 . note that @xmath464 + \kappa b\e [ f(\xi^{(t)}_{1,1})^2 ] \big\ } = \mu^{(t+1)}\ , , \ ] ] where the last equality follows by the calculation above . hence , by applying the central limit theorem to each of the four terms in eq . ( [ eq : sb ] ) and noting that they are independent , we conclude that @xmath465 converges in distribution to @xmath466 . define the event @xmath468 , and write @xmath414 for @xmath469 . from eq . ( [ eq : psuccfree ] ) we have @xmath470 using eq . ( [ eq : symmetry ] ) , and the fact that @xmath471 , we get @xmath472 call @xmath473 . by the initialization ( [ eq : freebc ] ) , @xmath474 . taking exponential moments of eq . ( [ eq : generalcavity1 ] ) , we get @xmath475 + \kappa b \,\e\left[\left(\frac{1+\rho\ , e^{\xi_1^{(t ) } } } { 1 + e^{\xi_1^{(t ) } } } \right)^2\right ] \right\}\ , . \ ] ] note that by eq . ( [ eq : symmetry ] ) , for any measurable function @xmath435 such that the expectations below make sense , we have @xmath476 applying this to @xmath477 , we get @xmath478 \right\}\ , . \ ] ] now we claim that , for @xmath479 , we have @xmath480 this can be checked , for instance , by multiplying both sides by @xmath481 and simplifying . using @xmath482 and @xmath483 , we get @xmath484 let @xmath485 be the solution of the above recursion with equality , i.e. @xmath486 and @xmath487 it is a straightforward exercise to see that @xmath485 is monotone increasing in @xmath227 and @xmath131 . further , for @xmath488 , @xmath489 the smallest positive solution of @xmath490 , and @xmath491 . hence @xmath492 which , together with eq . ( [ eq : basictvbound ] ) finishes the proof . note that by monotonicity @xmath493 , and hence it is sufficient to lower bound the limit of the latter quantity . by lemma [ lemma : gaussianapprox ] , we have @xmath494 where @xmath495 is the gaussian distribution , and @xmath336 is defined recursively by eq . ( [ eq : mutrecursion ] ) with @xmath337 . hence for all @xmath334 @xmath496 it is therefore sufficient to prove that @xmath497 now by monotone convergence , we have @xmath498 further @xmath499 increases monotonically towards its limit as @xmath40 . furthermore , @xmath500 is increasing in @xmath78 for any fixed @xmath501 . by induction over @xmath227 we prove that @xmath502 ( the limit being monotone from below ) , where @xmath503 and for all @xmath334 @xmath504 in order to prove this claim , note that the base case of the induction is trivial and ( writing explicitly the dependence on @xmath54 @xmath505 on the other hand for a fixed @xmath506 @xmath507 the claim follows since @xmath508 can be taken arbitrarily small . noga alon , michael krivelevich , and benny sudakov , _ finding a large hidden clique in a random graph _ , proceedings of the ninth annual acm - siam symposium on discrete algorithms , society for industrial and applied mathematics , 1998 , pp . 594598 . emmanuel abbe and andrea montanari , _ conditional random fields , planted constraint satisfaction and entropy concentration _ , approximation , randomization , and combinatorial optimization . algorithms and techniques , springer , 2013 , pp . 332346 . dimitris achlioptas and federico ricci - tersenghi , _ on the solution - space geometry of random constraint satisfaction problems _ , proceedings of the thirty - eighth annual acm symposium on theory of computing , acm , 2006 , pp . 130139 . aurelien decelle , florent krzakala , cristopher moore , and lenka zdeborov , _ asymptotic analysis of the stochastic block model for modular networks and its algorithmic applications _ , physical review e * 84 * ( 2011 ) , no . 6 , 066106 . alexandre gaudillire , benedetto scoppola , elisabetta scoppola , and massimiliano viale , _ phase transitions for the cavity approach to the clique problem on random graphs _ , journal of statistical physics * 145 * ( 2011 ) , no . 5 , 11271155 . dandan hu , peter ronhovde , and zohar nussinov , _ phase transitions in random potts systems and the community detection problem : spin - glass type and dynamic perspectives _ , philosophical magazine * 92 * ( 2012 ) , no . 4 , 406445 . subhash khot , _ improved inapproximability results for maxclique , chromatic number and approximate graph coloring _ , foundations of computer science , 2001 . proceedings . 42nd ieee symposium on , ieee , 2001 , pp . 600609 . florent krzakala , cristopher moore , elchanan mossel , joe neeman , allan sly , lenka zdeborov , and pan zhang , _ spectral redemption in clustering sparse networks _ , proceedings of the national academy of sciences * 110 * ( 2013 ) , no . 52 , 2093520940 . andrea montanari , daniel reichman , and ofer zeitouni , _ on the limitation of spectral methods : from the gaussian hidden clique problem to rank one perturbations of gaussian tensors _ , arxiv:1411.6149 ( 2014 ) .	we consider a random sparse graph with bounded average degree , in which a subset of vertices has higher connectivity than the background . in particular , the average degree inside this subset of vertices is larger than outside ( but still bounded ) . given a realization of such graph , we aim at identifying the hidden subset of vertices . this can be regarded as a model for the problem of finding a tightly knitted community in a social network , or a cluster in a relational dataset . in this paper we present two sets of contributions : @xmath0 we use the cavity method from spin glass theory to derive an exact phase diagram for the reconstruction problem . in particular , as the difference in edge probability increases , the problem undergoes two phase transitions , a static phase transition and a dynamic one . @xmath1 we establish rigorous bounds on the dynamic phase transition and prove that , above a certain threshold , a local algorithm ( belief propagation ) correctly identify most of the hidden set . below the same threshold _ no local algorithm _ can achieve this goal . however , in this regime the subset can be identified by exhaustive search . for small hidden sets and large average degree , the phase transition for local algorithms takes an intriguingly simple form . local algorithms succeed with high probability for @xmath2 and fail for @xmath3 ( with @xmath4 , @xmath5 the average degrees inside and outside the community ) . we argue that spectral algorithms are also ineffective in the latter regime . it is an open problem whether any polynomial time algorithms might succeed for @xmath3 .
improved numerical tools and the increasing power of modern supercomputers have brought considerable progress in modeling stellar core collapse in the past years . it is possible now to simulate the complex physical processes in the deep interior of supernovae with unprecedented sophistication and detailedness . it has become clear meanwhile that the explosions of massive stars are a generically multi - dimensional phenomenon . this insight was fostered by the fact that spherically symmetric ( 1d ) simulations , which became available with a fully energy dependent solution of the boltzmann transport problem for the neutrinos only recently ( see @xcite for an overview and comparison of different numerical approaches ) confirmed and solidified older 1d results of the 1980 s and 1990 s , namely that explosions in the 1d models could not be obtained , neither by the prompt bounce - shock nor by the delayed neutrino - heating mechanism , at least not for progenitor stars of more than 10@xmath1 and on a timescale of roughly one second after core bounce @xcite . moreover , the latest generation of multi - dimensional simulations has provided evidence for a variety of routes that can lead to explosions when nonradial phenomena are accounted for . these routes seem to depend on the properties and conditions in the progenitor stars like their mass and structure and the amount of angular momentum in their core . in the following we will briefly review these recent developments in the multi - dimensional modeling of stellar core collapse and explosion , and we will critically discuss the status of the present simulations . a better understanding of the explosion mechanism of core - collapse supernovae is not only important for interpreting the observable properties of the blast , for predicting gravitational - wave and neutrino signals , and for determining the conditions of nucleosynthesis processes that occur during the explosion . it is also and in particular essential for establishing the link between the progenitor stars and their compact remnants , thus answering questions like that of the mass distribution of neutron stars and of the stellar mass limit for black hole formation , which may happen either directly during the core collapse or by later massive fallback when the disrupted star does not become completely unbound during the explosion . so far , estimates for such scenarios have been made only on the basis of still rather crude self - consistent explosion simulations @xcite or by invoking assumptions about the mass cut and energy in models with piston - driven artificial explosions ( e.g. , @xcite ) . due to the huge gravitational binding energy liberated in neutrinos , which carry away hundred times more energy than needed for the explosion , these particles have long been speculated to be the driving agent of the stellar explosion . colgate and white @xcite in a seminal paper in 1966 not only proposed gravitational binding energy to be the primary energy source of core - collapse supernovae , but also that the intense flux of escaping neutrinos transfers the energy from the imploding core to the ejected stellar mantle . nearly twenty years later , bethe and wilson @xcite were the first who described in detail the way how this might happen , interpeting thereby the physics that played a role in hydrodynamic simulations performed by wilson . they concluded that electron neutrino and antineutrino absorptions on the free neutrons and protons that are abundantly present in the shock - dissociated matter behind the stalled accretion shock in the supernova core , are the primary agents of the energy transfer . these pioneering computer simulations of the so - called delayed neutrino - driven explosion mechanism were still conducted in spherical symmetry . the mechanism turned out to be successful only when the neutron star was assumed to become a more luminous neutrino source by mixing instabilities accelerating the energy transport out of its dense interior . the thus enhanced neutrino emission led to stronger neutrino heating in the overlying layers of the exploding star . theoretical studies and multi - dimensional computer models , however , suggest that convection and mixing instabilities inside the neutron star do not have the necessary big effect ( see , e.g.@xcite ) . instead , the first multi - dimensional simulations , which became available only in the mid 1990 s , demonstrated that the neutrino - heated layer around the forming neutron star is unstable to vigorous convective overturn @xcite . this can raise the efficiency of the neutrino energy deposition and thus can have a supportive effect on the supernova explosion . the first such two - dimensional ( i.e. axisymmetric ) and three - dimensional simulations , however , suffered from a severe drawback : the physics of the neutrino transport and of neutrino - matter interactions , which is essential for discussing the power input to the explosion , is so complex that it could be treated only in a grossly simplified way . in the best models at that time this was done by the so - called `` grey diffusion approximation '' . this means that the energy - dependence of the neutrino interactions ( the cross sections of the most important neutrino processes typically scale with the squared neutrino energy ) was ignored and replaced by a `` grey '' ( spectrally averaged ) description . moreover , the spatial propagation was approximated by assuming that neutrinos diffuse through the dense neutron star medium until they decouple and stream away from a chosen position , usually from a layer somewhat outside of the `` neutrinosphere '' , close to the surface of the compact remnant . the historical development of these theoretical studies of the supernova explosion mechanism is resumed in a recent review @xcite . [ cols="<,^ , > " , ] recently , burrows et al . @xcite came up with the suggestion that supernovae might be energized by a strong flux of acoustic power originating from the neutron star . in their 2d simulations they found that the compact remnant is instigated to large - amplitude bipolar oscillations , @xmath2 core gravity modes , by the anisotropic accretion of gas . the pulsating compact remnant sends pressure waves into its environment , which carry a sizable energy flux and can even steepen into shocks , thus dissipating their energy in the surrounding medium and raising the entropy there . the ringing neutron star acts as a transducer that converts some part of the gravitational binding energy released by the accreted gas into sonic power . the radiated sound was found to be the crucial supply of the developing explosion with energy and momentum and thus triggers acoustically driven explosions . while this appears as an interesting alternative to initiate the explosion if neutrino heating fails , the question is how the acoustic energy input compares to neutrino heating . burrows et al . @xcite reported that in their simulations the acoustic energy flux dominates the neutrino energy deposition later than several 100@xmath3ms after bounce . although we do not observe the large @xmath2 core g - modes in the neutron star seen by them at late post - bounce times , and on the basis of our present simulations we can neither judge nor exclude this possibility , also in our models in the full - scale supernova calculations for @xmath0815@xmath1 stars as well as in the parametric explosions we can clearly identify the presence of strong turbulence in the neutron star surface layer . the gas there is stirred by the violent impact of accretion downflows , thus creating vigorous sonic activity around the neutron star ( figs . [ janka_fig : waves1 ] and [ janka_fig : waves2 ] ) . very approximately , we can estimate the outgoing flux of sonic energy ( making the assumptions of spherical symmetry and negligible energy dissipation in the flow ) as @xcite @xmath4 where @xmath5 is the fluid velocity in the sound waves , @xmath6 is the sound speed , and @xmath7 denotes the amplitude of the ripples on the background density @xmath8 caused by the sound waves . the numerical value in the last expression turns out to be fairly independent of radius and time at some distance from the neutron star in figs . [ janka_fig : waves1 ] and [ janka_fig : waves2 ] . for @xmath7 of order unity our estimate agrees with values quoted in burrows et al . @xcite , which might account for a sizable contribution to the energy of the developing explosion . in our simulations , we indeed see that the perturbations can reach such amplitudes , in particular when the downflow becomes unstable and the perturbations it generates in and around the neutron star surface layer grow . outside of these transient periods , however , @xmath7 is usually significantly smaller . moreover , the net neutrino energy deposition rates , integrated over the volume of the gain layer , in all of our simulations are typically several @xmath9erg / s ( see eq . [ eq : dotenu ] ) , so that the energy input to the explosions in our models is clearly dominated by neutrinos . magnetohydrodynamically driven explosions have attracted renewed and increasing interest over the past years for a number of reasons one reason is the discovery of magnetars , which is taken as an indication that at least in some fraction of all cases very strong surface magnetic fields can be generated in neutron stars , possibly already during the evolution phases in which a supernova explosion is triggered . another reason is the established link of at least some long - duration gamma - ray bursts with very energetic and highly aymmetric hypernova explosions of presumably rapidly rotating massive stars . and a third reason is the interpretation of asymmetries observed in supernova explosions and supernova remnants as consequences and relics of jet - like eruptions or as hints to the ejection of highly collimated material . independent of how magnetic fields transfer energy to the explosion , e.g. , by hoop stresses , magnetic pressure , or viscous dissipation of energy , the initial magnetic fields in the stellar core must have been amplified during the collapse either by magnetic field wrapping or the magnetorotational instability . thus they tap the free energy that can be stored in the highly differential rotation of a spinning stellar core that collapses at angular momentum conserving conditions . this free energy , however , is typically a rather small fraction ( of order 10% ) of the total rotation energy , i.e. , @xmath10 where @xmath11 denotes the neutron star radius in @xmath12 cm and @xmath13 is the neutron star spin period . a millisecond neutron star typically requires the pre - collapse stellar core to rotate with a period of @xmath14 . this is significantly faster than predicted by current evolution models for rotating stars , in which the enhanced angular momentum loss due to angular momentum transport by magnetic field effects is taken into account @xcite . at the onset of core collapse , the stellar cores are estimated to have spin periods of @xmath15s ( i.e. , @xmath16rad / s ) . this will lead to neutron stars with @xmath17ms , which is much too slow for rotation to be an energy reservoir of mhd - driven supernovae . such a conclusion was also reached on the basis of detailed simulations in refs . nevertheless , the mhd mechansim may still need to be invoked for explaining the enormous energy output of long - duration gamma - ray bursts and associated hypernova explosions , which would imply that these events are linked to rare cases where massive stars have achieved to retain a large angular momentum at the time when they end their lives as collapsars @xcite . we have reviewed a variety of recent findings that shed new light on the processes that cause the explosions of massive stars and play a role during the birth of neutron stars . all recent 2d simulations that were performed with a full 180 degree grid agree that the standing accretion shock becomes unstable to low - mode , nonradial deformation . this sasi phenomenon plays a very important role in the supernova core . it was not only found to induce a large asymmetry of the developing blast but also to facilitate neutrino - driven explosions by stretching the time accreted matter can stay in the gain layer and can be exposed to neutrino heating . moreover , the sasi was observed to excite large - amplitude core g - modes in the neutron star , whose sonic damping could contribute to or be essential for powering the explosion . the asymmetries imprinted on the explosion by the sasi may lateron lead to neutron star kicks and might explain the observed velocities of young pulsars . in three dimensions the @xmath18 modes of the sasi can also have an influence on the spin of the forming neutron star @xcite . despite the general agreement about the importance of the sasi , the physical mechanism behind this phenomenon is still a matter of vivid debate ( see @xcite and references therein ) and the present calculations differ in many conclusions . partly this is may be so because of the significantly different numerical approaches taken , e.g. , regarding the hydrodynamics and neutrino transport , the computational grid , the description of gravity , and the employed equation of state and neutrino interactions . some of the current discrepancies and controversies , however , will find an explanation when more simulations become available and comparisons are made . ultimately , however , most of the processes and consequences mentioned above will have to be addressed by 3d models , which are currently still out of reach because of the enormous computational demands of the energy - dependent neutrino transport . we thank k. nomoto , a. heger and s. woosley for providing us with their progenitor data and are indebted to r. buras , k. kifonidis , e. mller , and m. rampp for fruitful collaborations . this project was supported by the deutsche forschungsgemeinschaft through the transregional collaborative research centers sfb / tr 27 `` neutrinos and beyond '' and sfb / tr 7 `` gravitational wave astronomy '' , and the cluster of excellence `` origin and structure of the universe '' ( http://www.universe-cluster.de ) . the computations were only possible because of computer time on the ibm p690 of the john von neumann institute for computing ( nic ) in jlich , on the national supercomputer nec sx-8 at the high performance computing center stuttgart ( hlrs ) under grant number supern/12758 , on the ibm p690 of the computer center garching ( rzg ) , on the sgi altix 4700 of the leibniz - rechenzentrum ( lrz ) in munich , and on the sgi altix 3700 of the mpi for astrophysics . we also acknowledge support by astrogrid - d , a project funded by the german federal ministry of education and research ( bmbf ) as part of the d - grid initiative . bruenn , c.j . dirk , a. mezzacappa , et al . , `` modeling core collapse supernovae in 2 and 3 dimensions with spectral neutrino transport , '' in _ scidac 2006 , scientific discovery through advanced computing _ , edited by w.m . tang , et al . 46 , 2006 , pp . 393402 ; arxiv0709.0537 .	we report here on recent progress in understanding the birth conditions of neutron stars and the way how supernovae explode . more sophisticated numerical models have led to the discovery of new phenomena in the supernova core , for example a generic hydrodynamic instability of the stagnant supernova shock against low - mode nonradial deformation and the excitation of gravity - wave activity in the surface and core of the nascent neutron star . both can have supportive or decisive influence on the inauguration of the explosion , the former by improving the conditions for energy deposition by neutrino heating in the postshock gas , the latter by supplying the developing blast with a flux of acoustic power that adds to the energy transfer by neutrinos . while recent two - dimensional models suggest that the neutrino - driven mechanism may be viable for stars from @xmath08@xmath1 to at least 15@xmath1 , acoustic energy input has been advocated as an alternative if neutrino heating fails . magnetohydrodynamic effects constitute another way to trigger explosions in connection with the collapse of sufficiently rapidly rotating stellar cores , perhaps linked to the birth of magnetars . the global explosion asymmetries seen in the recent simulations offer an explanation of even the highest measured kick velocities of young neutron stars . address = max planck institute for astrophysics , karl - schwarzschild - str . 1 , d-85741 garching , germany address = max planck institute for astrophysics , karl - schwarzschild - str . 1 , d-85741 garching , germany address = max planck institute for astrophysics , karl - schwarzschild - str . 1 , d-85741 garching , germany address = max planck institute for astrophysics , karl - schwarzschild - str . 1 , d-85741 garching , germany
the study of electron - proton ( @xmath0 ) collisions at hera has improved our understanding of the structure of the proton as well as about the dynamics of the strong interactions at high energies ( for a review see e.g. ref . in particular , the study of diffractive processes has been one of the most successful areas at hera , with vector meson production and deeply virtual compton scattering ( dvcs ) in exclusive processes ( @xmath4 , with @xmath5 ) being important probes of the transition between the soft and hard regimes of qcd . these processes have been the subject of intensive theoretical and experimental investigations , with one of the main motivations for these studies being the possibility to probe the qcd dynamics at high energies , driven by the gluon content of the proton which is strongly subject to non - linear effects ( parton saturation ) @xcite . an important lesson from the analysis of the hera data at small values of the bjorken - @xmath6 variable is that the inclusive and diffractive processes can be satisfactorily described using a unified framework the color dipole formalism . this approach was proposed many years ago in ref . @xcite and considers that the high energy photon can be described by a color quark - antiquark dipole and that the interaction of the dipole with the target can be described by the color dipole cross section @xmath7 , with the transverse size of the dipole @xmath8 frozen during the interaction process . in this approach all information about the target and strong interaction physics is encoded in @xmath7 , which is determined by the imaginary part of the forward amplitude of the scattering between a small dipole ( a colorless quark - antiquark pair ) and a dense hadron target , denoted by @xmath9 , where the dipole has transverse size given by the vector @xmath10 , with @xmath11 and @xmath12 being the transverse vectors of the quark and antiquark , respectively , and @xmath13 is the impact parameter . in the color glass condensate ( cgc ) formalism @xcite , @xmath14 contains all the information about non - linear and quantum effects in the hadron wave function . it can be obtained by solving an appropriate evolution equation in the rapidity @xmath15 , which in its simplest form is the balitsky - kovchegov ( bk ) equation @xcite . alternatively , the scattering amplitude can be obtained using phenomenological models based on saturation physics constructed taking into account the analytical solutions of the bk equation which are known in the low and high density regimes . as demonstrated in @xcite , the combination between the color dipole formalism and saturation physics are quite successful to describe the recent and very precise hera data on the reduced inclusive cross section as well as the data on the exclusive processes in a large range of photon - proton center - of - mass energies @xmath16 , photon virtualities @xmath17 and @xmath6 values . hera has also provided high precision experimental data on semi - inclusive @xmath18 processes , where the incident proton is converted into a neutron via a charge exchange @xcite . very recently the first measurements of exclusive @xmath1 photoproduction associated with leading neutrons ( @xmath19 ) were presented @xcite . the description of these leading neutron processes is still a theoretical challenge . in particular , the @xmath20 ( feynman momentum ) distribution of leading neutrons remains without a conclusive theoretical description @xcite . in ref . @xcite we extended the color dipole formalism to leading neutron processes and demonstrated that the experimental data on the semi - inclusive reactions can be well described by this approach . our goal in this paper is to further extend our previous analysis to exclusive processes and try to show that the color dipole formalism may also provide a unified description of leading neutron processes . using the same assumptions made in ref . @xcite , we compare our predictions with the hera data on @xmath1 exclusive photoproduction and estimate the contribution of the absorption corrections to exclusive leading neutron processes . taking into account these corrections we present predictions for the exclusive @xmath2 , @xmath3 and @xmath21 production associated with a leading neutron in @xmath0 collisions at the energies of hera and future electron - proton colliders . this paper is organized as follows . in the next section we present a brief review of leading neutron production in @xmath0 collisions and we discuss the treatment of exclusive processes with the color dipole formalism . in section [ results ] we analyse the dependence of our predictions on the models of the vector meson wave function , on the pion flux and on the dipole scattering amplitude . a comparison with the recent hera data on exclusive @xmath1 photoproduction is performed and predictions for the exclusive @xmath2 , @xmath3 and @xmath21 production associated to a leading neutron in @xmath0 collisions for the energies of hera and of the future electron - proton colliders are presented . finally , in section [ conc ] we summarize our main conclusions . [ cols="^,^ " , ] finally , in order to estimate the photon - pion cross section we must specify the dipole - pion scattering amplitude @xmath22 . as considered in ref . @xcite for the semi - inclusive processes , we will assume the validity of the approximation expressed by eq . ( [ doister ] ) , with the dipole proton scattering amplitude @xmath23 being given by the bcgc model , proposed in ref . @xcite and recently updated in ref . @xcite , which is based on the cgc formalism and takes into account the impact parameter dependence of the dipole - proton scattering amplitude . as demonstrated in refs . @xcite , this model is able to describe the vector meson production in @xmath0 and ultra peripheral hadronic collisions . in the bcgc model the dipole - proton scattering amplitude is given by @xcite @xmath24 with @xmath25 , where @xmath26 is the lo bfkl characteristic function . the coefficients @xmath27 and @xmath28 are determined uniquely from the condition that @xmath29 , and its derivative with respect to @xmath30 , are continuous at @xmath31 . in this model , the proton saturation scale @xmath32 depends on the impact parameter : @xmath33^{\frac{1}{2\gamma_s}}. \label{newqs}\ ] ] the parameter @xmath34 was adjusted to give a good description of the @xmath35-dependence of exclusive @xmath36 photoproduction . moreover , the factors @xmath37 and @xmath38 were taken to be free . the set of parameters which will be used here is the following : @xmath39 , @xmath40 , @xmath41 gev@xmath42 , @xmath43 , @xmath44 and @xmath45 . moreover , in order to estimate the dependence of our predictions on the choice of the model for @xmath23 , we also will consider the iims @xcite and gbw @xcite models , as well as the numerical solution of the bk equation obtained in ref . @xcite . such models were discussed in detail in ref . @xcite . for these models , we assume @xmath46 and @xmath47 , with the normalization of the dipole cross section ( @xmath48 ) being fitted to data , and that the @xmath49 - dependence of the photon - pion cross section can be approximated by an exponential ansatz , @xmath50 , with the slope being given by @xmath51 . it is important to emphasize that the conclusions obtained in @xcite are not modified if the bcgc model is used as input in the calculations . before discussing our results , a comment is in order . as in the semi - inclusive case , our predictions for the exclusive processes associated with a leading neutron are essentially parameter free , depending only on the choices of the models for the pion flux and on the dipole scattering amplitude . the main uncertainties are associated with the choice of @xmath52 ( in eq . ( [ doister ] ) ) and the magnitude of the absorption effects which can arise by soft rescatterings . these latter are difficult to calculate @xcite but are expected to modify almost uniformly all the @xmath20 spectrum of the leading neutrons . as in ref . @xcite , in what follows we will assume that these effects can be mimicked by a factor @xmath53 , which multiplies the right side of eq . ( [ crossgen ] ) changing the normalization of the spectra and which should be estimated from the analysis of experimental data . in spite of the efforts made in several studies of absorptive corrections in semi - inclusive processes @xcite , the magnitude of these effects in exclusive processes remains an open question . photoproduction considering two different models for the vector meson wave function ( boosted gaussian and gauss - lc ) and two different models for the pion flux ( @xmath54 and @xmath55).,title="fig : " ] photoproduction considering two different models for the vector meson wave function ( boosted gaussian and gauss - lc ) and two different models for the pion flux ( @xmath54 and @xmath55).,title="fig : " ] photoproduction considering different models of the pion flux . data from ref . ] photoproduction considering different models for the dipole scattering amplitude and of the pion flux . data from ref . @xcite.,title="fig : " ] photoproduction considering different models for the dipole scattering amplitude and of the pion flux . data from ref . @xcite.,title="fig : " ] let us start our analysis considering the exclusive @xmath1 photoproduction associated with leading neutrons as analysed by the h1 collaboration @xcite . in what follows we will assume that @xmath56 gev , @xmath57 gev@xmath58 and that @xmath59 gev , where @xmath60 is the transverse momentum of the leading neutron . moreover , we will assume initially that @xmath61 and the bcgc dipole model . in fig . [ fig:2 ] we analyse the dependence of our predictions on the choice of the vector meson wave function . we present our results for two different models of the pion flux . we find that the predictions are similar , with the gauss - lc results being a lower bound . this conclusion is also valid for other models of the pion flux and for the @xmath2 and @xmath3 production . consequently , in what follows we will consider only the gauss - lc model for the vector meson wave functions . let us now compare our predictions with the experimental data @xcite considering different models for the pion flux . in order to constrain the value of the @xmath53 factor associated to absorptive corrections , our strategy will be following : for a given model of the pion flux , @xmath52 and dipole cross section , we will estimate the total cross section . the value of @xmath53 will be the value necessary to make our prediction consistent with the h1 data @xcite . in fig . [ fig:3 ] we present our predictions for the leading neutron spectra in exclusive @xmath1 photoproduction considering different models for the pion flux . the corresponding @xmath53 values are also presented . we obtain a reasonable agreement with the experimental data , with @xmath53 values in the range @xmath62 . it is important to emphasize that these values of @xmath53 are strongly correlated with our choice for @xmath52 . for example , if instead of @xmath61 we assume @xmath63 , the corresponding @xmath53 values should be multiplied by 4 , since the exclusive cross sections depend quadratically on the dipole scattering amplitude . if we assume _ a priori _ that the magnitude of the absorptive correction factor is of the order of @xmath64 for exclusive processes , as predicted in @xcite for semi - inclusive one , this implies a preference for the value @xmath63 . however , as the magnitude of these corrections for exclusive processes is still an open question , as well as the value of @xmath52 , we refrain from drawing strong conclusions . therefore , in what follows we will only present results assuming @xmath61 , but the reader should keep in mind the quadratic correlation between @xmath53 and @xmath52 , implying that the same fits could be obtained with much bigger values of @xmath53 and smaller values of @xmath52 . with more data on different processes with leading neutron production , it may be possible to disentangle @xmath53 from @xmath52 . it is interesting to notice that leading neutron production is dominated by pion emission from the proton , i.e. @xmath65 . in all existing theoretical approaches this pion is soft and takes only a small fraction of the incoming proton energy , leaving the neutron with most of it . this is the physical reason for the peak seen at @xmath66 in the @xmath20 spectrum of the leading neutron . in fig . [ fig:4 ] we analyse the dependence of our predictions on the choice of the dipole scattering amplitude for two different models of the pion flux . as done before , we will constrain the value of @xmath53 by adjusting the predictions of the different dipole models to the experimental value of the total cross section . we find that the different predictions for the @xmath20 spectra are very similar . however , the effective value of the absorptive correction @xmath53 depends on the model of the dipole scattering amplitude as expected , since they predict different values for the @xmath28 slope , which determines the normalization of the photon - pion cross section . figs . [ fig:3 ] and [ fig:4 ] are in a sense complementary , since what changes in the former ( the flux factor ) is kept fixed in the latter ( where the dipole model is changed ) and vice - versa . in each curve the overall constant @xmath67 is chosen so as to bring our calculations as close as possible to the experimental points . comparing the curves we can conclude that the shape of the leading neutron spectrum is much more sensitive to the flux factor than to the dipole scattering amplitude @xmath68 . the normalization of the spectrum is hence determined by @xmath53 and @xmath52 , since @xmath68 is fixed from the analysis of other data . the values used for @xmath53 are significantly smaller than those found in theoretical estimates . larger values of @xmath53 would be more plausible implying a deviation from the valence quark scaling and the consequent change in the factor @xmath69 . a comparison between the curves in fig . [ fig:4 ] favors the choice of the pion flux @xmath55 , which , unlike @xmath54 , contains a t - dependent form factor . a similar preference was found in ref . @xcite , where a combined analysis of e866 and hera data was performed . in what follows we will only consider the bcgc model , which successfully describes the hera data on exclusive processes . in fig . [ fig:5 ] we present our predictions taking into account the experimental uncertainty present in the h1 data for the total cross section @xcite , which in our analysis translates into a range of possible values for the @xmath53 factor . these results indicate that the experimental data are better described using the pion flux @xmath70 . as a cross check of our results , we can compare our predictions with other h1 data obtained assuming @xmath71 gev . assuming the same range of values for @xmath53 obtained in fig . [ fig:5 ] we can see in fig . [ fig:6 ] that our predictions describe these data quite well . photoproduction obtained considering the possible range of values of the @xmath53 factor and two models for the pion flux . data from ref . @xcite.,title="fig : " ] photoproduction obtained considering the possible range of values of the @xmath53 factor and two models for the pion flux . data from ref . @xcite.,title="fig : " ] photoproduction obtained considering the possible range of values of the @xmath53 factor fixed using the other set of experimental data and two models for the pion flux . data @xcite obtained assuming that @xmath71 gev.,title="fig : " ] photoproduction obtained considering the possible range of values of the @xmath53 factor fixed using the other set of experimental data and two models for the pion flux . data @xcite obtained assuming that @xmath71 gev.,title="fig : " ] , @xmath3 and dvcs production in the hera kinematical range : @xmath56 gev and @xmath59 gev . , title="fig : " ] , @xmath3 and dvcs production in the hera kinematical range : @xmath56 gev and @xmath59 gev . , title="fig : " ] , @xmath3 and dvcs production in the hera kinematical range : @xmath56 gev and @xmath59 gev . , title="fig : " ] , @xmath2 , @xmath3 and dvcs production in the kinematical range of the future @xmath0 colliders ( @xmath72 gev@xmath58 ) . , title="fig : " ] , @xmath2 , @xmath3 and dvcs production in the kinematical range of the future @xmath0 colliders ( @xmath72 gev@xmath58 ) . , title="fig : " ] , @xmath2 , @xmath3 and dvcs production in the kinematical range of the future @xmath0 colliders ( @xmath72 gev@xmath58 ) . , title="fig : " ] , @xmath2 , @xmath3 and dvcs production in the kinematical range of the future @xmath0 colliders ( @xmath72 gev@xmath58 ) . , title="fig : " ] , @xmath2 , @xmath3 and dvcs production in the kinematical range of the future @xmath0 colliders ( @xmath73 tev).,title="fig : " ] , @xmath2 , @xmath3 and dvcs production in the kinematical range of the future @xmath0 colliders ( @xmath73 tev).,title="fig : " ] , @xmath2 , @xmath3 and dvcs production in the kinematical range of the future @xmath0 colliders ( @xmath73 tev).,title="fig : " ] , @xmath2 , @xmath3 and dvcs production in the kinematical range of the future @xmath0 colliders ( @xmath73 tev).,title="fig : " ] considering that the main inputs of our calculations have been fixed by the experimental data on exclusive @xmath1 photoproduction we can extend our analysis to other exclusive final states . we will assume the gauss - lc model for vector meson wave function , the bcgc dipole scattering amplitude , the @xmath55 model for the pion flux and the same @xmath53 values needed to describe the @xmath1 data . initially , let us consider the kinematical range probed by hera . as in the @xmath1 case , we will assume @xmath56 gev and @xmath59 gev . however , for @xmath2 and @xmath3 production we assume @xmath57 gev@xmath58 , while for the dvcs we consider that @xmath74 gev@xmath58 . the corresponding predictions for the leading neutron spectra in exclusive @xmath2 and @xmath3 production as well as in dvcs are presented in fig . [ fig:7 ] . for the hera kinematical range we predict @xmath75 nb , @xmath76 nb and @xmath77 nb , where the uncertainty is estimated taking into account the range of possible values of @xmath53 . finally , let us present our predictions for the kinematical range which may be probed in future @xmath0 colliders assuming @xmath59 gev . in figs . [ fig:8 ] and [ fig:9 ] we show our results for the energy and photon virtualities , respectively . as expected , the leading neutron spectra increases with the energy at fixed @xmath17 and decreases with the virtuality at fixed @xmath16 . in particular , for @xmath73 tev and @xmath72 gev@xmath58 we predict @xmath78 nb , @xmath79 nb , @xmath80 nb and @xmath81 nb . we believe that for these values of total cross sections , the experimental analysis of the exclusive processes associated with a leading neutron is feasible in future @xmath0 colliders . in particular , as the cross sections strongly increase when @xmath82 , the analysis of the vector meson photoproduction in @xmath0 collisions can be useful to understand leading neutron spectra , which are of crucial importance in particle production in cosmic ray physics . another possibility is the study of this process in ultraperipheral hadronic collisions , with the leading neutron being a tag for exclusive production . in principle these processes can be studied in the future at the lhc . such proposition will be discussed in detail in a forthcoming publication . one of the important goals in particle physics is to understand the production of leading particles , i.e. the production of baryons which have large fractional longitudinal momentum ( @xmath83 ) and the same valence quarks ( or at least one of them ) as the incoming particles . recent measurements of leading neutron spectra in @xmath0 collisions at hera have shed a new light on this subject . however , the description of the semi - inclusive and exclusive leading neutron processes remains without a satisfactory theoretical description . in a previous work @xcite , we proposed to study semi - inclusive leading neutron production using the color dipole formalism , which successfully describes both inclusive and diffractive hera data , taking into account the qcd dynamics and its non - linear effects , which are expected to be present at high energies . making use of very simple assumptions about the relation between the dipole - pion and the dipole - proton scattering amplitudes and about the absorptive corrections , we demonstrated that the semi - inclusive data can be described by the dipole formalism and that feynman scaling is expected at high energies . in this paper we have extended our analysis to exclusive processes associated with a leading neutron . considering the same assumptions used for the semi - inclusive case , we have analysed in detail the dependence of our predictions on the choices of the vector meson wave function , of the dipole model and and of the pion flux . we demonstrated that the hera data on the exclusive @xmath1 photoproduction associated with a leading neutron can be quite well described by the color dipole formalism . assuming the validity of this approach , we have presented for the first time predictions for the exclusive @xmath2 , @xmath3 and @xmath21 production in @xmath0 collisions for the energies of hera and future colliders . our results indicate that the experimental analysis of these processes is feasible and that they can be used to understand this long standing problem in high energy physics . finally , it is important to emphasize that the current sources of uncertainties in the computation of leading neutron spectra are : i ) the strength of the absorptive corrections represented by the factor @xmath53 ; ii ) the validity of the additive quark model for the photon - pion cross section ; iii ) the strength of the contribution from direct fragmentation of the proton into neutrons ; iv ) the precise form of the pion flux ; v ) the precise form of the dipole cross section . with sufficient experimental information , we can rule out candidates of the pion flux and of the photon - pion cross section . we believe that it is possible to constrain the unknown numbers and assumptions with the help of more experimental data on other processes with tagged leading neutrons , such as those on @xmath84 production @xcite and those with dijet production @xcite . work along this line is in progress . f. gelis , e. iancu , j. jalilian - marian and r. venugopalan , ann . nucl . part . sci . * 60 * , 463 ( 2010);e . iancu and r. venugopalan , arxiv : hep - ph/0303204 ; h. weigert , prog . part . phys . * 55 * , 461 ( 2005 ) ; j. jalilian - marian and y. v. kovchegov , prog . phys . * 56 * , 104 ( 2006 ) ; j. l. albacete and c. marquet , prog . * 76 * , 1 ( 2014 ) . j. jalilian - marian , a. kovner , l. mclerran and h. weigert , phys . d * 55 * , 5414 ( 1997 ) ; j. jalilian - marian , a. kovner and h. weigert , phys . d * 59 * , 014014 ( 1999 ) , _ ibid . _ * 59 * , 014015 ( 1999 ) , _ ibid . _ * 59 * 034007 ( 1999 ) ; a. kovner , j. guilherme milhano and h. weigert , phys d * 62 * , 114005 ( 2000 ) ; h. weigert , nucl . phys . * a703 * , 823 ( 2002 ) ; e. iancu , a. leonidov and l. mclerran , nucl.phys . * a692 * , 583 ( 2001 ) ; e. ferreiro , e. iancu , a. leonidov and l. mclerran , nucl . * a701 * , 489 ( 2002 ) . j. olsson [ h1 collaboration ] , pos dis * 2014 * , 156 ( 2014 ) ; v. andreev _ et al . _ [ h1 collaboration ] , eur . j. c * 74 * , 2915 ( 2014 ) ; f. d. aaron _ et al . _ [ h1 collaboration ] , eur . j. c * 68 * , 381 ( 2010 ) . frankfurt , l. mankiewicz and m.i . strikman , z. phys . a * 334 * , 343 ( 1989 ) ; t.t . chou and c.n . yang , phys . d * 50 * , 590 ( 1994 ) ; k. golec - biernat , j. kwiecinski and a. szczurek , phys . d * 56 * , 3955 ( 1997 ) ; m. przybycien , a. szczurek and g. ingelman , z. phys . c * 74 * , 509 ( 1997 ) ; w. melnitchouk , j. speth and a.w . thomas , phys . d * 59 * , 014033 ( 1998 ) ; a. szczurek , n.n . nikolaev and j. speth , phys . b * 428 * , 383 ( 1998 ) . see for example : f. carvalho , f. o. duraes , f. s. navarra , m. nielsen and f. m. steffens , eur . j. c * 18 * , 127 ( 2000 ) ; s. paiva , m. nielsen , f. s. navarra , f. o. duraes and l. l. barz , mod . a * 13 * , 2715 ( 1998 ) ; w. melnitchouk and a.w . thomas , phys . b * 414 * , 134 ( 1997 ) . v. p. goncalves and m. v. t. machado , phys . c * 84 * , 011902 ( 2011 ) ; v. p. goncalves , b. d. moreira and f. s. navarra , phys . c * 90 * , 015203 ( 2014 ) ; v. p. goncalves , b. d. moreira and f. s. navarra , phys . b * 742 * , 172 ( 2015 ) .	in this paper we extend the color dipole formalism to the study of exclusive processes associated with a leading neutron in @xmath0 collisions at high energies . the exclusive @xmath1 , @xmath2 and @xmath3 production , as well as the deeply virtual compton scattering , are analysed assuming a diffractive interaction between the color dipole and the pion emitted by the incident proton . we compare our predictions with the hera data on @xmath1 production and estimate the magnitude of the absorption corrections . we show that the color dipole formalism is able to describe the current data . finally , we present our estimate for the exclusive cross sections which can be studied at hera and in future electron - proton colliders . lu tp 15 - 55 + december 2015
a superconformal algebra in four dimensions is specified by the number of supercharges , @xmath8 . theories with @xmath9 have been studied extensively in the last few decades , and played essential roles in the development of field theories . @xmath0 theories , however , had not attracted much attention until quite recently . a reason for this is that the only @xmath0 multiplet of free fields is the vector multiplet , which has the same contents as the @xmath1 vector multiplet , and if we try to make an @xmath0 theory in a perturbative way we always end up with an @xmath1 theory . this means if there exist genuine @xmath0 theories they should necessarily be strongly coupled . this is consistent to the fact that the absence of marginal deformations in @xmath0 theories @xcite . an explicit construction of @xmath0 theories was recently proposed by garca - etxebarria and regalado in @xcite theories by using ads / cft correspondence . ] . they realize @xmath0 theories on a stack of d3-branes by performing a @xmath10 orbifolding that project out one of four supercharges on the d3-branes . such an orbifold can be defined by combining @xmath11 symmetry of type iib supergravity @xcite and a rotation of the transverse space @xmath12 . the @xmath11 is broken by the quantization of @xmath13 string charges down to a discrete subgroup depending on the value of the complex field @xmath14 . for generic @xmath15 the subgroup is @xmath16 and then the orbifolded theory is still @xmath1 , while for some special values @xmath10 with @xmath17 are unbroken and then we can define genuine @xmath0 theories . the resulting orbifolds , which we call s - folds following @xcite , are labelled by two integers @xmath18 and @xmath19 ; @xmath18 is the order of the orbifold group and @xmath19 is a divisor of @xmath18 related to the discrete torsion . therefore , the theory on the d3-branes are labeled by @xmath18 , @xmath19 , and @xmath20 , where @xmath20 is the number of mobile d3-branes . aharony and tachikawa @xcite listed the dimensions of the coulomb branch operators in these theories . in particular , they contain an operator with dimension @xmath21 , which becomes @xmath22 or @xmath23 for some of the theories . aharony and evtikhiev showed in @xcite that the dimensions of coulomb branch operators in a genuine @xmath0 theory must be equal to or greater than @xmath24 . this was confirmed for rank-@xmath22 @xmath0 theories in @xcite . this fact implies that the @xmath0 theories with @xmath25 or @xmath23 are in fact @xmath1 theories @xcite , which can be specified by giving the gauge group @xmath26 . in @xcite some bps states were identified with @xmath27 bps w - bosons , and the gauge groups were determined ( table [ n4list ] ) . .theories in which @xmath0 is enhanced to @xmath1 are shown . [ cols="^,^",options="header " , ] substituting ( [ ppqq2z ] ) into ( [ qfqs2 ] ) we obtain @xmath28,\quad q_2^{(o)}=\frac{1+i}{2}[(p_1-q_2)+i(p_2+q_1)].\end{aligned}\ ] ] these are similar to ( [ qfforso5 ] ) and only difference up to unimportant signatures are the extra factors @xmath29 . therefore , the quantization and constraint for @xmath30 can be obtained from the previous ones by simply rotate the lattice in figure [ lat ] by @xmath31 degrees and expand it by the factor @xmath32 . the resulting lattice is the same as figure [ ox ] . if we set @xmath33 the charges @xmath30 take the form @xmath34 , and @xmath35 can be expressed as @xmath36 this is the constraint ( [ attachable ] ) for o3@xmath37-plane . we can also set @xmath38 . then @xmath39 can be expressed as @xmath40 this is the constraint ( [ attachable ] ) for @xmath41-plane . these two assignments of @xmath42 give two @xmath43 theories with different @xmath44-angle , which are dual to each other . in both cases , the spectrum of the central charge @xmath2 in the s - fold matches completely with that of the orientifold . once we obtain these relations , we can rewrite @xmath3 as a function of the s - fold variables . for the @xmath7 matching with ( [ so5zy ] ) and ( [ qfqs1 ] ) and the @xmath43 matching with ( [ sp2zy ] ) and ( [ qfqs2 ] ) we obtain @xmath45 the overall sign depends on the choice of string orientation and is the matter of conventions . this result is the same as ( [ z3olz ] ) up to sign . as in the @xmath46 case let us illustrate the junctions corresponding to strings in s - fold . strings connecting two independent d3-branes give w - bosons and they have been already analyzed above , and here we focus on strings connecting a d3-brane and one of its mirror branes . first we consider two strings @xmath47 these are shown in figure [ z4wall.eps ] ( a ) in a situation with @xmath48 . the dyonic charges of these strings are @xmath49 by the relation ( [ qfqs1 ] ) we obtain the charges of the corresponding junctions : @xmath50 junctions carrying these charges are shown in figure [ z4wall.eps ] ( b ) . as in figure [ wall.eps ] the red and blue junctions in figure [ z4wall.eps ] ( b ) are holomorphic and anti - holomorphic , respectively . another pair of two strings in s - fold is @xmath51 we can easily check that these strings carry dyonic charges that are as @xmath29 times as those in ( [ z4juncq1q2 ] ) . the corresponding junctions are ones in figure [ z4wall.eps ] ( b ) with charges multiplied by @xmath29 . for @xmath53 we expect @xmath1 sym with the gauge group @xmath54 . the discrete torsion group for @xmath53 is trivial , and an arbitrary pair of the dyonic charges @xmath55 is allowed . the central charge @xmath2 is given by @xmath56 although no perturbative realization of @xmath54 theory is known in string theory and we can not directly read off the other central charge @xmath3 from junctions , it is natural to guess from the results for @xmath57 that @xmath58 up to overall phase . indeed , if we assume * @xmath2 is given by ( [ zqzz ] ) , * @xmath3 is a holomorphic bilinear form of @xmath59 and @xmath60 . namely , @xmath3 is given by @xmath61 with some coefficients @xmath62 , and * @xmath63 for charges satisfying the electric condition ( [ electric ] ) , then ( [ olzqz ] ) is the unique solution up to phase ambiguity . in the previous sections we show that there are two types of @xmath64 bps junctions : holomorphic and anti - holomorphic ones . they may decay or make the transition between two types of bps states as we move d3-branes . in this section we demonstrate how this happens for some junctions . the purpose of this section is not to give a comprehensive analysis of general junctions but simply to show typical processes that occur for simple junctions when we move d3-branes . thus , we consider only strings in s - folds that are shown in figure [ wall.eps ] and figure [ z4wall.eps ] . let us consider @xmath4 case first . as a starting point we take the red string and junction in figure [ wall.eps ] , which are also shown in figure [ moduli.eps ] . case are shown . @xmath65 , @xmath66 , and the origin are fixed and @xmath67 and @xmath68 moves . walls , regions , and a typical example of wall crossing are illustrated . ] as we previously mentioned the junction is holomorphic . let @xmath69 be the intersection point of the junction . if @xmath69 coincide @xmath70 , @xmath68 , or @xmath66 , the junction decays into a pair of strings . let us first consider the wall with @xmath71 . because the angles at the junction point are all @xmath72 , @xmath73 is equivalent to @xmath74 using ( [ z1z2byy1y2 ] ) we can rewrite this as the following relation between @xmath67 and @xmath65 : @xmath75 which means @xmath67 is an internally dividing point of @xmath76 and @xmath77 . if we illustrate relative position of d3-branes in the s - fold so that the position @xmath65 is fixed , this gives a wall of the transition on the @xmath67-plane . the other two walls @xmath78 and @xmath79 are given respectively by @xmath80 these three walls form the triangle with vertices at @xmath65 , @xmath76 , and @xmath77 , which is shown as purple dashed lines in figure [ moduli.eps ] ( a ) . the corresponding walls in the flat background are shown in ( b ) by the same color of dashed lines . when @xmath68(@xmath67 ) crosses these walls the junction decays into two strings ( see ( b-2 ) in figure [ moduli.eps ] ) . let us continue the deformation . when @xmath81 all d3-branes in the flat background are aligned on a line , and then the relation @xmath82 holds , which means d3-branes in the s - fold are on a circle . ( see orange dashed lines and @xmath83 in figure [ moduli.eps ] . ) nothing special happens on this locus . as @xmath67 goes further away from the origin , finally , it crosses the walls defined by @xmath84 where @xmath85 is , again , real positive . these walls are shown as green dashed lines in figure [ moduli.eps ] . note that these equations also represent that @xmath65 is on a side of the triangle @xmath67 , @xmath86 , and @xmath87 . after the crossing the pair of strings form an anti - holomorphic junction . ( see ( b-5 ) in figure [ moduli.eps ] . ) having illustrated the walls , it is clear that the gray regions in figure [ moduli.eps ] are regions where the junctions can not be sustained , hence , they are regions of non - bps states . the other regions are ones of holomorphic or anti - holomorphic bps states . let us move on to @xmath5 case as another example . we look into the red string and junction illustrated in figure [ z4wall.eps ] . there are two different points compared to @xmath4 case . one is that the flat background in @xmath4 case is replaced by the orientifold . the other is that the angles at the junction point are not all @xmath72 but @xmath88 , @xmath89 , and @xmath89 . by taking these into account we can easily determine the walls , and the results are illustrated in figure [ z4moduli.eps]-pronged junctions in the orientifold . ] . s - fold is shown . @xmath65 , @xmath66 , and the origin are fixed and @xmath67 and @xmath68 moves . in the shaded regions the junction becomes non - bps . the d3-brane motion shown by dashed arrows cause a deformation of the junction similar to what shown in figure [ moduli.eps ] . some string configurations in the motion along solid arrows are shown on the right . ] let us follow a typical motion of d3-branes . we move the d3-brane at @xmath67 with @xmath65 kept intact . if we move @xmath67 outward as shown in figure [ z4moduli.eps ] ( a ) by the dashed arrow , the deformation of the corresponding junction is quite similar to that of @xmath4 case , and we do not repeat the analysis for it . we comment on another type of wall that we do not meet in the @xmath4 case . it is illustrated as light - blue lines in figure [ z4moduli.eps ] . let us move d3-branes following the solid arrows in the figure . in ( b ) we show both the junction and its mirror image by the reason which will become clear shortly . on the light - blue line each of the junction and its mirror becomes a pair of strings as in ( b-2 ) . if the brane pass through the light - blue line two strings ending on @xmath68 and @xmath90 form a new junction . although the topology of the junction changes in this process the junction keeps itself holomorphic . in this paper we investigated junctions in s - folds for the purpose of understanding the supersymmetry enhancement proposed by aharony and tachikawa . we determined the spectrum of dyonic charges of the theory from the information of junctions in the s - fold , and in the @xmath46 and @xmath91 cases we confirmed that the spectra are the same as those of @xmath1 @xmath6 and @xmath7 gauge theories by using another perturbative brane realization of each theory . this agreement seems highly non - trivial , and gives a strong support to the realization of @xmath1 supersymmetry . there are many unsolved problems . what is the most desired would be a direct determination of the non - perturbative central charges @xmath3 . due to the lack of the information of @xmath3 we could not directly determine the masses of bps states . when we established the correspondence of the spectrum of junctions and @xmath1 dyonic particles , we use the information of one of the central charge @xmath2 , which can be seen perturbatively in the s - fold . however , to determine bps saturating masses , we also need the other central charge @xmath3 . because this central charge is generated non - perturbatively together with the fourth supercharge in the supersymmetry enhancement , we can not determine it by simply drawing the shapes of junctions . there is another problem related to the marginal deformation . in this paper we showed that the spectrum of s - fold side gives dyonic spectrum of @xmath1 theory with a particular value of the marginal deformation parameter @xmath15 . however , if @xmath1 is realized , we should be able to freely change the deformation parameter . the change of the parameter affects the central charges , and we should have the corresponding parameter on the s - fold side . the identification of the parameter in the s - fold is very important problem to understand the supersymmetry enhancement . even if we could obtain the central charges , it would be another problem to determine the bps spectrum . the existence of bps saturating states is highly non - trivial problem and we need to perform quantum analysis of the junctions to determine it . as far as we know this is open problem even for orientifolds . in section [ wall.sec ] we studied wall crossing for simple string configurations . we determined walls by using another brane realization of the theory which is expected to be equivalent to the s - fold system . it is still unknown what happens on the s - fold side when the wall is crossed . we believe any of these questions is important to understand dynamics of @xmath0 theories and s - folds . we hope to return to these problems in future . we would like to thank tetsuji kimura for valuable discussions . the work of yi is partially supported by grand - in - aid for scientific research ( c ) ( no.15k05044 ) , ministry of education , science and culture , japan . dy is supported by the erc starting grant n. 304806 , `` the gauge / gravity duality and geometry in string theory '' . 99 o. aharony and m. evtikhiev , `` on four dimensional n=3 superconformal theories , '' arxiv:1512.03524 [ hep - th ] . c. cordova , t. t. dumitrescu and k. intriligator , `` deformations of superconformal theories , '' arxiv:1602.01217 [ hep - th ] . i. garca - etxebarria and d. regalado , `` n=3 four dimensional field theories , '' arxiv:1512.06434 [ hep - th ] . s. ferrara , m. porrati and a. zaffaroni , `` n=6 supergravity on ads(5 ) and the su(2,2/3 ) superconformal correspondence , '' lett . phys . * 47 * , 255 ( 1999 ) [ hep - th/9810063 ] . j. h. schwarz , `` covariant field equations of chiral n=2 d=10 supergravity , '' nucl . phys . b * 226 * , 269 ( 1983 ) . j. h. schwarz and p. c. west , `` symmetries and transformations of chiral n=2 d=10 supergravity , '' phys . b * 126 * , 301 ( 1983 ) . o. aharony and y. tachikawa , `` s - folds and 4d n=3 superconformal field theories , '' arxiv:1602.08638 [ hep - th ] . t. nishinaka and y. tachikawa , `` on 4d rank - one n=3 superconformal field theories , '' arxiv:1602.01503 [ hep - th ] . p. c. argyres , m. lotito , y. l and m. martone , `` expanding the landscape of @xmath92=2 rank 1 scfts , '' arxiv:1602.02764 [ hep - th ] . j. h. schwarz , `` lectures on superstring and m theory dualities : given at ictp spring school and at tasi summer school , '' nucl . * 55b * , 1 ( 1997 ) [ hep - th/9607201 ] . k. dasgupta and s. mukhi , `` bps nature of three string junctions , '' phys . b * 423 * ( 1998 ) 261 [ hep - th/9711094 ] . a. sen , `` string network , '' jhep * 9803 * ( 1998 ) 005 [ hep - th/9711130 ] . s. j. rey and j. t. yee , `` bps dynamics of triple ( p , q ) string junction , '' nucl . b * 526 * ( 1998 ) 229 [ hep - th/9711202 ] . o. bergman , `` three pronged strings and 1/4 bps states in n=4 superyang - mills theory , '' nucl . phys . b * 525 * ( 1998 ) 104 [ hep - th/9712211 ] . e. witten , `` baryons and branes in anti - de sitter space , '' jhep * 9807 * , 006 ( 1998 ) [ hep - th/9805112 ] . y. imamura and s. yokoyama , `` superconformal index of n=3 orientifold theories , '' arxiv:1603.00851 [ hep - th ] . y. hyakutake , y. imamura and s. sugimoto , `` orientifold planes , type i wilson lines and nonbps d - branes , '' jhep * 0008 * , 043 ( 2000 ) [ hep - th/0007012 ] . c. fraser and t. j. hollowood , `` semiclassical quantization in n=4 supersymmetric yang - mills theory and duality , '' phys . lett . b * 402 * , 106 ( 1997 ) [ hep - th/9704011 ] .	we study supersymmetry enhancement from @xmath0 to @xmath1 proposed by aharony and tachikawa by using string junctions in s - folds . the central charges carried by junctions play a central role in our analysis . we consider planer junctions in a specific plane . before the s - folding they carry two complex central charges , which we denote by @xmath2 and @xmath3 . the s - fold projection eliminates @xmath3 as well as one of the four supercharges , and when the supersymmetry is enhanced @xmath3 should be reproduced by some non - perturbative mechanism . for the models of @xmath4 and @xmath5 s - folds which are expected to give @xmath6 and @xmath7 @xmath1 theories we compare the junction spectra with those in perturbative brane realization of the same theories . we establish one - to - one correspondence so that @xmath2 coincides . by using the correspondence we also give an expression for the enhanced central charge @xmath3 .
the discovery of superfluidity in helium ii ( @xmath0he and @xmath1he ) has aroused interest in boson liquids . this was recently renewed and intensified due to the experimental obtaining of bose einstein condensates ( becs ) in alkali metals . especially interesting is the formation and time evolution of both curved and straight vortices which appear in these media @xcite . in two preceding papers @xcite , referred to as part i and part ii , the time evolution of a pair of oppositely polarized linear vortices was examined . in part i , it was demonstrated that if such a pair is appropriately perturbed , a persistent set of identical circular vortices can be created . this was in full agreement with feynman s hypothesis @xcite mentioned in the title . in part ii , examples were given for the creation of several _ different _ circular vortices from one linear pair . in both cases , the bec was examined within the commonly used gross pitaevskii model @xcite . the authors checked that all circular vortices created during the evolution satisfied known relations between their velocities and radii @xcite . here , in part iii , we examine the interaction of circular vortices with either linear ones , or else with other circular vortices of different radii , again within the gross pitaevskii model . the results are compared with those of other authors @xcite@xcite . in the present paper , as in parts i and ii , we describe a one - component bec by a single particle wavefunction of @xmath2 bosons of mass @xmath3 that obeys the nonlinear schrdinger ( nls ) type equation as formulated by gross @xcite and pitaevskii @xcite , @xmath4 where positive @xmath5 is the strength of the assumed @xmath6-function repulsive potential between bosons . both linear pair and circular stationary vortex solutions are known , of which only the last ones were shown to be stable ( see part i for references ) . similarly , and for the same reasons as in parts i and ii , we introduce dimensionless variables defined by the transformation @xmath7 where @xmath8 is the chemical potential and @xmath9 is the background mass density . now , we obtain from ( [ trans ] ) and ( [ nls ] ) @xmath10 an initial condition used in parts i and ii , was @xmath11 where ( @xmath12 here has been interchanged with @xmath13 ) @xmath14 @xmath15 which is a model representing two oppositely polarized rectilinear vortices parallel to the @xmath16 plane and moving along the @xmath12 axis . the constant @xmath17 was chosen such that the velocity of the two rectilinear vortices ( along @xmath12 ) in the simulation would agree with @xmath18 , from the theory , see @xcite , table 2 . in spite of its simplicity , this model satisfies several criteria . the variable @xmath19 is scaled correctly , the wavefunction is zero at the centres of the two vortices , and far field behaviour is approximately correct ( of the order of @xmath20 ) . ( we have doubts about the formula of jones and roberts , see appendix . ) when the vortices are well separated , fetters formula @xcite is recovered ( @xmath21 , and @xmath22 as @xmath23 ) . in this paper , an initial condition for a circular vortex will be used . it will be described by formulae of similar form to ( [ incond])([inconf ] ) , but with @xmath19 and @xmath12 replaced by @xmath24 the vortex in question will move along the @xmath12 axis with velocity close to @xmath18 taken from table 1 in @xcite , and the vortex lines ( @xmath25 ) at @xmath26 will intersect the plane @xmath27 at two points : @xmath28 . it is clear that the accuracy of such a model will increase as we increase the radius @xmath29 of the circular vortex . however , as we observed in all considered cases , even for not very large values of @xmath29 , the oscillations of the respective circular vortices during their motion were negligible . the nonlinear schrdinger equation in three dimensions ( [ nlstr ] ) was numerically solved by using a discrete fast fourier transform in @xmath19 , @xmath12 , and @xmath13 to calculate the space derivatives ( pseudospectral algorithm ) , along with the leapfrog timestep . calculations were performed in a box : @xmath30 , @xmath31 , @xmath32 , with the number of mesh points @xmath33@xmath34 , and @xmath35@xmath36 . periodic boundary conditions were assumed , and the timestep was determined from the numerical stability condition . the details of our calculations are described in @xcite . figures 1 and 2 present collisions of a circular vortex parallel to the @xmath16 plane ( of radius @xmath37 ) with an arc of another such vortex of much larger radius @xmath38 ( @xmath39 ) . this arc , along with its periodic continuations to neighbouring periodicity boxes , models a linear vortex . ( we have evidence that in this model , small disturbances of the initial condition for linear vortex , also involving small discontinuities in derivatives at the boundaries , only introduce small oscillations but do not change the main features of the evolution . ) as the advection velocity ( along the @xmath12 axis ) of a circular vortex is a decreasing function of its radius ( see e.g. figure 2 in part i ) , the circular vortex of radius @xmath37 will be much faster than the `` linear '' one which , in the situation shown in figures 1 and 2 , will imply their collision . in figure 1 , the @xmath19 components of the polarization vectors for the linear vortex and the neighbouring part of the circular one during collision have opposite signs , which locally resembles the situation analysed in parts i and ii , i.e. a pair of oppositely polarized rectilinear vortices . in figure 2 , the pertinent signs are the same . nevertheless , strange as it may seem , the result of the collision in both cases is topologically the same , i.e. a ( distorted ) circular vortex and a linear vortex . this is in contradiction to the description of the two above mentioned collisions given by schwarz , see the second and third situation presented in figure 16 of @xcite . in both these situations ( in the schwarz description ) the reconnection occurs only at one of the crossing points and the result of the collision is topologically different from the initial state ( one linear vortex only in the final state ) . it should be mentioned that different behaviour at the two crossing points ( occurrence and lack of reconnection ) is in fact also in contradiction to the results of koplik and levine who examined the evolution ( within the gp model ) of two neighbouring rectilinear vortices for various angles between their vorticity vectors . reconnection was demonstrated for angles between @xmath40 and @xmath41 , and no reconnection for @xmath42 . thus , in the case of two rectilinear vortices , the angle @xmath42 implied the same behaviour as that for @xmath43 . for the collisions shown in our figures 1 and 2 , the angles between vorticity vectors at two intersection points are the same and one should not expect different behaviour there . another question , however , is the applicability of the results of koplik and levine to the collisions shown in figures 1 and 2 , where one can only locally think in terms of rectilinear vortices during collision . plane ( of radius @xmath44 ) with an arc of another such vortex , but of much larger radius @xmath45 , such that @xmath46 . the distance between the `` rectilinear '' vortex and the axis of the circular one is 4 . consecutive times are : 0 , 5@xmath47 , 7@xmath47 , @xmath48 . polarizations of the nearest segments of the approaching vortices are opposed.,title="fig : " ] plane ( of radius @xmath44 ) with an arc of another such vortex , but of much larger radius @xmath45 , such that @xmath46 . the distance between the `` rectilinear '' vortex and the axis of the circular one is 4 . consecutive times are : 0 , 5@xmath47 , 7@xmath47 , @xmath48 . polarizations of the nearest segments of the approaching vortices are opposed.,title="fig : " ] plane ( of radius @xmath44 ) with an arc of another such vortex , but of much larger radius @xmath45 , such that @xmath46 . the distance between the `` rectilinear '' vortex and the axis of the circular one is 4 . consecutive times are : 0 , 5@xmath47 , 7@xmath47 , @xmath48 . polarizations of the nearest segments of the approaching vortices are opposed.,title="fig : " ] + , and @xmath49 ( @xmath46 ) , and the distance between the `` rectilinear '' vortex and the axis of the circular one is 9 . consecutive times are : @xmath50 , @xmath51 , @xmath52 , @xmath53 , @xmath54 . the first view is along the @xmath19 axis.,title="fig : " ] , and @xmath49 ( @xmath46 ) , and the distance between the `` rectilinear '' vortex and the axis of the circular one is 9 . consecutive times are : @xmath50 , @xmath51 , @xmath52 , @xmath53 , @xmath54 . the first view is along the @xmath19 axis.,title="fig : " ] , and @xmath49 ( @xmath46 ) , and the distance between the `` rectilinear '' vortex and the axis of the circular one is 9 . consecutive times are : @xmath50 , @xmath51 , @xmath52 , @xmath53 , @xmath54 . the first view is along the @xmath19 axis.,title="fig : " ] , and @xmath49 ( @xmath46 ) , and the distance between the `` rectilinear '' vortex and the axis of the circular one is 9 . consecutive times are : @xmath50 , @xmath51 , @xmath52 , @xmath53 , @xmath54 . the first view is along the @xmath19 axis.,title="fig : " ] + here we pause for some comments . firstly , we repeated the calculation but for the segment of the large circular vortex with opposite curvature . no difference between the results in both cases was observed . secondly , we performed a calculation for a similar initial vortex configuration , but with both colliding circular vortices having comparable radii . we obtained essentially the same results ( i.e. topologically , both results could be treated as identical ) . so the exact values of the radii of the colliding vortices are not important , if only the radius of one of the vortices is finite . even more , a similar scenario of the collision of two circular oppositely polarized vortices is observed if the vortices have identical radii and their symmetry axes are parallel to each other but do not coincide . again two reconnections at the crossing points and two distorted circular vortices after collision are obtained , see @xcite . note also that the initial conditions presented in figures 1 and 2 are symmetric with respect to the plane @xmath55 passing through both centres of colliding rings . this implies that this symmetry lasts during the evolution . this was the case in our results even if it is not obvious from figures 1 and 2 . as we have mentioned earlier , collisions as considered here were also investigated by k w schwarz @xcite who performed the first important step towards a formal description of the dynamics of vortices in becs kws85 , kws88 . schwarz was able to explain many features of the vortex dynamics in these media , including the interaction with walls . in order to describe some simple observed effects , such as `` avoiding collision '' between parallel vortices and `` attraction '' by antiparallel ones , schwarz introduced explicitly an `` artificial force '' acting between cores of neighbouring vortex lines . this made it possible to reach many detailed results , but also introduced the risk of some of these results being questionable . the disagreement with our results has an important consequence . if in the course of the collision in question only one reconnection point appears , the two vortex lines join together into one vortex line and the degree of entanglement of the system of these lines increases . our result , on the contrary , states that such coplanar collisions can not increase the degree of entanglement and such a quick generation of vortex entanglement as showed schwarz in his figure 4 in @xcite can not come about in this fashion . plane with comparable radii , @xmath56 , @xmath57 ( @xmath58 ) . the periodicity box has transverse dimensions ( @xmath59 ) comparable to the larger diameter @xmath60 ( @xmath61 , i.e. @xmath62 ) . consecutive times are : @xmath50 , @xmath63 , @xmath64 , @xmath65.,title="fig : " ] plane with comparable radii , @xmath56 , @xmath57 ( @xmath58 ) . the periodicity box has transverse dimensions ( @xmath59 ) comparable to the larger diameter @xmath60 ( @xmath61 , i.e. @xmath62 ) . consecutive times are : @xmath50 , @xmath63 , @xmath64 , @xmath65.,title="fig : " ] plane with comparable radii , @xmath56 , @xmath57 ( @xmath58 ) . the periodicity box has transverse dimensions ( @xmath59 ) comparable to the larger diameter @xmath60 ( @xmath61 , i.e. @xmath62 ) . consecutive times are : @xmath50 , @xmath63 , @xmath64 , @xmath65.,title="fig : " ] + in parts i and ii the authors gave several examples of how circular vortices can be created from pairs of oppositely polarized rectilinear vortices perturbed sinusoidally in space . here we will show that similar circular vortices can be created , if a circular vortex passes by another , oppositely polarized coaxial circular vortex of somewhat larger radius . this seems to contradict what koplik and levine say when considering a coaxial collision of oppositely polarized circular vortices of different radii , ( , p. 4746 ) : `` two cases occur if two rings of _ different _ sizes approach on axis : if the ratio of radii is large , the rings simply leapfrog each other . if , however , the radii are comparable , one again sees annihilation similar to the equal - sized case , except that the partially overlapping rings continue to translate during the merger stage preceding annihilation . '' different behaviour , resulting in reconnections and formation of circular vortices , should be possible if the the collision of the circular rings is somehow perturbed . in the case of rectilinear vortices , this perturbation was introduced externally . in the case of circular vortices , it could result from interaction ( in the sense of schwarz @xcite ) of the colliding vortices with other circular vortices close by during the collision . such situations are very probable in real condensates . in this paper , they have been modelled in the simplest possible way , i.e. by choosing the dimensions of the periodicity box in the plane parallel to the colliding vortices ( @xmath16 plane ) to be comparable to the diameters of the colliding vortices . with this choice , the perturbation is due to the interaction of the colliding vortices with their neighbouring periodic images parallel to the @xmath16 plane . there are four closest neighbours in the directions of the @xmath19 and @xmath13 axis , four further ones in the directions rotated by @xmath42 along the axis of the colliding vortices , still more along the @xmath19 and @xmath13 axis , etc . this evidently suggests that the number of circular vortices after collision should be a multiple of four . this prediction was confirmed by the results of our calculations as shown in figures 37 . ) but with somewhat larger periodicity box ( @xmath66 , i.e. @xmath67 ) . the formation of four smaller circular vortices after collision can be seen . consecutive times are : @xmath50 , @xmath68 , @xmath69 , @xmath70 , @xmath71.,title="fig : " ] ) but with somewhat larger periodicity box ( @xmath66 , i.e. @xmath67 ) . the formation of four smaller circular vortices after collision can be seen . consecutive times are : @xmath50 , @xmath68 , @xmath69 , @xmath70 , @xmath71.,title="fig : " ] ) but with somewhat larger periodicity box ( @xmath66 , i.e. @xmath67 ) . the formation of four smaller circular vortices after collision can be seen . consecutive times are : @xmath50 , @xmath68 , @xmath69 , @xmath70 , @xmath71.,title="fig : " ] ) but with somewhat larger periodicity box ( @xmath66 , i.e. @xmath67 ) . the formation of four smaller circular vortices after collision can be seen . consecutive times are : @xmath50 , @xmath68 , @xmath69 , @xmath70 , @xmath71.,title="fig : " ] + ) but with a relatively large periodicity box ( @xmath72 , i.e. @xmath73 ) . no smaller circular vortices after collision can be seen . consecutive times are : @xmath50 , @xmath68 , @xmath70 , @xmath74.,title="fig : " ] ) but with a relatively large periodicity box ( @xmath72 , i.e. @xmath73 ) . no smaller circular vortices after collision can be seen . consecutive times are : @xmath50 , @xmath68 , @xmath70 , @xmath74.,title="fig : " ] ) but with a relatively large periodicity box ( @xmath72 , i.e. @xmath73 ) . no smaller circular vortices after collision can be seen . consecutive times are : @xmath50 , @xmath68 , @xmath70 , @xmath74.,title="fig : " ] + plane with similar radii , @xmath75 , @xmath76 ( @xmath77 ) . the periodicity box is not too large ( @xmath78 , i.e. @xmath79 ) . consecutive times are : @xmath50 , @xmath80 , @xmath51 , @xmath81 , @xmath82.,title="fig : " ] plane with similar radii , @xmath75 , @xmath76 ( @xmath77 ) . the periodicity box is not too large ( @xmath78 , i.e. @xmath79 ) . consecutive times are : @xmath50 , @xmath80 , @xmath51 , @xmath81 , @xmath82.,title="fig : " ] plane with similar radii , @xmath75 , @xmath76 ( @xmath77 ) . the periodicity box is not too large ( @xmath78 , i.e. @xmath79 ) . consecutive times are : @xmath50 , @xmath80 , @xmath51 , @xmath81 , @xmath82.,title="fig : " ] plane with similar radii , @xmath75 , @xmath76 ( @xmath77 ) . the periodicity box is not too large ( @xmath78 , i.e. @xmath79 ) . consecutive times are : @xmath50 , @xmath80 , @xmath51 , @xmath81 , @xmath82.,title="fig : " ] + ) but with the periodicity box somewhat larger ( @xmath83 , i.e. @xmath84 ) . consecutive times are : @xmath85 . a `` massive '' production of smaller circular vortices continues . note that in frame ( c ) , the @xmath16 plane is parallel to the page.,title="fig : " ] ) but with the periodicity box somewhat larger ( @xmath83 , i.e. @xmath84 ) . consecutive times are : @xmath85 . a `` massive '' production of smaller circular vortices continues . note that in frame ( c ) , the @xmath16 plane is parallel to the page.,title="fig : " ] ) but with the periodicity box somewhat larger ( @xmath83 , i.e. @xmath84 ) . consecutive times are : @xmath85 . a `` massive '' production of smaller circular vortices continues . note that in frame ( c ) , the @xmath16 plane is parallel to the page.,title="fig : " ] + figures 37 present head - on collisions of coaxially moving pairs of circular vortices with comparable radii . we first assume that the radius @xmath38 of the larger circular vortex is not too close to that of the smaller vortex , @xmath37 , e.g. for @xmath58 , see figures 35 . if the transverse box dimension is close enough to the size of the larger vortex , the interaction of the colliding pair with its images in the four closest neighbouring cells predominates over the main collision . as a result , the structure of the periodic vortex system changes , and four ( distorted ) circular vortices extending to the neighbouring cells are formed , see figure 3 , where @xmath62 . only for more distant boundaries can we treat the interaction of the colliding pair with neighbouring pairs as merely a perturbation of the main collision . this perturbation can lead to the production of four smaller circular rings after collision , if the ratio @xmath86 is not too large , see figure 4 , where @xmath67 . otherwise , the perturbation is too weak to switch on the reconnection , and the colliding rings pass through each other without decaying into smaller rings , see figure 5 , where @xmath87 . in all above cases ( @xmath58 ) , only the interaction with four nearest cells could be strong enough to turn on the reconnection . the interaction with further cells can be significant if the larger ring is sufficiently close to the smaller one , e.g. for @xmath77 , see figures 6 and 7 , where respectively eight and twelve smaller circular vortices are formed after collision . in parts i and ii the authors demonstrated how circular vortices can be created from pairs of oppositely polarized line vortices and confirmed the @xmath88 dependence , proving feynman s hypothesis . in this paper we present two alternate scenarios leading to the creation of smaller ring vortices from larger ones . use was made of the equipment of the interdisciplinary centre for mathematical and computational modelling ( icm ) , warsaw university . professor g. wilk should be thanked for exerting pressure on us to write up these results . jones and roberts @xcite give formulas for the wavefunction in the far field for both the three dimensional case ( circular vortex ) and the two dimensional case ( two counterstreaming line vortices ) . one might wish to compare their 2d model with our equation ( [ incond ] ) in the far field . however , we have doubts about their derivation . they first linearize equation ( [ nlstr ] ) , obtaining in the far field ( @xmath13 is interchanged with @xmath12 in our notation ) : @xmath89^{-1},\ ] ] @xmath90[y^2 + ( 1 - 2 u^2)x^2]^{-2}.\ ] ] they then find that the leading nonlinear term @xmath91 is at least of the same order as ( [ jr4p3lin ] ) , @xmath92 , and simply tack it on : @xmath93 - \case12 m y^2\}[y^2 + ( 1 - 2 u^2)x^2]^{-2}.\ ] ] this is not proper procedure . the full , nonlinear equation ( [ nlstr ] ) must be solved in the far field . two inhomogeneous equations are obtained for @xmath94 and @xmath95 ( the equation for @xmath94 is free of @xmath95 ) . we add the good news that their other calculation is valid in 3d , the nonlinear correction being of higher order in @xmath96 than the linear terms ( @xmath97 and @xmath98 respectively ) .	in two preceding papers ( infeld and senatorski 2003 _ j. phys . : condens . matter _ * 15 * 5865 , senatorski and infeld 2004 _ _ j . phys . : condens . matter _ _ * 16 * 6589 ) the authors confirmed feynman s hypothesis on how circular vortices can be created from oppositely polarized pairs of linear vortices ( first paper ) , and then gave examples of the creation of several _ different _ circular vortices from one linear pair ( second paper ) . here in part iii , we give two classes of examples of how the vortices can interact . the first confirms the intuition that the reconnection processes which join two interacting vortex lines into one and thus would increase the degree of entanglement of the vortex system , practically do not occur . the second shows that new circular vortices can also be created from pairs of oppositely polarized _ coaxial circular _ vortices . this seems to contradict the results for such pairs given in koplik and levine 1996 _ phys . rev . lett . _ * 76 * 4745 .
the theory of ordinary and generalized hermite polynomials has largely benefited of the operational formalism . the two variable hermite - kamp de frit polynomials @xcite can be defined using the following identity @xcite : @xmath0 which involves the action of an exponential operator , containing a second order derivative , on a monomial . the explicit form of the polynomials @xmath1 can be obtained by means of a straightforward expansion of the exponential in eq . , which yields : @xmath2 } \ , \frac{x^{n - 2r}\,y^r}{(n - 2r)!\,r!}\;,\ ] ] where the variables @xmath3 independent each other as a parameter , the standard hermite form are recovered by the identities @xmath4 and @xmath5 . ] . by keeping , therefore , the derivative of both sides of eq . with respect to @xmath6 , we find that the hermite polynomials can be viewed as the solution of the following heat equation @xmath7 for @xmath8 they can be written in terms of the gauss - weierstrass transform @xcite : @xmath9 which is a standard mean of solutions for the heat type problems . the higher order hermite polynomials @xcite , widely exploited in combinatorial quantum field theory @xcite , can be expressed as a generalization of the operational identity , and indeed they write , denoting the order of the polynomials , is omitted for hermite polynomials of order 2 . ] @xmath10 } \ , \frac{x^{n - mr}\,y^r}{(n - mr)!\,r!}\;.\ ] ] therefore , we can ask whether an integral transform , a sort of generalization of the gauss - weierstrass transform , also holds for the higher order case . we start discussing the case of hermite polynomials of even order with negative values of the @xmath6 parameter , namely : @xmath11 we express this family of polynomials in terms of a suitable transform following the procedure , put forward in @xcite , which considers the operator function @xmath12 where @xmath13 is a function admitting a fourier transform @xmath14 . with this assumption we find that the operator @xmath15 can be written as : @xmath16 and , therefore , we can express the action of the operator @xmath15 on a given function @xmath17 as the integral transform indicated below @xmath18 we can now apply the same procedure to express the exponential operator intervening in the definition of @xmath19 ( see eq . ) , thus obtaining the following integral transform yielding the even order hermite polynomials @xmath20{\|y\|}\right)^n\ ] ] with @xmath21 which is motivated by the the request that the integrals defining @xmath22 is convergent . ] , after a redefinition of the variable , can also be written as @xmath23{\|y\|}}\ , \int_{- \infty}^{\infty}\ , \mathrm{d}\xi\ , \xi^n \,\tilde{e}_{2p}\ , \left(\frac{x - \xi}{\imath \sqrt[2p]{\|y\|}}\right)\;.\ ] ] it must remarked that the same procedure , applied to the case @xmath24 , does not lead to a gauss - weierstrass transform as in eq . , which holds only for @xmath25 . in the next section we will extend the formalism developed in these introductory remarks , and prove that the airy transform and the associated polynomials can be framed within the same context . the higher order hermite polynomials satisfy the following recurrences @xcite : @xmath26 with the combination of the first and third recurrences yielding @xmath27 therefore , the higher order hermite polynomials satisfy a generalized heat equation , and this justifies the operational definition given in eq . . furthermore , by interpreting @xmath6 as a parameter , we can use the first two recurrences to prove tha they satisfy the @xmath28-th order ode : @xmath29 in the previous section we have considered even order hermite polynomials only . here we will discuss the third order case and their important relationship with the airy transform and the airy polynomials @xcite . before getting into this specific aspect of the problem , we note that the following identity holds : @xmath30{\lambda } xt}\ , ai(t)\;,\ ] ] where @xmath31 is the airy function @xcite . let us now consider the third order pde @xmath32 whose formal solution writes @xmath33 by applying the identity given in eq . , and by limiting ourselves to the case @xmath8 , we find : @xmath34{3}}\ , \int_{- \infty}^{\infty}\ , \mathrm{d}k\ , ai\left(\frac{k}{\sqrt[3]{3}}\right)\ , \mathrm{e}^{\sqrt[3]{y } k \partial_x}\ , f(x ) \nonumber \\ \frac{1}{\sqrt[3]{3}}\ , \int_{- \infty}^{\infty}\ , \mathrm{d}k\ , ai\left(\frac{k}{\sqrt[3]{3}}\right)\ , f(x + \sqrt[3]{y } k ) \nonumber \\ \!\!&=&\!\ ! \frac{1}{\sqrt[3]{3 y}}\ , \int_{- \infty}^{\infty}\ , \mathrm{d}\xi\ , ai\left(- \frac{x - \xi}{\sqrt[3]{3 y}}\right)\ , f(\xi)\ ; , \qquad \qquad ( y > 0)\end{aligned}\ ] ] which is recognized as the airy transform of the function @xmath13 . the concept of the airy transform was introduced in ref . @xcite , and has found noticeable applications in classical and quantum physics @xcite . the airy transform of a monomial , namely @xmath35{3 y}}\ , \int_{- \infty}^{\infty}\ , \mathrm{d}\xi\ , ai\left(- \frac{x - \xi}{\sqrt[3]{3 y}}\right)\ , \xi^n\ ; , \qquad \qquad ( y > 0)\ ] ] has been defined as airy polynomials @xcite , but , according to the discussion of the previous section , they are also recognized as the third order hermite polynomials @xmath36 . the characteristic recurrences ( specialized to the case @xmath37 ) , can also be directly inferred from eq . . for further convenience we will introduce the following two variable extension of the airy functions @xmath38 that , in terms of the ordinary airy function writes @xmath39{3y}}\,ai\left(\frac{x}{\sqrt[3]{3y}}\right)\;.\ ] ] it is also easily shown that it satisfies the ode @xmath40 and that any other function linked to @xmath41 by @xmath42 satisfies the ode @xmath43 \,=\,0\ ] ] which on account of the identity @xmath44 can be written as @xmath45 if we assume @xmath46 , @xmath47 , and @xmath48 , we get the following generalization of the airy function @xmath49 satisfying the ode @xmath50 as an example a comparison between the ordinary airy function and its generalization of order 7 is shown in fig . [ fig : aicmp ] the point of view developed in sec . 2 can be generalized to the case with @xmath51 . we consider the generalized hermite polynomials of odd order ( @xmath52 ) and note that the same procedure exploited in the previous sections leads to the following integral representation @xmath53{(2p + 1 ) \|y\|}}\ , \int_{- \infty}^{\infty}\ , \mathrm{d}\xi\ , ai^{(2p + 1 ) } \left(\frac{\xi - x}{\sqrt[2p + 1]{(2p + 1 ) \|y\|}}\right)\,\xi^n\;,\ ] ] the use of a further variable or parameter in the theory of special functions and/or polynomials offers a further degree of freedom , which may be helpful to derive new properties , otherwise hidden by the loose of symmetry deriving from the fact that a specific choice of the variable has been done . this is indeed the case of the hermite polynomials , which acquire a complete new flavor by the use of the @xmath6 variable and the case of the airy function given in eq . which is easily shown to be the natural solution of the pde @xmath54 and the translation property @xmath55 according to eq . , we can write the solution of the equation @xmath56 as follows @xmath57 it is evident that the further generalization @xmath58 satisfies a higher order heat equation and can be exploited to obtain the solution of the same family of equation in terms of an appropriate integral transform . before concluding this note let us consider the following schrdinger equation @xmath59 describing the motion of a particle in a linear potential ( @xmath60 is a constant with the dimension of a force ) . this equation can be written in the more convenient form @xmath61 where has the dimensions of a squared length and @xmath62 of an inverse cube length , so that @xmath63 has the dimensions of a length . ] @xmath64 as discussed in refs . @xcite , the previous equation can be treated using different means , a very simple solution being offered by the use of the following auxiliary function : @xmath65 which satisfies the equation @xmath66 the solution of eq . can therefore be written as @xmath67 where the operator @xmath68 , once written in an integral form , is recognized as the airy transform . the solution of the equation @xmath69 can be obtained in an analogous way , and it is given by @xmath70 where the operator @xmath71 can be expressed in terms of integral transforms , linked to the higher order airy functions discussed in these sections . further examples of generalization of the airy function are due to watson @xcite , who introduced a complete class of functions with interesting properties for physical applications . as an example we consider the function expressed as @xmath72 which is compared to the ordinary airy function in fig . [ fig : wacmp ] , and satisfies the ode the use of the unitary transformation @xmath74 shows that it can be associated to the first ode @xmath75 the transformation has removed the second order derivative and has simplified the underlying group of symmetry , which has been reduced from @xmath76 to the dilatation group . we have mentioned this example because it may open a new interesting point of view to schrdinger type equations involving quadratic potentials . before concluding this paper we discuss whether the airy polynomials can be exploited to obtain a series expansion of a given function . the problems associated with the orthogonal properties of ordinary and higher order hermite polynomials has been thoroughly considered in a previous publications ( see refs . @xcite ) ; here we will use the point of view developed in ref . we consider indeed the following expansion @xmath77 and we will show that the use of the operational tools developed in this paper provides , in a fairly natural way , the coefficients @xmath78 , and we will display the orthogonal properties of this family of polynomials . on account of the definition of the higher orders hermite polynomials ( see eq . ) , eq . yields : @xmath79 the use of the identity allows to specify the action of the exponential operator on the function @xmath13 as : @xmath80{3 \|y\|}}\,\int_{- \infty}^{\infty}\ , \mathrm{d}\xi\ , ai\left(- \frac{x - \xi}{\sqrt[3]{3 \|y\|}}\right)\,f(\xi ) \,=\ , \sum_{n = 0}^\infty\,a_n\,x^n\;,\ ] ] and the insertion of the explicit expression of the airy function ( see eq . ) yields : @xmath80{3 \|y\|}}\,\frac{1}{2 \pi}\,\int_{- \infty}^{\infty}\ , \mathrm{d}\xi\ , \int_{- \infty}^{\infty}\ , \mathrm{d}t\,\exp{\left\{\frac{\imath}{3}\,t^3\,+\,\imath\ , \frac{\xi - x}{\sqrt[3]{3 \|y\|}}\,t\right\}}\,f(\xi ) \,=\ , \sum_{n = 0}^\infty\,a_n\,x^n\;.\ ] ] the expansion of the @xmath81-dependent part of the exponential on the lhs of the previous equation yields : @xmath82{3 \|y\|}}\,\int_{- \infty}^{\infty}\ , \mathrm{d}t\ , f(\xi)\,\partial_\xi^n\,ai\left(\frac{\xi}{\sqrt[3]{3 \|y\|}}\right)\;,\ ] ] which holds only if the integral on the rhs of this equation converges . in these concluding remarks we have touched on several problems , which are worth to be explored thoroughly . the link between the watson and the @xmath83 functions , the possibility of obtaining more general airy - type transforms and their relevance to bell - type polynomials , and a more rigorous formulation of the orthogonality properties of the airy polynomials will be the topic of a forthcoming investigation . g. dattoli and e. sabia , `` the binomial transform '' , submitted to journal of integral transforms and special functions . o. valle and m. soares , _ airy functions and application to physics _ , world scientific , london ( 2004 ) .	the airy transform is an ideally suited tool to treat problem in classical and quantum optics . even though the relevant mathematical aspects have been thoroughly investigated , the possibility it offers are wide and some aspects , as the link with special functions and polynomials , still contains unexplored aspects . in this note we will show that the so called airy polynomials are essentially the third order hermite polynomials . we will also prove that this identification opens the possibility of developing new conjectures on the properties of this family of polynomials .
@xmath0re is an attractive therapeutic radioisotope with broad clinical applications . oncology applications range from palliation of metastatic bone pain to bone marrow ablation and in general to the use of @xmath0re labelled therapeutic agents to target specific cancerous tissues ( see for instance reference @xcite ) . other therapeutic applications comprise the inihibition of restenosis after percutanueous transluminal coronary angioplasty ( ptca ) , radiation synovectomy , and intravasal brachytherapy . in the field of metabolic radiotherapy @xmath0re shows several favourable characteristics . it can be produced carrier free using a w - re generator ( see for instance reference @xcite ) and its chemistry is similar to that of @xmath3tc which is the most used radioisotope in nuclear medicine ( medical imaging ) . the @xmath0w parent has a 69 d half - life , which permits to use the generator for a relatively long period , the equilibrium between parent and daughter setting up within about two days . @xmath0re decays to @xmath0os@xmath6 in about 0.7 days via the emission of a @xmath7-ray with a maximum energy of 2.12 mev ( 0.78 mev average energy ) , which can be used for destroying cancerous cells . in addition @xmath0os@xmath6 emits promptly ( 0.69 ns ) a 155 kev @xmath2-ray ( 15% ) , which can be used for imaging . given the chemical similarity with @xmath3tc , it can be linked to molecules of hyaluronic acid ( ha ) which have the function of carrying it to specific sites in the body , e.g. with the production of an accumulation and retention of the drug in the liver @xcite . hence the potential interest for treating liver cancers . on the other hand the @xmath7 radiation could interfere with the labelling process and destroy the molecule used to carry the radioisotope in the body to the target organ , thus reducing the therapeutic effect . two additional points have to do with the relatively long lifetime and with the rich photon spectrum of @xmath0re , which extends to high energy , compared to the single 140 kev photon emitted by @xmath3tc with a half - life of about 6 h. taking into account the branching ratios ( br ) and the lifetimes , the relative @xmath0re/@xmath3tc counting rate , for an equal number of @xmath8-moles , is about 9% . this implies that photon detectors developed for imaging with @xmath3tc are not necessarily sensitive enough . the usual thickness of a low - energy pb collimator , say 20 mm , will be almost transparent to @xmath0re photon lines at 300 kev or higher energies , even with brs depressed by a factor of ten or more , producing background counts which blur the image and worsen the spatial resolution . to study the effect of metabolic radiotherapy in mice , we have therefore built a new small high - sensitivity @xmath2-camera , following the experience with the yap - camera @xcite@xcite@xcite , which is used routinely to image small animals ( mice ) with @xmath3tc - ha at the laboratori nazionali di legnaro , italy ( e.g. @xcite@xcite ) . the gamma camera is based on a matrix of yttrium aluminium perovskite doped with cerium ( yap : ce or yalo@xmath9:ce ) crystals @xcite , with friendly mechanical properties ( no igroscopicity ) , fast response ( @xmath1025 ns decay time ) , high density ( 5.37 g / cm@xmath5 ) and good x- and @xmath2-ray absorption . there are 66@xmath466 pixels , each [email protected]@xmath410 mm@xmath5 , covering a field - of - view ( fov ) of 40@xmath440 mm@xmath12 . the pixels are covered laterally by a special 5 @xmath8 m thick reflective coating which provides also the optical separation between neighboring elements . the scintillator is read out by a r2486 hamamatsu positionsensitivepmt @xcite with a 76 mm diameter photocathode . the anode consists of 16 plus 16 wires crossing at 90@xmath13 and connected by two resistive chains , defining the @xmath14 and @xmath15 directions . the wires define an active area with a diameter of about 50 mm . a 20 mm thick lead parallel hexagonal - hole collimator @xcite , with hole diameter 1.5 mm and septa of 0.18 mm , is placed in front of the yap matrix . the detector is triggered using the last dynode and the ends of the @xmath14 and @xmath15 resistive chains ( x@xmath16 , x@xmath17 , y@xmath16 , y@xmath17 ) are amplified , stretched and read out by a pc using a pci 6023e card @xcite . the coordinates of the photon impact point are then reconstructed by charge division , @xmath18 and similarly for @xmath15 . the energy response of the detector to a spatially uniform source ( flat field ) of @xmath3tc 140 kev photons prior to energy equalization has been determined using a solution containing @xmath3tc which covered the whole fov and was located a few centimeters in front of the collimator surface . the @xmath19 has been arbitrarily divided in pixels and the average energy computed in each pixel . the corrections to the measured energy extracted from this calibration are shown in fig . [ fig : flat ] and have been used in the following . as shown below with the @xmath0re spectrum , these corrections improve the energy resolution of the detector . apart from the energy equalization , no other correction has been applied . with @xmath3tc the sensitivity of the gamma camera is found to be @xmath102@xmath410@xmath20 or @xmath107@xmath410@xmath5 cps / mci , which agrees roughly with calculations . tc.,width=6 ] pointlike @xmath21am ( 60 kev photons ) , @xmath22co ( @xmath10122 kev photons ) and @xmath23cs ( 660 kev ) sources located in different positions a few millimeters distant from the collimator have been used both to simulate the @xmath0re energy spectrum and to evaluate the spatial resolution of the detector . in addition @xmath21am and @xmath22co , together with @xmath3tc photons , permit the calibration of energy scale . the overall energy spectrum is presented in fig . [ fig : acc - en ] and the cumulative image of the sources is visible in fig . [ fig : acc - im ] . the 60 kev is well prominent and a 122 kev shoulder is also visible ; @xmath23cs instead produces a broad shoulder at about half of the photon energy . using appropriate energy cuts the three images can be separated . the intrinsic spatial resolution of the system is quite good , since individual collimator holes can be clearly seen in fig . [ fig : acc - im ] . the resolution however is worsened in practice by the small thickness of the septa relative to the hole diameter of the present collimators , making them partially transparent to radiation , so the actual effective resolution is @xmath24@xmath14 @xmath10 3 mm fwhm at 122 kev . am , @xmath22co and @xmath23cs superimposed spectra measured with the yap camera.,width=6 ] am ( center ) , @xmath22co ( top ) and @xmath23cs ( bottom ) images measured with the yap camera.,width=6 ] the @xmath0re photon spectrum measured with a ge detector is shown in fig . [ fig : re - ge ] . the 155 kev line is prominent , but many more lines are present at higher energy , in some cases , e.g. at @xmath10300 kev , with brs only a factor of ten lower , or at 800 - 1200 kev with intensities lower only by a factor of 100 @xcite . re spectrum measured with a ge detector.,width=8 ] the @xmath0re spectrum measured with the yap camera during the imaging of a c57 mouse ( see next section ) is shown in fig . [ fig : re - yap ] . the spectrum is shown before and after applying the corrections for energy non - uniformity , and a clear improvement is observed with the shrinking of the 155 kev peak . after the corrections , the energy resolution is @xmath24e / e @xmath10 33% @xmath25 155 kev . re spectrum measured with the yap camera : the shaded spectrum has been corrected for energy non - uniformity.,width=6 ] the labelling reaction of ha using @xmath0re was carried out with good yields ( 65 - 70% ) . the radiolabelled compound was purified with a size exclusion chromatographic method before being used for biodistribution studies . stability studies in rat serum confirmed the maintaining of the re linked to the polymer and there was no evidence of radio - decomposition after a few hours . the radiotoxicity of @xmath0re has been tested `` in vitro '' and compared with @xmath3tc , with which no effect is expected . cells of the m5076 tumor line have been treated with @xmath0re and @xmath3tc solutions , and irradiated with x - rays . the number of binucleate cells and of micronuclei in the cells is then counted . m5076 cells turn to out to be highly sensitive to x - rays ( 0.25 - 2 gy)@xcite . activities of 150 - 300 @xmath8ci of @xmath3tc show no effect , but @xmath0re @xmath7-rays , with similar initial activities integrated during 72 h , seem to be quite efficient in inducing dna damage@xcite . to test the full chain , from the radiolabelling to to the imaging `` in vivo '' , a c57 black mouse ( healthy , female ) has been injected with @xmath0re - ha . after general anesthesia , the solution with an activity of about 250 @xmath8ci was injected in the caudal vein . the mouse was positioned along the diagonal of the fov , with the locus of injection outside it , and was monitored for about three hours . the image collected in the first five minutes shows a large spot close to the locus of injection in the tail ( fig . [ fig : c57a ] ) . after 5 mins the activity concentrates roughly in the centre of the body , in a volume which contains the liver ( fig . [ fig : c57b ] ) . the activity is slowly decreasing during the 3 h of the measurement . after 3 h the mouse was sacrificed , and the organs were extracted and measured with a microcurimeter ( fig . [ fig : att ] ) . the liver contains 60% of the residual activity and close by organs another 20% , in agreement with the scintigrafic image ( fig . [ fig : c57b ] ) , where individual organs are not resolved . even with limited resolution the test shows that it is possible to monitor the biodistribution of @xmath0re in mice , with a potential saving in the number of animals needed for testing the @xmath0re therapy . re - ha in the caudal vein.,width=6 ] re - ha in the caudal vein . the volume of large activity corresponds to the liver.,width=6 ] @xmath26''.,width=6 ] preliminary results obtained using a new yap camera in imaging @xmath0re sources and c57 black mice injected with a @xmath0re - ha solution have been presented . to increase the space resolution without losing sensitivity , and to obtain different projections simultaneously , we are building two new cameras to be positioned at 90 degrees around the animal . they use a csi(tl ) matrix with 1@xmath41@xmath45 mm@xmath5 pixels read out by h8500 hamamatsu flat panel pmt @xcite . parallel - hole pb collimators 20 mm thick , with 1 mm diameter hexagonal holes and 0.2 mm thick septa , will be mounted in front of the scintillators . also specially made collimators with thicker septa and/or different absorber material will be used . the front - end electronics for the 64 channels of the h8500 has been designed using mpx-08 chips @xcite . the system will be mounted on a rotating support in order to produce tomographic images . labelling techniques of biomolecules for targeted radiotherapy , iaea - tecdoc-1359 , vienna july 2003 a. antoccia et al . , dati preliminari e danno al dna del @xmath3tc - ha in sistemi `` in vitro '' ed `` in vivo '' , suppl . sirr vii ( 2004 ) 43 f. de notaristefani et al . , yap camera : a small field camera with sub - millimiter spatial resolution , eur . j. nucl . med . 22 ( 1995 ) 337 f. de notaristefani et al . , first results from a yap : ce gamma camera for small animal studies , ieee trans . nucl . 43 ( 1996 ) 3264 f. de notaristefani , f. vittori and t. malatesta , development of yap : ce multi - crystal detectors , physica medica 13 ( 1997 ) 237 r. rossin et al . , in vivo biodistribution studies on hyaluronan butyrate by means of 99mtc direct labelling and yap camera , in : technetium , rhenium and other metals in chemistry and nuclear medicine ( m. nicolini , u. mazzi eds . ) , pp 689 - 693 , sgeditoriali , padova , 2002 . giron et al . , yap - camera device for biodistribution studies in mice of 99mtc - radiotracers , poster at the 6th int . on technetium in chemistry and nuclear medicine , brixen , italy , 2002 + http://tc2002.unipd.it/posters_main.htm crytur preciosa a.s . , palackeho 175 , 511 19 turnov , czech republic . a. antoccia et al . , study of metabolic radiotherapy with @xmath0re via small animal imaging , presented at the annual meeting of the italian physical society , brescia 20 - 25 settembre 2004 http://www.sif.it	@xmath0re is a @xmath1 ( emax = 2.12 mev ) and @xmath2 ( 155 kev ) emitter . since its chemistry is similar to that of the largely employed tracer , @xmath3tc , molecules of hyaluronic acid ( ha ) have been labelled with @xmath0re to produce a target specific radiopharmaceutical . the radiolabeled compound , i.v . injected in healthy mice , is able to accumulate into the liver after a few minutes . to study the effect of metabolic radiotherapy in mice , we have built a small gamma camera based on a matrix of yap : ce crystals , with [email protected]@xmath410 mm@xmath5 pixels , read out by a r2486 hamamatsu pspmt . a high - sensitivity 20 mm thick lead parallel - hole collimator , with hole diameter 1.5 mm and septa of 0.18 mm , is placed in front of the yap matrix . preliminary results obtained with various phantoms containing a solution of @xmath0re and with c57 black mice injected with the @xmath0re - ha solution are presented . to increase the space resolution and to obtain two orthogonal projections simultaneously we are building in parallel two new cameras to be positioned at 90 degrees . they use a csi(tl ) matrix with 1@xmath41@xmath45 mm@xmath5 pixels read out by h8500 hamamatsu flat panel pmt .
the nos - hoover model , @xcite,@xcite , is widely used in molecular dynamics for simulating a system s behavior at constant temperature . the central idea of the model , that is the introduction ancila dynamic variables to control kinetic energy , admits of various implementations . in the most simple and exploited form it amounts to considering a minimal non - linear extension of the original equations for the system . it should be noted that hamiltonian versions of the model drew considerable attention , @xcite . in the present paper we study the nos - hoover model in the form generally employed in molecular dynamics , that is a system constructed from initial hamiltonian equations , i.e. newton s second law , by employing non - linear dissipative terms on their right - hand sides and an additional equation for the dissipation constant,@xmath3 , which is allowed to vary in time , so that the equations of evolution read @xmath4 in this setting the nos - hoover model is a hamiltonian system with dissipation . for small values of @xmath1 the dissipative effects can be treated within the framework of perturbation theory . considerable criticism has been levelled at the nos - hoover approach , ( see paper @xcite and references therein ) , because it runs across difficulties in providing the correct thermodynamic behavior for simple harmonic systems , but it is still generally accepted that its efficiency improves with the increase of complexity and dimension of the simulated system,@xcite . in fact , from the point of view of dynamical theory it models a non - linear time - dependent dissipative environment and , therefore , has some proper physical interest . in this paper we focus our attention on ensembles of harmonic oscillators ; the importance of such systems follows from the fact that among these , is the harmonic lattice , familiar in the theories of solids and molecules . the potential energy of a harmonic lattice is given by the quadratic form @xmath5 in which @xmath6 is the @xmath7-th coordinate of the @xmath8-th particle . to see the symmetry properties of the model we may cast eq.([n - h1 ] ) in the matrix form @xmath9 in which @xmath10 is the matrix of force constants , @xmath11 . it can be transformed by an appropriate orthogonal transformation , @xmath12 , to the diagonal form @xmath13 assume that all @xmath14 are equal , @xmath15 , and let @xmath12 be the matrix of the orthogonal transformation mentioned above , and @xmath16 an @xmath17-dimensional vector of coordinates with respect to the new coordinate system determined by @xmath12 . since @xmath18 is constant , and @xmath3 is an invariant of orthogonal transformation , we may cast eq.([n - h1 ] ) in the form @xmath19 so that the equation acquires the form @xmath20 in which the matrix @xmath21 is the diagonal one . thus , we have transformed the original problem of harmonic lattice to that for a set of harmonic oscillators , which do not interact with each other . the latter problem is more tractable , from analytical point of view . at this point it is worthwhile to notice that problems of molecular dynamics involves dynamical systems of extremely high dimension , and this circumstance brings about specific difficulties for numerical simulation . the best approach to the problem is to use analytical methods in conjunction with the numerical ones , especially in cases like the nos - hoover model , where systematic investigation of high dimensional problems is particularly interesting , ( see @xcite ) . for this end we use the method of windows worked out for the needs of the relaxation dynamics of spin in superfluid @xmath22 , ( see the review article @xcite ) to obtain a general picture of the nos - hoover dynamics for an ensemble of harmonic oscillators . we show that it is characterized by the presence of oscillations round the stationary solution corresponding to an equilibrium , for which the phase - space sampling is different from the normal law . the numerical analysis of the nos - hoover model given by eqs.([n - h1 ] ) , which is a high - dimensional non - linear dissipative system , is a serious challenge , and , in fact , it is generally confined to low dimensional situations,(see paper @xcite , in which the case of two oscillators is considered ) . in treating high - dimensional problems the key point is the wise choice of output variables . in the present case it is dictated by the physics of the problem , and taking into account the structure of the nos - hoover model , that is its being a hamiltonian system with dissipation , we employ , to the effect , the total energy of the system @xmath23 , @xmath24 and the dissipative variable @xmath3 . directing the output in plane @xmath25 , we obtain a kind of two - dimensional window on the phase - space of the model , which has dimension @xmath26 , @xmath17 is the number of oscillators . since we aim at studying the situations in which @xmath17 is large , the two - dimensional reduction is of primary importance . next , one should look for the distribution law for the the system in phase - space , and one could expect that it should be either the microcanonical or the gibbs one ; the first is characterized by its being centered on a particular value of energy , @xmath27 , whereas the latter by its being of the characteristic bell - shape . nothing of the sort happens . energy regions of the phase - space against visiting time ; the number of oscillators @xmath28 . dashed line corresponds to eq.([arcsin ] ) provided by the averaged equations . , width=288,height=192 ] to find the distribution law we consider the partition of the phase - space in regions @xmath30 corresponding to the energy intervals @xmath31 assumed to be of equal size , and compile a record of periods of time @xmath32 which the system spends in regions ( [ regions ] ) ; the total time of simulation reads @xmath33 the frequencies for the system s visiting the regions are given by the equation @xmath34 it is convenient to use a representation for the set of frequencies by means of histogram , that is rectangles whose widths represent the energy intervals ( [ regions ] ) and whose heights represent corresponding frequencies . it is worth noting that the partition of the phase - space in the energy regions ( [ regions ] ) can be effected in a more graphic form with the help of the window @xmath25 on the phase space . in fact , the numerical simulation gives the picture of the system s motion as a thin ring , the times @xmath35 are those spent in the bands @xmath36 . therefore , the frequencies ( [ visiting ] ) could be found with the two - dimensional window , @xmath37 , see fig.([build ] ) . trajectories visiting regions @xmath38 in ( @xmath23 - @xmath3 ) window . @xmath39 , @xmath40 , @xmath29 . , width=288,height=192 ] recall that we shall consider a set of n _ one - dimensional _ harmonic oscillators . by following the prescription given above we obtain the histogram given in fig.([hyst1p ] ) , describing the probability distribution @xmath41 for the system in the phase - space . it is quite different from the microcanonical one or the gibbs one . it is important that the window @xmath25 provides a means for elucidating the form of the fluctuations round the stationary state given by the equations @xmath42 the circular motion seen from the window @xmath43 is characterized by a mean angular velocity given by the law @xmath44 to within one - thousandth . the fact agrees with the histogram given in fig.([hyst1p])(see section [ sec : averaging ] for the details ) . for the case of ideal gas the oscillation law @xmath45 was found in @xcite ( see eqs.(5 ) and ( 8) of paper @xcite ) . it should be noted that the shape of the histogram depends on the amplitude of the oscillations , that is the size of deviations in the energy @xmath23 from the stationary value @xmath46 determined by the temperature parameter of the model . in the next section we shall find the distribution using averaging method , and it is worth noting that the numerical results are in good agreement with those given by the averaging , as is seen from fig.([maxwell1p ] ) . the distribution in energy that should correspond to gibbs canonical ensemble for the set of harmonic oscillators at temperature @xmath47 , is given by the equation @xmath48 and is totally different from fig.([hyst1p ] ) . the discrepancy between the obtained and expected distributions indicate that the nos - hoover model describes a kind of non - linear dissipative system having special properties . bars , energy distribution , obtained from numerical experiment . curve , gibbs distribution for the set of harmonic oscillators . @xmath28 , @xmath29 . , the simulation of the last section , which was thoroughly checked by calculating with different algorithms and comparing their results , might nonetheless be subject to artifacts and errors . in this respect , it is important that analytical means capable of verifying section [ sec : histogram ] , have been used and resulted in agreement with the numerical work . let us notice that according to the prescription described in introduction ( see eqs . ( [ n - h1 ] ) ) the ensemble of @xmath17 harmonic oscillators confined to the nos - hoover thermostat is described by the system of equations @xmath49 it is important that the dissipative constant @xmath1 is suggested to be small , @xmath50 , therefore , we may consider , as the first approximation the non dissipative regime for which @xmath51 , @xmath52 . in this case there is an exact solution given by the equation @xmath53 the energy of the @xmath8-th oscillator devided by mass @xmath2 reads @xmath54 all the masses of the particles assumed equal , @xmath55 , oscillators differing by their frequencies @xmath56 . equations ( [ xx ] ) , from the topological point of view , mean that the ensemble s motion belongs to an @xmath17-dimensional torus , the whole phase space being foliated by the tori . we shall consider the system at temperature @xmath47 . since the parameter @xmath1 is small , we may take into account the nonlinear dissipative terms on the right hand sides of eqs . ( [ nos1_1 ] ) , within the framework of the averaging approach , that is by substituting the basic equations ( [ xx ] ) into the right hand sides of the exact equations @xmath57 and cancelling out the oscillating terms . thus , we obtain the averaged equations @xmath58 or @xmath59 as follows from eq.([energy1 ] ) . since the total energy of the system is given by the equation @xmath60 we obtain the following two equations @xmath61 which have the stationary solution : @xmath62 describing the oscillators at the temperature parameter @xmath63 close to the stationary solutions we have the equations for energy @xmath64 in which @xmath65 is small . therefore the equations for @xmath65 and @xmath3 acquire the form @xmath66 on linearizing the equations indicated above we obtain @xmath67 and hence the equation for @xmath65 @xmath68 which has the form of harmonic oscillator with the frequency @xmath69 solutions to eqs.([nos1_1 ] ) are illustrated in plane ( @xmath23,@xmath3 ) , in fig.([oscil1p ] ) oscillations round the stationary solution , the characteristic frequency @xmath70 . @xmath28 , @xmath29 . dashed line corresponds to the averaged motion of @xmath3 and @xmath23 . , width=288,height=192 ] we may use eq.([z ] ) for finding the time spent by the system in the region of the phase - space corresponding to the energy interval , @xmath71 ; it is given by the equation @xmath72 \label{arcsin}\end{aligned}\ ] ] which leads to the agreement with histogram of fig.([hyst1p ] ) . we have studied the dynamics of an ensemble of harmonic oscillators confined to the nos - hoover thermostat , the number of the oscillators , @xmath17 , being large , and the dissipative constant , @xmath1 , small . the numerical simulation , and the analytical averaging method , indicate that the nos - hoover model does not provide for _ thermodynamically meaningful distribution _ of the system s samples in the phase - space , e.g. there is no gibbs distribution . the @xmath44-frequency law deserves some attention . indeed , if the mass @xmath2 of a particle , corresponding to the oscillator , be @xmath73 , and @xmath74 , the frequency of the oscillations generated by the thermostat dynamics , should be in the region of @xmath75 ghz , that is low frequency region of molecular vibrations . therefore in applying the nos - hoover thermostat to simulating molecular dynamics in ghz - region , one might expect spurious effects of parametric resonance . at the same time it is worth noting that the nos - hoover model corresponds with a hamiltonian system confined to dissipative environment , that is it comprises a base hamiltonian system , e.g the oscillators , and a dissipative extension formed by ancila variables , in the present case it is the variable @xmath3 . a similar systems , even though more sophisticated , is the leggett - takagi theory of spin dynamics in superfluid phases of helium-3 , @xcite , in which the equation describing the spin motion are augmented by an equation for the order parameter , which contains a dissipative term . the situation is reminiscent of that taking place in the hydrodynamics treatment of viscous phenomena in the ghz - region , where according to the theory worked out by mandelstam and leontovic , see @xcite , the effects of dissipation can be accommodated by employing an ancila dynamical variable @xmath76 , which describes certain states of the system , e.g. concentration of a chemical reagent . the evolution equations for @xmath76 have dissipative character , for they should describe the system s coming to equilibrium , though the initial equations for the system could be of hamiltonian form . the nos - hoover model may turn out to be of a similar kind and thus helpful in studying interesting physical problems . 99 s.nos , progr.theor.phys . suppl . * 103 * , 1 - 46 ( 1991 ) . w.g . hoover and b.l . holian , phys.lett . * a211 * , 253 ( 1996 ) . hoover , phys.rev . * a31 * , 1695 ( 1985 ) . m.dalessandro , a.tenenbaum , and a.amadei , j.physical chemistry b 106 , 5050 ( 2002 ) . h.a.posch , w.g.hoover , and f.j.vesely , phys.rev . * a33 * , 4253 ( 1985 ) . golo , int.j.shape modeling , * 1 * , 109 ( 1994 ) . a.j.leggett and s.takagi , ann.phys . * 106 * , 79 ( 1977 ) . v.l . golo and a.a.leman , spin relaxation in superfluid @xmath77 in turned - off and weak magnetic fields , in `` helium three '' , eds . w.p.halperin and l.p.pitaevskii , elsevier ( 1990 ) . l.d.landau and e.m.lifshitz , hydrodynmics , ch.viii , nauka ( 1988 ) .	we study the dynamics of an ensemble of non - interacting harmonic oscillators in a nonlinear dissipative environment described by the nos - hoover model . using numerical simulation we find the histogram for total energy , which agrees with the analysis of the nos - hoover equations effected with the method of averaging . the histogram does not correspond to gibbs canonical distribution . we have found oscillations at frequency proportional to @xmath0 , @xmath1 the dissipative parameter of thermostat and @xmath2 the characteristic mass of particle , about the stationary state corresponding to equilibrium . the oscillations could have an important bearing upon the analysis of simulating molecular dynamics in the nos - hoover thermostat .
oh megamasers ( ohms ) are rare luminous 18 cm masers associated with major galaxy merger - induced starbursts . the hosts of ohms are ( ultra)luminous ir galaxies ( [ u]lirgs ) , and the ohm fraction in ( u)lirgs peaks at about 1/3 in the highest luminosity mergers @xcite . it is not known whether all major mergers experience an ohm stage or what detailed physical conditions produce ohms , but it is clear that ohms are a radically different phenomenon from the aggregate oh maser emission associated with `` normal '' ( galactic ) modes of star formation in galaxies . @xcite posed a key question : why do @xmath7 of lirgs show no ohm activity ? to reframe the question : given two merging systems with similar global ir and radio continuum properties in the same morphological stage of merging , why does one show ohm emission while the other does not ? what is the difference between the two systems ? perhaps there is no difference and the fraction of ohms among mergers simply reflects beaming or oh abundance . or perhaps ohm activity depends on small - scale conditions that are decoupled from global properties of mergers . the provenance of ohm emission vis - - vis the host galaxy has been extensively investigated in the radio through x - ray bands by comparing samples of ohm galaxies to similarly selected non - masing control samples . for example , @xcite and @xcite studied radio and ir properties vis - - vis the agn versus starburst contributions to ohm activity , @xcite and @xcite investigated the ohm fraction in ( u)lirgs versus star formation rate and ir color , @xcite and @xcite used optical spectral classification to distinguish populations and to quantify agn fraction in ohm hosts , and @xcite conducted an x - ray study of the contribution of agns to ohm hosts . while some of these studies pointed to minor differences in statistical samples of ohm hosts versus nonmasing systems , they could not identify on a case - by - case basis which systems would harbor ohms and which would not based on any observable quantity except the oh line itself . theoretical modeling of ohm formation has seen a recent renaissance : @xcite model the @xmath8 pc molecular torus in iii zw 35 and show how ohm emission is a stochastic amplification of unsaturated emission by multiple overlapping clouds , and @xcite show how the general excitation of ohms is fundamentally different from galactic oh maser emission and predict that a single excitation temperature governs all 18 cm oh lines . while the physics of ohms is crystallizing , and models predict that beaming is not likely to be the dominant factor in the ohm fraction among ( u)lirgs , it remains unclear on a case - by - case basis what conditions found in starbursts drive or prohibit ohm formation . here we describe a dense gas trigger for ohm formation , at last identifying physical observable properties that differentiate ohms from nonmasing mergers . we identify ohms , oh absorbers , oh kilomasers , and oh non - detections in the gao & solomon ( 2004a ; hereafter gs04a ) sample ( [ sec : sample ] ) and employ co(@xmath9 ) and hcn(@xmath9 ) molecular gas tracers to show that while oh absorbers appear nearly uniformly distributed in @xmath10 and @xmath5 , ohms represent the _ majority _ of the nonlinear population in the ir - co relation ( [ sec : results ] ) . in combination with a kennicutt - schmidt - based star formation model of co line emission by @xcite , we identify a high mean molecular density driving ohm emission , and from the hcn / co ratio we find that ohm galaxies are exclusively high dense gas fraction starbursts ( [ sec : results ] ) . now that we can at last observe quantities that are highly predictive of ohm activity , we can employ ohms at high redshifts as probes of major galaxy mergers and extreme star formation ( [ sec : discussion ] ) . the somewhat diverse gs04a hcn sample that forms the basis for this study includes most ir- and co - bright galaxies ( by flux ) and most local northern ulirgs ( @xmath11 km s@xmath12 ) . table [ tab : hcnsample ] lists basic properties and line luminosities of all galaxies in the sample that have been observed in the 1667 mhz oh line by various groups . the hcn sample includes 8 ohms , 12 oh absorption systems , 4 oh kilomasers , and 40 oh nondetections . while the division between oh kilomasers and ohms at @xmath13 is rather arbitrary , the @xmath14 values for the oh kilomasers are well separated from the ohms in table [ tab : hcnsample ] by 2 orders of magnitude . three of the four oh kilomasers in this sample show both emission and absorption . oh types marked with an asterisk are somewhat uncertain and have been omitted from all subsequent analysis and figures . we have included in figure [ fig : ir_vs_co ] four additional ohms that have been detected in co by @xcite but have not yet been observed in hcn : iras 03521 + 0028 , 14070 + 0525 , 16090@xmath150139 , and 18368 + 3549 . sorting the gs04a sample by oh type megamaser , kilomaser , absorber , or non - detection reveals striking properties of ohm host galaxies that set them apart from other starburst galaxies . figure [ fig : ir_vs_co ] shows that ohms comprise the majority of the population that is offset from the linear @xmath10-@xmath16 relation . oh absorbers and kilomasers , however , generally follow the linear ir - co relation . using the relationship between @xmath17 and the mean h@xmath18 density , @xmath19 , derived from kennicutt - schmidt laws by ( * ? ? ? * fig.2 ) , we show in figure [ fig : ir_vs_co ] that all ohms in the sample are produced in starburst volumes of @xmath20@xmath21 pc@xmath22 ( radii @xmath23130290 pc ) and that all but one ohm have extremely high volume - averaged molecular densities , @xmath24@xmath1 @xmath2 . in fact , 7 of 10 objects in the hcn sample in this density range are ohms and one is a nonabsorbing oh kilomaser . note that @xmath19 is the _ h@xmath18 density ; the clouds responsible for ohms within these regions must be significantly denser than the mean . there are many nonmasing systems with high @xmath10 at lower densities and larger volumes than the ohms , demonstrating that molecular density , not the ir radiation field , is the ohm trigger . equally striking is the segregation of ohms from nonmasing starbursts in a plot of @xmath10 versus @xmath25 , a proxy for dense molecular gas fraction @xcite . figures [ fig : ir_vs_dense ] and [ fig : hists ] show that all 8 ohms ( and the nonabsorbing oh kilomaser ) have @xmath26 . there are also 2 oh absorbers and 5 oh nondetections this regime , so ohms comprise roughly half of this unusual population . there are no ohms with @xmath27 , but there are many other luminous systems , including oh absorption systems and oh absorbers with coincident kilomaser emission . galaxies with oh absorption are fairly uniformly distributed in @xmath10 and dense gas fraction , indicating that oh abundance is not a factor in ohm formation . the @xmath28 fraction of ohms in dense starbursts constrains ohm beaming be a minor contributor to the ohm fraction in lirgs , and we conclude that the ohm emission solid angle must be greater than @xmath6 steradians . a dense gas trigger for ohm formation is consistent with the modeling work by @xcite showing that the critical component for ohm formation is cloud - cloud overlap ; a starburst - scale high mean molecular density and high dense gas fraction both provide the required overlap of many dense clouds . what is not yet clear is whether ohm activity is a density effect or simply a concentration effect . the rough size scales bracketing all ohms in figure [ fig : ir_vs_co ] are also consistent with the 100200 pc ohm emission regions observed with vlbi ( e.g. , * ? ? ? * ; * ? ? ? it is somewhat surprising that significant differences between ohm hosts and nonmasing systems are seen at all in unresolved observations because masing nuclei are `` contaminated '' by nonmasing nuclei within mergers . we expect that the observed differences between ohms and nonmasing starbursts would intensify in resolved observations . however , high dipole moment molecules such as hcn and cs may be good ohm location - selective tracers that obviate the need to obtain subarcsecond resolution . it is certain that the ohm phase is transitory because star formation rates found in these systems are sustainable for @xmath20@xmath21 yr , whereas the complete merging process requires of order @xmath29 yr . what is not known , however , is whether this mode of star formation is a universal , inevitable stage or an uncommon event in major galaxy mergers . tidal torques spike multiple times during mergers , and many major mergers are likely to experience a ulirg phase , but will most mergers experience the even more extreme ohm phase ? the simple observation that about @xmath30 of lirgs with @xmath31 show ohm activity @xcite suggests that if _ all _ lirgs experience an ohm stage , then the ohm lifetime is of order @xmath30 of the lirg lifetime . if only a subset of lirgs have an ohm stage , then the ohm lifetime must be longer . it is also possible that ohms draw from a larger `` pool '' of galaxies that begin at lower @xmath32 than the lirg sample , which would allow a shorter ohm lifetime . constraints on the ohm lifetime are clearly critical to understanding ohms and the modes of star formation in major galaxy mergers . we have identified three closely related physical properties that differentiate ohms from other starburst galaxies : ohm hosts have the highest mean molecular gas densities , they are a distinct population in the nonlinear part of the ir - co relation , and they reside in galaxies with exceptionally high dense molecular gas fractions . we conclude that molecular gas must be concentrated and massive in order to reach the mean density required to form an ohm in a galactic nucleus . ir luminosity is not a condition for ohm formation ; both star formation and ohm activity are consequences of the tidal density enhancements accompanying galaxy interactions . the fraction of ohms in dense starbursts constrains ohm beaming to be a minor effect : ohm solid angle emission must be greater than @xmath6 steradians . these conclusions are in good agreement with the stochastic cloud - cloud overlap amplification model by @xcite . the rather uniform distribution of oh absorbers in ir , hcn , and co luminosity suggests that oh abundance is not a significant factor in ohm formation . the main caveat to these conclusions is that the sample of ohms with hcn observations remains small , and should be expanded , particularly to higher redshifts to include `` typical '' ohms . ohms are signposts of the most intense , compact , and unusual modes of star formation in the local universe , and surveys for ohms will now provide detailed information about the detected host galaxies and their mode of star formation . the missing datum required for a complete interpretation of ohm surveys , however , is the ohm lifetime . this work benefited significantly from comments by the anonymous referee . 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 . baan , w. a. , haschick , a. d. , buckley , d. & schmelz , j. t. 1985 , , 293 , 394 baan , w. a. , haschick , a. d. , & schmelz , j. t. 1985 , , 298 , l51 baan , w. a. , haschick , a. d. , & henkel , c. 1989 , , 346 , 680 baan , w. a. 1989 , , 338 , 804 baan , w. a. , haschick , a. , & henkel , c. 1992 , , 103 , 728 1998 , , 509 , 633 2006 , , 449 , 559 bottinelli , l. , gouguenheim , l. , le squeren , a. m. , martin , j. m. , & paturel , g. 1990 , iau circ . 4977 2002a , , 124 , 100 2006 , , 132 , 2596 gallimore , j. f. , baum , s. a. , odea , c. p. , brinks , e. , & pedlar , a. 1996 , , 462 , 740 gao , y. & solomon , p. m. 2004a , , 152 , 63 ( gs04a ) gao , y. & solomon , p. m. 2004b , , 606 , 271 haschick , a. d. & baan , w. a. 1985 , nature , 314 , 144 henkel , c. & wilson , t. l. 1990 , , 229 , 431 krumholz , m. r. & thompson , t. a. 2007 , , submitted ( astro - ph/0704.0792 ) 2005 , , 43 , 625 lockett , p. , & elitzur , m. 2007 , , submitted martin , j. m. , le squeren , a. m. , bottinelli , l. , gouguenheim , l. , & dennefeld , m. 1988 , , 201 , l13 martin , j. m. , le squeren , a. m. , bottinelli , l. , gouguenheim , l. , & dennefeld , m. 1989 , , 208 , 39 nguyen - q - rieu , mebold , u. , winnberg , a. , guibert , j. , & booth , r. 1976 , , 52 , 467 norris , r. p. , gardner , f. f. , whiteoak , j. b. , allen , d. a. , & roche , p. f. 1989 , , 237 , 673 2005 , , 443 , 383 pihlstrm , y. m. , conway , j. e. , booth , r. s. , diamond , p. j. , & polatidis , a. g. 2001 , , 377 , 413 rovilos , e. , diamond , p. , lonsdale , c. j. , lonsdale , c. j. , & smith , h. e. 2003 , , 342 , 373 rickard , l. j. , turner , b. e. , & bania , t. m. 1982 , , 252 , 147 schmelz , j. t. , baan , w. a. , haschick , a. d. , & eder , j. 1986 , , 92 , 1291 schmelz , j. t. & baan , w. a. 1988 , , 95 , 672 solomon , p. m. , downes , d. , radford , s. j. e. , & barrett , j. w. 1997 , , 478 , 144 staveley - smith , l. , norris , r. p. , chapman , j. m. , allen , d. a. , whiteoak , j. b. , & roy , a. l. 1992 , , 258 , 725 unger , s. w. , chapman , j. m. , cohen , r. j. , hawarden , t. g. , & mountain , c. m. 1986 , , 220 , 1p 2005 , , 364 , 99 whiteoak , j. b. & gardner , f. f. 1973 , , 15 , 211 llccccccc 00450@xmath152533 & ngc253 & 2.5 & 2.1 & 4.6 & 0.27 & abs+kilo & ( @xmath151.3 ) & 1 + 01053@xmath151746 & ic1623 & 81.7 & 46.7 & 130.5 & 8.5 & non & & 2 + 01219 + 0331 & ngc520 & 31.1 & 8.5 & 16.3 & 0.64 & abs+kilo & ( @xmath150.4 ) & 3 + 01403 + 1323 & ngc660 & 14.0 & 3.7 & 7.3 & @xmath330.26 & abs & & 4 + 02071 + 3857 & ngc828 & 75.4 & 22.4 & 58.5 & 1.3 & non & & 2 + 02193 + 4207 & ngc891 & 10.3 & 2.6 & 11.0 & 0.25 & non & & 2 + 02360@xmath150653 & ngc1022 & 21.1 & 2.6 & 4.2 & 0.20 & non & & 2 + 02391 + 0013 & ngc1055 & 14.8 & 2.1 & 13.3 & @xmath340.37&non & & 5 + 02401@xmath150013 & ngc1068 & 16.7 & 28.3 & 20.7 & 3.61 & kilo & ( @xmath150.3 ) & 6 + 02526@xmath150023 & ngc1144&117.3 & 25.1 & 108.9 & 2.67 & non & & 2 + 03317@xmath153618 & ngc1365 & 20.8 & 12.9 & 58.7 & 3.10 & non * & & 7 + 03419 + 6756 & ic342 & 3.7 & 1.4 & 9.5 & 0.47 & non & & 2 + 04315@xmath150840 & ngc1614 & 63.2 & 38.6 & 24.5 & 1.25 & non & & 2 + 05083 + 7936 & viizw31&223.4 & 87.1 & 125.0 & 9.8 & non & & 2 + 05189@xmath152524 & & 170.3 & 118.1 & 67.0 & 6.2 & non * & & 7 + 06106 + 7822 & ngc2146 & 15.2 & 10.0 & 12.5 & 0.96 & non & & 2 + 07101 + 8550 & ngc2276 & 35.5 & 6.2 & 10.2 & 0.40 & non & & 2 + 09126 + 4432 & arp55 & 162.7 & 45.7 & 125.0 & 3.8 & non & & 2 + 09293 + 2143 & ngc2903 & 6.2 & 0.83 & 2.3 & @xmath330.09 & non & & 2 + 09320 + 6134 & ugc05101&160.2 & 89.2 & 50.8 & 10.0 & ohm & 1.61 & 8 + 09517 + 6954 & m82 & 3.4 & 4.6 & 5.7 & 0.30 & abs+kilo & ( @xmath151.7 ) & 9 + 09585 + 5555 & ngc3079 & 16.2 & 4.3 & 24.0 & @xmath231.0 & abs & & 10 + 10566 + 2448 & & 173.3 & 93.8 & 61.5 & 10.2 & non & & 2 + 11010 + 4107 & arp148 & 143.3 & 36.5 & @xmath3347.0 & 4.0 & ohm & 1.98 & 11 + 11085 + 5556 & ngc3556 & 10.6 & 1.35 & @xmath334.5&@xmath330.09 & non & & 2 + 11176 + 1315 & ngc3627 & 7.6 & 1.26 & 4.4 & @xmath330.08 & non & & 12 + 11176 + 1351 & ngc3628 & 7.6 & 1.01 & 7.1 & 0.24 & abs & & 4 + 11257 + 5850 & arp299 & 43.0 & 62.8 & 29.0 & 2.1 & ohm & 1.38 & 13 + 11460 + 4859 & ngc3893 & 13.9 & 1.15 & 4.1 & 0.23 & non & & 2 + 11578@xmath150049 & ngc4030 & 17.1 & 2.14 & 15.2 & 0.54 & non * & & 2 + 11596 + 6224 & ngc4041 & 18.0 & 1.70 & 3.9 & 0.18 & non & & 2 + 12239 + 3129 & ngc4414 & 9.3 & 0.81 & 4.6 & 0.16 & non & & 5 + 12396 + 3249 & ngc4631 & 8.1 & 2.0 & 2.3 & @xmath230.08&non & & 5 + 12540 + 5708 & mrk231 & 170.3 & 303.5 & 82.2 & 18.6 & ohm & 2.87 & 14 + 12542 + 2157 & ngc4826 & 4.7 & 0.26 & 1.3 & @xmath330.04 & non & & 12 + 13025@xmath154911 & ngc4945 & 3.7 & 2.6 & 5.8 & @xmath230.27 & abs & & 1 + 13086 + 3719 & ngc5005 & 14.0 & 1.4 & 8.2 & 0.41 & non * & & 2 + 13135 + 4217 & ngc5055 & 7.3 & 1.1 & 8.6 & @xmath330.10 & non & & 2 + 13183 + 3423 & arp193 & 92.7 & 37.3 & 39.8 & 9.5 & abs & & 5 + 13229@xmath152934 & ngc5135 & 51.7 & 13.8 & 31.3 & 2.73 & non * & & 7 + & m51 & 9.6 & 4.2 & 19.4 & 0.50 & non & & 2 + 13341@xmath152936 & m83 & 3.7 & 1.4 & 8.1 & 0.35 & non * & & 15 + 13428 + 5608 & mrk273 & 152.2 & 129.9 & 65.0 & 15.2 & ohm & 2.53 & 14 + 14306 + 5808 & ngc5678 & 27.8 & 3.0 & 17.2 & 0.75 & non & & 2 + 14376@xmath150004 & ngc5713 & 24.0 & 4.2 & 8.1 & 0.22 & non & & 2 + 14514 + 0344 & ngc5775 & 21.3 & 3.8 & 10.9 & 0.57 & abs * & & 5 + 15327 + 2340 & arp220 & 74.7 & 140.2 & 78.5 & 9.2 & ohm & 2.58 & 13 + 16504 + 0228 & ngc6240 & 98.1 & 61.2 & 79.0 & 11.0 & abs & & 3 + 17208@xmath150014 & & 173.1 & 234.5 & 146.9 & 37.6 & ohm & 3.02 & 16 + 18293@xmath153413 & & 72.1 & 53.7 & 85.5 & 4.03 & non & & 7 + 18425 + 6036 & ngc6701 & 56.8 & 11.2 & 34.0 & 1.38 & abs & & 17 + & ngc6921 & 60.3 & 11.4 & 17.5 & @xmath232.81 & non & & 5 + 20338 + 5958 & ngc6946 & 5.5 & 1.6 & 9.2 & 0.49 & non & & 2 + 21453@xmath153511 & ngc7130 & 65.0 & 21.4 & 44.9 & 3.27 & non & & 18 + 22132@xmath153705 & ic5179 & 46.2 & 14.1 & @xmath2326.4&3.42&non * & & 7 + 22347 + 3409 & ngc7331 & 15.0 & 3.5 & @xmath3310.7&@xmath330.44 & non & & 2 + 23007 + 0836 & ngc7469 & 67.5 & 40.7 & 37.1 & 2.19 & abs * & & 5 + 23024 + 1203 & ngc7479 & 35.2 & 7.4 & 26.7 & 1.12 & non & & 5 + 23365 + 3604 & & 266.1 & 142.0 & 85.0 & 15.0 & ohm & 2.45 & 19 + 23488 + 1949 & ngc7771 & 60.4 & 21.4 & 90.8 & 6.5 & non & & 5 + 23488 + 2018 & mrk331 & 75.3 & 26.9 & 52.1 & 3.35 & non & & 2 +	hcn and co line diagnostics provide new insight into the oh megamaser ( ohm ) phenomenon , suggesting a dense gas trigger for ohms . we identify three physical properties that differentiate ohm hosts from other starburst galaxies : ( 1 ) ohms have the highest mean molecular gas densities among starburst galaxies ; nearly all ohm hosts have @xmath0@xmath1 @xmath2 ( oh line - emitting clouds likely have @xmath3 @xmath2 ) . ( 2 ) ohm hosts are a distinct population in the nonlinear part of the ir - co relation . ( 3 ) ohm hosts have exceptionally high dense molecular gas fractions , @xmath4 , and comprise roughly half of this unusual population . oh absorbers and kilomasers generally follow the linear ir - co relation and are uniformly distributed in dense gas fraction and @xmath5 , demonstrating that ohms are independent of oh abundance . the fraction of non - ohms with high mean densities and high dense gas fractions constrains beaming to be a minor effect : ohm emission solid angle must exceed @xmath6 steradians . contrary to conventional wisdom , ir luminosity does not dictate ohm formation ; both star formation and ohm activity are consequences of tidal density enhancements accompanying galaxy interactions . the ohm fraction in starbursts is likely due to the fraction of mergers experiencing a temporal spike in tidally driven density enhancement . ohms are thus signposts marking the most intense , compact , and unusual modes of star formation in the local universe . future high redshift ohm surveys can now be interpreted in a star formation and galaxy evolution context , indicating both the merging rate of galaxies and the burst contribution to star formation .
the study of finite type invariants of virtual knots ( see @xcite ) has led to interest in gauss diagrams ( circles with arrows joining distinct pairs of points on it ) and arrow diagrams , which can be thought of as formal sums of gauss diagrams modulo relations that correspond to the virtual reidemeister moves . it can be shown that a degree-@xmath0 finite type invariant gives rise to a functional on arrow diagrams with @xmath0 arrows . in this paper we study such functionals coming from lie bialgebras which are themselves constructed from the classical lie algebras . we begin with the following definitions . let @xmath1 be a disjoint union of oriented line segments and circles . 1 . a chord diagram with skeleton @xmath1 is a diagram with chords joining distinct pairs of points on @xmath1 . 2 . an oriented chord diagram is a chord diagram together with an orientation on each of its chords . a trivalent diagram with skeleton @xmath1 is @xmath1 together with a graph @xmath2 with univalent and trivalent vertices so that all the univalent vertices are attached to distinct points on @xmath1 , and each trivalent vertex comes with an orientation ( i.e. an ordering of the incident edges modulo cyclic permutations ) . 4 . an arrow diagram is a trivalent diagram with an orientation on each of the edges of @xmath2 . an arrow diagram is acyclic if the underlying graph @xmath2 contains no oriented cycles . we consider the vector space generated by oriented chord diagrams modulo the relation 6 t . ( for origins see @xcite . ) it is depicted in figure [ fig:6 t ] . similarly we consider the vector space generated by arrow diagrams modulo the relations @xmath3 ( antisymmetry ) , @xmath4 ( no sink , no source ) , @xmath5 and @xmath6 in the space of arrow diagrams . ( see figures [ fig : as ] to [ fig : ihx ] . ) @xmath7 is the vector space of oriented chord diagrams on @xmath1 modulo 6 t . @xmath8 is the vector space of arrow diagrams on @xmath1 modulo @xmath3 , @xmath4 , @xmath5 and @xmath6 . from now on we will use ` oriented chord diagrams ' and ` arrow diagrams ' to refer to equivalent classes in @xmath7 and @xmath8 , respectively . a functional from @xmath7 or @xmath8 to @xmath9 is called a weight system . our goal is to study weight systems on @xmath7 . polyak ( @xcite ) proved that the space of oriented chord diagrams modulo 6 t is isomorphic to the subspace of @xmath8 generated by acyclic arrow diagrams . we will first study functionals on @xmath8 , which are easier to construct since they have a close realtion to lie bialgebras . given lie bialgebras constructed from classical lie algebras we will then give algorithms to compute the corresponding weight systems on diagrams in @xmath7 in section [ sec : bialg ] we will follow @xcite to construct a manin triple from a simple lie algebra . in section [ sec : tensor ] we follow @xcite to assign tensors to elements of @xmath8 and construct weight systems when the skeleton is a circle . in section [ sec : calc ] , using manin triples coming from the families @xmath10 , @xmath11 and @xmath12 , we give explicit formulas to turn @xmath7 into tensors . we also introduce the averaging map and compose it with our tensors and compare the results with those in @xcite on unoriented diagrams . finally in section [ sec : sample ] we apply the results from section [ sec : calc ] and do some sample calculations on @xmath13 . this paper is part of the author s ph.d . research at the university of toronto under the supervision of dror bar - natan . the author would like to thank him for his guidance and suggestions in the writing of this paper . a lie bialgebra @xmath14,\delta ) $ ] is a lie algebra @xmath14)$ ] equipped with an antisymmetric @xmath15 map @xmath16 satisfying the cojacobi identity @xmath17 and the cocycle condition @xmath18)=ad_{x}(\delta y)-ad_{y}(\delta x),\ ] ] for any @xmath19 , where @xmath20 is the cyclic permutation on @xmath21 . a finite dimensional manin triple is a triple of finite dimensional lie algebras @xmath22 , where @xmath23 is equipped with a metric ( a symmetric nondegenerate invariant bilinear form ) @xmath24 such that 1 . @xmath25 as a vector space and @xmath26 are lie subalgebras of @xmath23 . 2 . @xmath26 are isotropic with respect to @xmath24 . as a consequence , @xmath27 and @xmath28 are maximal isotropic subalgebras . suppose @xmath22 is a manin triple . the metric then induces a nondegenerate pairing @xmath29 , and hence a lie algebra structure on @xmath30 . let @xmath31 be the induced coalgebra structure on @xmath27 . we can check by direct computation ( @xcite ) that the cocycle condition is satisfied . @xmath32,\delta)$ ] is therefore a lie bialgebra . in fact the process can be reversed . given a lie bialgebra @xmath33 , we may define a symmetric nondegenerate bilinear form @xmath34 by @xmath35 . if @xmath36 is a basis of @xmath33 and @xmath37 is a basis of @xmath38 with @xmath39=c_{ij}^{k}e_{k}$ ] and @xmath40=\gamma^{rs}_{t}f^{t}$ ] , then we can define a lie algebra structure on @xmath41 by @xmath42=c^{r}_{st}f^{t}-\gamma^{rt}_{s}e_{t } \label{eqn : bracket}\ ] ] and keeping the bracket for @xmath33 and @xmath38 , with the consequence that @xmath34 is invariant . ( see section 1.3 of @xcite . ) @xmath34 is therefore a metric and @xmath43 is a manin triple . there is a standard way to obtain manin triples from simple lie algebras , and those are the ones we are going to use . the construction below follows chapter 4 of @xcite . given a simple lie algebra over @xmath9 with metric @xmath24 , we fix a cartan subalgebra @xmath44 and consider the corresponding positive and negative root spaces @xmath45 and @xmath46 . for each root @xmath47 we consider @xmath48 and @xmath49 ( where @xmath50 are the root spaces corresponding to @xmath51 ) such that @xmath52 . let @xmath53 $ ] . we consider the lie algebra @xmath54 where @xmath55 and with bracket defined by : @xmath56=0 $ ] , @xmath57=\alpha ( h)e_{\alpha}$ ] , + @xmath58=-\alpha ( h)f_{\alpha}$ ] , and @xmath59=\frac{1}{2}(h^{(1)}_{\alpha}+h^{(2)}_{\alpha})$ ] . we define the following metric on @xmath23 : @xmath60 we can check that @xmath61 is a manin triple . in fact @xmath23 is a lie bialgebra with r - matrix @xmath62 i.e. , @xmath63 , where @xmath64 is an orthonormal basis of @xmath44 with respect to @xmath24 . we define the projection @xmath65 where @xmath66@xmath67 this map endows @xmath33 with a quasitriangular lie bialgebra structure with r matrix @xmath68 so @xmath69 . the lie subalgebras @xmath70 and @xmath71 are lie subbialgebras . note that the map @xmath72 is a lie algebra homomorphism . if @xmath33 is given as a matrix lie algebra then @xmath72 is a representation . since polyak ( @xcite ) has proved that @xmath73 modulo 6 t is isomorphic to subspace of @xmath8 generated by the acyclic diagrams , a functional on @xmath8 induces a functional on @xmath73 . in this section we construct a funtional on @xmath8 by following haviv s method of assigning lie tensors to directed trivalent graphs and a representation to the skeleton ( @xcite ) and then calculate the trace . let @xmath33 be a lie bialagebra with basis \{@xmath74 } and let \{@xmath75 } be the dual basis of @xmath38 . we would like to assign to each directed trivalent graph a tensor in @xmath76 . first to the straight arrow we assign the element @xmath77 , which coresponds to the identity map . given the ns relation we have two types of vertices ( one in , two out and two in , one out ) . for the first type we assign to it the cobracket tensor @xmath78 , while to the second we assign the bracket tensor @xmath79=c^{k}_{ij}(f^{i}\otimes f^{j}\otimes e_{k})$ ] , where @xmath80 and @xmath81 are the structure constants for the bracket and the cobracket , respectively . ( see figure [ fig : bracket ] . ) it is worth noting that under this interpretation , the stu becomes the diagrammatic way of saying @xmath82=xy - yx$ ] in the universal enveloping algebra of @xmath23 ( see equation [ eqn : bracket ] ) , the 3-term ihx relations become the jacobi and cojacobi identities , and the 5-term ihx becomes the cocycle identity . to assign a tensor to a directed trivalent graph , we break it down into subgraphs with 0 or 1 vertex and at the points of gluing we contract using the metric . given a representation @xmath83 where @xmath84 has a basis @xmath85 , if the skeleton is part of the picture , we assign greek letters ranging over @xmath86 to each section of the skeleton . for example in figure [ fig : repexample ] , we assign @xmath87 and @xmath88 , respectively , to the pictures . note , since @xmath89 where @xmath90 is the inner product of @xmath91 with respect to the given basis , the same values can be written as @xmath92 and @xmath93 . here we do not distinguish @xmath94 or @xmath95 from its matrix with respect to the basis @xmath96 . if we restrict ourselves to the case where the skeleton is a circle , we can see that the construction above gives us the trace of the tensor in the given representation . in this section we will calculate weight systems coming from manin triples constructed from classical lie algebras . the metric in each of the lie algebras is @xmath97 . throughout this section @xmath98 or @xmath99 is the matrix whose @xmath100-th entry is 1 and zero everywhere else . let @xmath72 be the map given in section [ sec : bialg ] . we think of it as a representation and let @xmath101 and @xmath102 . we begin with @xmath103 . note that @xmath103 is not simple , but our construction in section [ sec : bialg ] , when applied to @xmath104 , results in the manin triple @xmath105 , where @xmath106 is the lie algebra of ( non - strictly ) upper triangular matrices with trace 0 and its dual @xmath107 is the lower triagular matrices with trace 0 . this implies that both @xmath106 and @xmath107 are lie bialgebras . let @xmath108 be the commutative lie algebra of scalar matrices , then @xmath109 is the lie algebra of upper triangular matrices . in fact it is a lie bialgebra if we set @xmath110 for any @xmath111 . now following section [ sec : bialg ] we consider @xmath112 where @xmath113 and consider the metric @xmath114 and the bracket @xmath115'$ ] on @xmath116 defined as follows : @xmath117 @xmath118'=[\mathfrak{u},\mathfrak{l}]_{\tilde{\mathfrak{u}}}$]@xmath119=0 $ ] for @xmath120 and any @xmath121 , where @xmath122 and @xmath115_{\tilde{\mathfrak{u}}}$ ] are the metric and the bracket in @xmath123 , respectively . the dual to @xmath124 under @xmath114 is then @xmath125 . if we let @xmath126 and @xmath127 then we consider @xmath128 to be the manin triple constructed from @xmath103 . in this paper we use @xmath103 instead of @xmath104 . the map @xmath72 is defined similarly as in section [ sec : bialg ] . we consider the tensor in figure [ fig : glnonearrow ] . note that @xmath129 forms a basis of @xmath33 and @xmath130 ( where @xmath131 ) forms the corresponding dual basis of @xmath38 under @xmath24 . using haviv s method we assign @xmath132 to each chord ( assuming the einstein summation convention ) . we have @xmath133 and @xmath134 ( so @xmath135 and @xmath136 , and similarly @xmath137 and @xmath138 , are the same matrices sitting in different spaces ) . the tensor is therefore given by @xmath139 @xmath140 the result can be expressed diagramatically as in figure [ fig : gln ] , where the dotted equality sign means that when @xmath141 , the resulting value is @xmath142 . for @xmath143 we use the basis @xmath144 and @xmath145 where @xmath146 and @xmath147 . we have : @xmath148 note this is not exactly a basis because @xmath149 but that only means in the calculation of @xmath150 we have some extra zero terms . @xmath150 in this case is given by @xmath151 in the expression above the first product is given by the case when @xmath152 while the second product corresponds to the case @xmath153 . we consider the first product . letting @xmath154 be the set @xmath155 and @xmath156 the set @xmath157 , we can assign @xmath154 or @xmath156 to each of @xmath47 , @xmath158 , @xmath159 and @xmath160 to indicate which set it belongs to . the four combinations that correspond to the four ways to expand the first summand in the last equation are as given in figure [ fig : so2nabk1 ] . starting at the top left corner and going clockwise the four pictures correspond to @xmath161 , @xmath162 , @xmath163 and @xmath164 , respectively . if we delete the arrow and join two greek letters to indicate that they are equal ( or equal modulo @xmath165 , if one comes from @xmath154 and the other from @xmath156 ) , then the first summand can be expressed diagramatically as in figure [ fig : so2nk1 ] . the inequality comes from the fact that @xmath166 and again the dotted equality means that when @xmath167 the resulting value comes with an extra factor of @xmath142 . now we look at the product corresponding to the case @xmath153 . the only possible assignment is given in figure [ fig : so2nabk2 ] . similarly we can represent the second summand diagramatically . see figure [ fig : so2nk2 ] . in this case the equality may be weighted with any factor because when @xmath167 we get zero . if we choose the factor to be @xmath142 , as we did in the beginning of this subsection , diagramatically the @xmath143 @xmath150 tensor may be expressed as in figure [ fig : so2n ] . like in @xmath143 we use the basis @xmath144 and @xmath145 where @xmath146 and @xmath169 , but the signs of some entries are different . we have : @xmath170 if we look at the tensor @xmath150 we get : @xmath171 the factor in the second product is due to the fact that @xmath172 . like in the case of @xmath143 we assign letters @xmath154 and @xmath156 to the four corners to indicate which set each of @xmath47 , @xmath158 , @xmath159 and @xmath160 belongs to . the only possible assignments are the same as the ones for @xmath143 . ( see figures [ fig : so2nabk1 ] and [ fig : so2nabk2 ] for the cases @xmath152 and @xmath153 , respectively . ) following the steps in the @xmath143 case we can see that the @xmath168 tensor may be given as in figure [ fig : spn ] . note that it is almost the tensor for @xmath143 except for some sign difference . we use the basis @xmath144 and @xmath145 where @xmath146 and @xmath174 . we have : @xmath175 the @xmath150 tensor is then given as @xmath176 using @xmath177 , @xmath154 and @xmath156 to denote the sets @xmath178 , @xmath179 and @xmath180 respectively , we can follow the same steps as in @xmath143 to get a diagrammatic representation of @xmath150 . the result is given in figure [ fig : so2n1 ] . the averaging map @xmath181 takes a chord diagram to an arrow diagram by summing over all possible ways to direct each chord . ( see figure [ fig : averaging ] for an example . ) let @xmath182 be the vector space of unoriented chord diagrams on @xmath1 modulo the 4 t relation . ( see @xcite and @xcite . ) if we have a weight system @xmath183 , then composing it with the averaging map @xmath181 gives us a map @xmath184 which satisfies the 4 t relation and is thus a weight system on @xmath185 . ( 4 t is a consequence of 6 t by repeatingly applying the averaging map . ) let @xmath33 be a lie algebra in the family @xmath10 , @xmath12 or @xmath11 and @xmath186 be the manin triple obtained from @xmath33 by the standard construction given in section [ sec : bialg ] . also let @xmath187 and @xmath188 be the weight systems they give rise to . then @xmath189 the equality occurs at the tensor level . we will look at each case separately . for @xmath10 the unoriented tensor is calculated as in figure [ fig : glnunori ] . note the absence of restrictions on the values of unconnected greek letters give us the weight system as given in @xcite . for @xmath168 we have figure [ fig : spnunori ] , which again is the weight system given in @xcite . for @xmath143 , we have figure [ fig : so2nunori ] . we consider a new basis @xmath190 so that @xmath191 of @xmath192 , where @xmath193\ ] ] and @xmath194 is the @xmath195 identity matrix . note @xmath196 is a unitary matrix , so @xmath197 where @xmath90 is the inner product with respect to the basis @xmath198 . using this new basis , whenever an @xmath154 appears at the tail of an arrow which is part of the skeleton we can replace it by @xmath199 . similarly we can replace @xmath156 by @xmath200 . if they appear at the head of an arrow that is part of the skeleton , however , we replace @xmath154 and @xmath156 by @xmath200 and @xmath199 , respectively , since the inner product is conjugate linear in the first argument . if we expand each of the diagrams linearly we obtain the same result as in @xcite . see figure [ fig : so2nunorichangebasis ] . for the case @xmath173 we use the change of basis matrix @xmath201\ ] ] and we obtain the same result after expansion and cancelling . see figure [ fig : so2n1unorichangebasis ] . we now do some sample calculations . in this section the skeleton is always oriented counterclockwise . first we calculate the two diagrams shown in figures [ fig : glncalc1 ] and [ fig : glncalc2 ] , using the @xmath103 weight system . for the first picture , each triple @xmath202 such that @xmath203 gives us a value of either 1 , @xmath142 or @xmath204 , depending on whether @xmath205 or @xmath206 . therefore the diagram should have weight @xmath207 , where @xmath181 is the number of triples @xmath208 such that @xmath209 , @xmath96 is the number of triples @xmath208 such that @xmath210 or @xmath211 , and @xmath212 is the number of triples @xmath208 such that @xmath213 . the number is now we calculate the weight of the picture in figure [ fig : so2ncalc ] using @xmath143 . we assign letters @xmath216 to each arc following the rules from section [ sec : calc ] ( see figure [ fig : so2ncalc ] ) . for each assignment we have one or four ways to resolve the diagram , and for a resolution with @xmath217 loops we count the number of @xmath217-tuples in @xmath218 such that the inequalities are satisfied , bearing in mind that each equality comes with a weight @xmath142 . the weights of the first two diagrams and the last two diagrams on the right hand side of the first equation in figure [ fig : so2ncalc ] are therefore @xmath219 and @xmath220 , respectively . the weight of the diagram is therefore finally we calculate the weight of the picture in figure [ fig : threearrows ] . according to section [ sec : calc ] we have ten possible ways to label the arcs on the skeleton ( figure [ fig : so2n1samplelabel ] ) . the weights of the ten diagrams ( first row left to right , then second row left to right ) are @xmath222 , @xmath223 , @xmath222 , @xmath165 , @xmath165 , @xmath224 , @xmath224 , @xmath225 , @xmath226 , and @xmath227 , respectively . summing these ten weights we get that the weight of figure [ fig : threearrows ] is @xmath228 .	in @xcite haviv gave a way of assigning lie tensors to directed trivalent graphs . weight systems on oriented chord idagrams modulo 6 t can then be constructed from such tensors . in this paper we give explicit combinatorial formulas of weight systems using manin triples constrcted from classical lie algebras . we then compose these oriented weight systems with the averaging map to get weight systems on unoriented chord diagrams and show that they are the same as the ones obtained by bar - natan ( @xcite ) . in the last section we carry out calculations on certain examples . [ 3][]#2 theorem # 3 _ _ * combinatorial formulas for classical lie weight systems on arrow diagrams * louis leung university of toronto [email protected]
modern astrophysical and cosmological data strongly indicate that a significant amount of matter in the universe exists in the form of _ dark matter_. the nature of dark matter ( dm ) remains completely unknown and its existence presents one of the major challenges to modern physics . it is commonly believed that the dark matter is made of particles . however , the standard model of elementary particles fails to provide a dm candidate . therefore the hypothesis of a particle nature of dm implies extension of the standard model . _ weakly interacting massive particles _ ( wimps ) are probably the most popular class of dm candidates . they appear in many extensions of the standard model ( see e.g. * ? ? ? * ; * ? ? ? these stable particles interact with the sm sector with roughly electroweak strength @xcite . to provide a correct dark matter abundance , wimps should have mass in the gev range . a significant experimental effort is devoted to the detection of the interaction of wimps in the galaxy s dm halo with laboratory nucleons ( _ direct detection experiments _ , see e.g. @xcite ) or to finding their annihilation signal in space ( _ indirect detection experiments _ , see e.g. @xcite ) . another large class of dm candidates are _ superweakly _ interacting particles ( so called _ super - wimps _ ) , i.e. particles , whose interaction strength with the sm particles is much more feeble than the weak one . super - wimps appear in many extensions of the standard model : extensions of the sm by right - handed neutrinos @xcite , supersymmetric theories @xcite , models with extra dimensions @xcite and string - motivated models @xcite . the feeble interaction strength makes the laboratory detection of super - wimps challenging ( see e.g. @xcite ) . on the other hand , many super - wimp particles possess a _ 2-body radiative decay _ channel : @xmath10 , producing a photon with energy @xmath11 . one can therefore search for the presence of such a monochromatic line with energy in the spectra of dm - containing objects @xcite . although the lifetime of any realistic decaying dm is much longer than the lifetime of the universe ( see e.g. @xcite ) , the huge amount of potentially decaying dm particles in a typical halo means a potentially detectable decay signal in the spectra of dm - dominated objects . searching for such a line provides a major way of _ detection of super - wimp dm particles_. the strategy of the search for decaying dark matter signal drastically differs from its annihilation counterpart . indeed , the _ decay _ signal is proportional to the _ dm column density _ integral of a dm distribution @xmath12 along the line of sight @xmath13 as opposed to @xmath14 in the case for annihilating dm . the dm column density varies slowly with the mass of dm objects @xcite and as a result a vast variety of astrophysical objects of different nature would produce a comparable decay signal @xcite . therefore without sacrificing the expected signal one has a freedom of choosing observational targets , avoiding complicated astrophysical backgrounds ; if a candidate line is found , its surface brightness profile may be measured ( as it does not decay quickly away from the centers of the objects ) and distinguished from astrophysical lines that usually decay in outskirts of galaxies and clusters . moreover , any tentative detection in one object would imply a signal of certain ( comparable ) signal - to - noise ratio from a number of other objects . this can be checked and the signal can either be unambiguously confirmed to be that of dm decay origin or ruled out . this allows to distinguish the decaying dm line from any possible astrophysical background and therefore makes astrophysical search for the decaying dm _ another type of a direct detection experiment_. in this paper we demonstrate the power of this approach . we check the recent conjecture by @xcite ( * ? ? ? * * lk09 * in what follows ) , who reported that a spectral feature at @xmath15 kev with the flux @xmath16 ( all errors at @xmath17 confidence level ) in the spectrum of the milky way satellite known as willman 1 @xcite may be interpreted as a dm decay line . according to lk09 , the line with such parameters is marginally consistent the restrictions on decaying dark matter from some objects ( see e.g. @xcite or the restrictions from the @xmath18 ks observation of ursa minor dwarf spheroidal in @xcite ) . the results of other works ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) are inconsistent at @xmath3 level with having @xmath19 of decaying dm with the best - fit parameters of lk09 . however , it is hard to exclude completely new unknown systematic uncertainties in the dm mass estimates and therefore in expected signal , from any given object . to this end , in this work we explicitly check for a presence of a line with parameters , specified above , by using archival observations by _ xmm - newton_and _ chandra _ of several objects , where comparable or stronger signal was expected . we compare the signals from several objects and show that the spatial ( angular ) behavior of the spectral feature is inconsistent with its dm origin so strongly , that systematic uncertainties can not affect this conclusion . we also discuss a possible origin of the spectral feature of lk09 . the particle flux of the decaying dark matter ( in photons @xmath20s ) ) is given by @xmath21 where @xmath22 is the mass of the dark matter particle , @xmath23 is the decay rate to photons ( depending on the particular model of decaying dark matter ) , @xmath24 is the dark matter mass within the instrument s field - of - view ( fov ) @xmath25 and @xmath26 is the luminous distance towards the object . ) is valid for all objects whose size is much smaller than @xmath26 . ] * willman 1 signal . * in lk09 the signal was extracted from the central @xmath27 of the willman 1 observation . assuming the dm origin of this signal , the observed flux @xmath28 corresponds to @xmath29 flux per unit solid angle . the mass within the fov of the observation of willman 1 was estimated in lk09 to be @xmath30 , using the best - fit values from @xcite . taking the luminous distance to the object to be @xmath31 , the average dm column density of willman 1 is @xmath32 ( see however discussion in the section [ sec : dwarf - spher - galax ] below ) . to check the presence of a dm decay line , we should look for targets with comparable signal . the signal - to - noise ratio for a weak line observed against a featureless continuum is given by ( see e.g. discussion in @xcite ) : @xmath33 where @xmath34 is the average dm column density within an instrument s field - of - view @xmath25 , @xmath35 is the exposure time , @xmath36 is an effective area of the detector and @xmath37 is the spectral resolution ( notice that the line in question has a much smaller intrinsic width than the spectral resolution of _ chandra _ and _ xmm - newton _ ) . for the purpose of our estimate we consider the effective areas of _ xmm - newton_mos cameras and _ chandra _ acis - i ( as well as their spectral resolution ) to be approximately equal , and effective area of pn camera is 2 times bigger . * galactic contribution . * the contribution of the milky way s dm halo along the line of sight should also be taken into account , when estimating the dm decay signal . using a pseudo - isothermal profile @xmath38 with parameters , adopted in @xcite ( @xmath39 , @xmath40 , @xmath41 ) , the corresponding milky way column density in the direction of willman 1 ( @xmath42 off galactic center ) is @xmath43 . as demonstrated in @xcite , the value of dm column density computed with this dm density profile coincides with the best - fit navarro - frenk - white ( * ? ? ? * * nfw * ) profiles of @xcite and @xcite within few @xmath44 for the off - center angle @xmath45 . other dark matter distributions produce systematically larger values of dark matter column density . for example , the best - fit pseudo - isothermal profile of @xcite would corresponds to an additional milky way contribution @xmath46 in the direction of willman 1 . the comparison between the dm column densities of several objects with that of milky way in their directions is shown in fig . [ fig : mw ] . for each _ xmm - newton_observation observation , we extracted both epic mos and pn spectra . we select events with ` pattern < = 12 ` for mos and ` pattern = = 0 ` for pn camera and use ` flag = = 0 ` . we used ` sas ` task ` edetect_chain ` to exclude point sources , detected with the likelihood values above 10 ( about 4@xmath47 ) . for epic pn camera , bright strips with out - of - time ( oot ) events were also filtered out , following the standard procedure . we imposed stringent flare screening criteria . to identify `` good time intervals '' we performed a light - curve cleaning , by analyzing temporal variability in the hard ( @xmath48 ) x - ray band ( as described e.g. in @xcite ) . the resulting light curves were inspected `` by eye '' and additional screening of soft proton flares was performed by employing a method similar to that described in e.g. @xcite ( i.e. by rejecting periods when the count rate deviated from a constant by more than @xmath49 ) . the extracted spectra were checked for the presence of residual soft proton contaminants by comparing count rates in and out of fov @xcite . in all spectra that we extracted the ratio of count rates in and out of fov did not exceed 1.13 which indicates thorough soft proton flare cleaning ( c.f . * ) . following the procedure , described in @xcite we extracted the filter - wheel closed background from the same region as signal ( in detector coordinates ) . the background was further renormalized based on the @xmath50 kev band count rates of the extracted spectra . the resulting source and background spectra are grouped by at least 50 counts per bin using ` ftool ` @xcite command ` grppha ` . cm^{-2}~s^{-1}~ { \ensuremath{\:\mathrm{kev}}}^{-1}})$ ] & @xmath51/dof + & & at 1 kev ( mos1 / mos2 / pn ) + + ` 0109270101 ` & 1.33 & 5.71 / 5.80 / 7.88 & 1328/1363 + ` 0112570101 ` & & 5.37 / 6.12 / 6.86 & + ` 0112570401 ` & & 6.13 / 6.66 / 6.37 & + + ` 0402560301 ` & 1.46 & 6.28 / 6.12 / 9.97 & 895/997 + + ` 0511380101 ` & 1.53 & 2.75 / 2.88 / 3.89 & 875/857 + ` 0109270401 ` & 1.47 & 5.02 / 4.41 / 5.22 & 606/605 + ` 0505760401 ` & 1.27 & 1.91 / 2.04 / 2.47 & 548/528 + ` 0505760501 ` & 1.28 & 2.25 / 1.85 / 3.03 & 533/506 + ` 0402561301 ` & 1.57 & 2.94 / 3.10 / 3.95 & 489/515 + ` 0402561401 ` & 1.47 & 4.33 / 3.58 / 4.47 & 751/730 + ` 0402560801 ` & 1.55 & 5.22 / 4.49 / 5.83 & 902/868 + ` 0402561501 ` & 1.44 & 2.70 / 4.29 / 6.67 & 814/839 + ` 0109270301 ` & 1.39 & 4.29 / 4.58 / 3.75 & 413/436 + + ` 0302500101 ` & 1.52 & 4.46 / 4.02 / 3.78 & 938/981 + the extracted spectra with subtracted instrumental background were fitted by the ` powerlaw ` model of ` xspec ` in the energy range 2.18.0 kev data for mos and 2.17.2 kev for the pn camera . this choice of the _ baseline model _ is justified by the fact that for all considered objects we expect no intrinsic x - ray emission above 2 kev and therefore the observed signal is dominated by the extragalactic diffuse x - ray background ( xrb ) with the powerlaw slope @xmath52 and the normalization @xmath53 at 1 kev ( 90% cl errors ) ( see e.g. @xcite ) . galactic contribution ( both in absorption and emission ) is negligible at these energies and we do not take it into account in what follows . for all considered observations ` powerlaw ` model provided a very good fit , fully consistent with the above measurements of xrb . the ` powerlaw ` index was fixed to be the same for all cameras ( mos1 , mos2 , pn ) , observing the same spatial region , while the normalization was allowed to vary independently to account for the possible off - axis calibration uncertainties between the different cameras , see e.g. @xcite . the best - fit values of the ` powerlaw ` index and normalizations are presented in table [ tab : best_fit_values ] . to check for the presence of the lk09 line in the spectra , we added a narrow ( @xmath54 ) gaussian line ( ` xspec ` model ` gaussian ` ) to the baseline ` powerlaw ` model . assuming the dm origin of this line , we fix its normalization at the level @xmath55 , derived in lk09 , multiplied by the ratio of total ( including milky way contribution ) dm column densities and instrument s field - of - view . we vary the position of the line in the interval corresponding to the @xmath3 interval of lk09 : 2.30 2.72 kev . ( [ eq:3 ] ) allows to estimate the expected significance of the line . when searching for weak lines , one should also take into account uncertainties arising from the inaccuracies in the calibration of the detector response and gain . this can lead to systematic residuals caused by calibration inaccuracies , at the level @xmath56 of the model flux ( some of them having edge - like or even line - like shapes ) ( see e.g. * ? ? ? * ; * ? ? ? * ) as well as discussion in @xcite for similar uncertainties in _ chandra_. to account for these uncertainties we perform the above procedure with the @xmath57 of the model flux added as a systematic error ( using ` xspec ` command ` systematic ` ) . the andromeda galaxy is the closest spiral galaxy to the milky way . its dark matter content has been extensively studied over the years ( see e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * and references therein ) for an incomplete list of recent works . the total dynamical mass ( out to @xmath58 kpc ) can be determined from the rotation curve measured from hi kinematics . the major uncertainty in determination of dark matter content is then related to a separation of contributions of baryonic ( stellar bulge and especially , extended stellar disk ) and dark components to the total mass . the baryonic mass is often obtained from the ( deprojected ) surface brightness profile ( optical or infrared ) , assuming certain mass - to - light ratio for the luminous matter in the bulge and the disk of a galaxy . @xcite ( * b08 * in what follows ) analyzed dark matter distributions of m31 , existing in the literature at that time in order to provide the most conservative estimate of the expected dm decay signal . the dm column density @xmath34 in more than 10 models from the works @xcite turns out to be consistent within a factor of @xmath59 for off - center distances greater than @xmath60 kpc ( see gray lines in fig.[fig : dm_distrib - m31 - 2010 ] ) . the most conservative estimate of dark matter column density was provided by one of the models of @xcite ( called in that work m31b ) , marked as red solid line on fig.[fig : dm_distrib - m31 - 2010 ] . recently two new hi surveys of the disk of m31 were performed @xcite and new data on mass distribution of m31 became available . in @xcite modeling of hi rotation curve between @xmath61 kpc and 38 kpc was performed . @xcite used _ r - band _ photometric information @xcite to determine the relative contribution of the stellar disk and the bulge . based on this information and taking into account foreground and internal extinction / reddening effects ( see section 8.2.2 in * ? ? ? * for details ) they determined mass - to - light ratio , using stellar population synthesis models @xcite @xmath62 ( in agreement with those of @xcite ) . aiming to reproduce the data in the inner several kpc , they explored different disk - bulge decompositions with two values of the bulge s mass - to - light ratio @xmath63 and @xmath64 ( all mass - to - light ratios are in solar units ) . @xcite analyzed several dark matter density profiles ( nfw , einasto , cored profile ) . according to their 6 best - fit models ( 3 dm density profiles times two choices of mass - to - light ratio ) the dm column density in the andromeda galaxy is higher than the one , adapted in b08 ( model m31b ) everywhere but inside the inner 1 kpc ( see red dashed line on fig.[fig : dm_distrib - m31 - 2010 ] ) . @xcite used circular velocity data , inferred from the hi rotation curves , also extending out to @xmath65 . the authors exclude from the analysis the inner 8 kpc , noticing the presence of structures in the inner region ( such as a bar ) , associated with non - circular motions . @xcite also use optical data of @xcite and determine an upper and lower bounds on the disk mass - to - light ratio using the same models @xcite as @xcite . they obtain possible range of values of the _ b - band _ mass - to - light ratio @xmath66 . these values do not take into account corrections for the internal extinction . ) . the authors of @xcite had chosen not to include any uniform extinction corrections due to the presence of a gradient of b - r index ( e. corbelli , private communication ) . ] @xcite fit the values of @xmath67 rather than fixing it to a theoretically preferred value . the mass - to - light ratio of the disk of a spiral galaxy is known to be poorly constrained in such a procedure since the contributions of the disk and dm halo are similar ( see e.g. discussions in @xcite ) . @xcite analyzed variety of dm models and mass - to - light ratios , providing good fit to the data . the mass model that maximizes the contribution of the stellar disk ( @xmath68 ) has the core of the burkert profile @xcite @xmath69 ( if one imposes additional constraint on the total mass of m31 within several hundred kpc ) . the dm column density remains in this model remains essentially flat from the distance @xmath70 inward ( see fig.[fig : dm_distrib - m31 - 2010 ] ) . notice , that the `` maximum disk '' fitting of @xcite ( blue dashed double - dotted line on fig.[fig : dm_distrib - m31 - 2010 ] ) has higher dm content . therefore in this work we will adopt the burkert dm model of @xcite as a _ minimal _ possible amount of dark matter , consistent with the rotation curve data on m31 . the corresponding dm column density is factor 23 lower than the one given by the previously adapted model m31b in the inner @xmath71 kpc . in what follows we will provide the restrictions for both m31b ( as the _ most conservative _ among physically motivated models of dm distribution in m31 ) and burkert model of @xcite . three observations of the central part of m31 ( obsids ` 0112570401 ` , ` 0109270101 ` and ` 0112570101 ` ) were used in b08 to search for a dark matter decay signal . after the removal of bright point sources , the diffuse spectrum was extracted from a ring with inner and outer radii @xmath27 and @xmath72 , centered on m31 @xcite ( at this distance @xmath73 corresponds to @xmath74 ) . ] ( see b08 for details , where this region was referred to as ` ring5 - 13 ` ) . we call these three observations collectively ` m31on ` in what follows . the baseline ` powerlaw ` model has the total @xmath51 1328 for 1363 d.o.f . ( reduced @xmath75 ) . .cleaned exposures and fov after the removal of point sources and oot events ( calculated using backscal keyword ) of three ` m31on ` observations . [ cols="^,^,^ " , ] this work provides more than @xmath76 times improvement of statistics of observations , as compared to lk09 ( factor @xmath77 times larger total exposure time and factor @xmath78 improvement in both the effective area and the field - of - view ) . even under the most conservative assumption about the dm column density such an improvement should have led to an about @xmath79 detection of the dm decay line . in our analysis no significant lines in the position , predicted by lk09 were found . however , in the _ xmm - newton _ observations that we processed there are several spectral features in the range @xmath80 ( see table [ tab:3sigma ] ) . this is not surprising , as the _ xmm - newton _ gold - coated mirrors have au absorption edge at @xmath81 @xcite and therefore their effective areas possess several prominent features in the energy range @xmath80 . according to _ xmm - newton_calibration report @xcite , there is about 5% uncertainty at the modeling of the effective area of both mos and pn cameras at these energies . these uncertainties in the effective area can lead to artificial spectral features due to the interaction of the satellite with cosmic rays . in particular , the solar protons with energies of few hundreds kev can be interpreted as x - ray photons ( so called _ soft proton flares _ ) . the interaction efficiency of solar protons with the instrument is known to be totally different from that of real photons @xcite . in particular , their flux is not affected by the au edge of the _ xmm - newton_mirrors . according to @xcite , the spectrum of soft proton flares for epic cameras is well described by the _ unfolded _ ( i.e. assuming that the instrument s response is energy - independent ) broken powerlaw model with the break energy around @xmath82 kev , ` bknpow / b ` in ` xspec v.11 ` . therefore , modeling the spectrum that is significantly contaminated with soft proton flares in a standard way ( i.e. by using _ folded _ ` powerlaw ` model ) will produce artificial residual excess at energies of sharp decreases of the instrument s effective area . in our data analysis we performed both the standard flare screening @xcite that uses the inspection of high - energy ( 1015 kev ) light curve and an additional soft proton flare cleaning . as found by @xcite , it is possible to screen out the remaining flares , e.g. by the visual inspection of the cleaned lightcurve at low and intermediate energies . to provide the additional cleaning after the rejection of proton flares at high energies , we followed the procedure of @xcite , leaving only the time intervals , where the total count rate differs from its mean value by less than @xmath49 . to test the possible instrumental origin of the features , discussed in this section , we first performed only a high - energy light curve cleaning @xcite and determined the maximally allowed flux in a narrow line in the energy interval of interest . after that we performed an additional soft proton cleaning . this procedure improved the quality of fit . the significance contours of the resulting `` line '' ( for ` m31out ` region ) are shown on the left panel in fig . [ fig : m31out ] . in deriving these limits we allowed both the parameters of the background , the position of the line and its total normalization to vary ( the latter from negative values ) . finally , we added an unfolded broken powerlaw component ` bknpow / b ` to model a contribution of remaining soft proton flares ( see e.g. @xcite ) . the break energy was fixed as 3.2 kev so no degeneracy with the parameters of the ` gaussian ` is expected . the ` bknpow / b ` powerlaw indices at low and high energies were fixed to be the same for different cameras observing the same spatial region . further improved the quality of fit and made the line detection totally insignificant ( see right fig . [ fig : m31out ] for details ) . our analysis clearly suggests the instrumental origin of the line - like features at the energy range 2.12.8 kev . as a final note , we should emphasize that the _ chandra _ acis instrument used in lk09 has the ir absorption edges near 2.11 and 2.55 kev ( see @xcite for additional discussion ) . they also report negative result of searching for the lk09 feature , in agreement with our findings . in this work we demonstrated that the dm decay line can be unambiguously distinguished from spectral features of other origin ( astrophysical or instrumental ) by studying its spacial distribution . many dm - containing objects would provide a comparable dm decay signal @xcite which makes such a study possible ( cf . * ? ? ? * ; * ? ? ? * ; * ? ? ? if an interesting candidate line is found in an object with an unusually high column density , the differences ( always within an order of magnitude ) in column densities with other objects can be compensated by longer exposure or bigger grasp of observing instruments . to illustrate the power of this strategy , we have applied this approach to the recent result of lk09 who put forward a hypothesis about a decaying dm origin of a @xmath83 spectral feature in chandra observations of the milky way satellite known as willman 1 . although the parameters of decaying dm , corresponding to such an interpretation , would lie in the region , already excluded at least @xmath3 level by several existing works @xcite we performed a new dedicated analysis of several archival observations of m31 , fornax and sculptor dsphs . the exclusions provided by m31 are extremely strong ( at the level of @xmath84 , depending on the observed region ) . the @xmath0 bound is obtained using the off - center ( @xmath85 kpc ) observations and a very conservative model of dm distribution with a heavy stellar disc and a very large ( @xmath86 ) dm core ( the model of @xcite , see section [ sec : dm - m31 ] ) . such a model leaves dm column density almost unchanged from about @xmath87 inward . combining all _ xmm - newton _ observations of m31 at the distances @xmath88@xmath89 off - center provides more than @xmath90 bound assuming the model by @xcite and the @xmath91 bound with the model m31b of @xcite ( with more physically motivated values of the mass of the stellar disk of m31 ) . fornax dsph provides an exclusion between @xmath92 an @xmath93 ( depending on how far from its nominal value @xmath94 kev the position of the line is allowed to shift ) instead of @xmath95 detection expected if the spectral feature of lk09 was of the dm origin . the bound of @xmath92 corresponds to the interval @xmath96 kev and the @xmath97 exclusion corresponds to @xmath98 kev interval . _ to summarize , _ by comparing the strength and position of the feature found by loewenstein & kusenko with observations of several dm - dominated objects ( m31 , fornax and sculptor dsphs ) we found that the hypothesis of dark matter origin of lk09 is excluded with the combined significance exceeding @xmath8 even under the most conservative assumptions . this is possible because of the large increase of the statistics in our observations as compared with the observation used in lk09 . to change this conclusion , a number of extreme ( and not otherwise motivated ) systematic errors would need to be present in determination of dm content of m31 , fornax and sculptor dsph . the dm origin of the spectral feature of lk09 would remain consistent with existing archival data only if the dm amount in m31 , fornax and sculptor are strongly overestimated and/or the mass of willman 1 is grossly underestimated even compared to the best - fit model of @xcite . it should be stressed that in our calculations of the expected signal in other objects , we have adopted for willman 1 the dm column density that results from the mass modeling of @xcite . our own modeling indicates that the mass of willman 1 is highly uncertain . even supposing that the measured velocity dispersion of willman 1 provides a clean estimate of the mass , the general models of @xcite allow the mass of willman 1 to vary over several orders of magnitude within the region of interest . furthermore , plausible scenarios such as tidal heating and/or contamination from unresolved binaries in the kinematic sample would cause the willman 1 mass to be over - estimated . we conclude that the column density derived from the modeling of @xcite is approximately an upper limit . in contrast , the large kinematic samples of fornax and sculptor provide tighter mass constraints under the assumptions of our modeling , and given the larger inherent velocity dispersion , the modeling assumptions are more secure for fornax and sculptor than they are for willman 1 . as a result , we expect that significance of our exclusion of dark matter origin of lk09 feature to be very conservative . small line - like features ( with the significance @xmath99 ) are present in the _ xmm - newton_spectrum at energies between 2 and 3 kev ( see fig . [ fig : m31out ] , left and table [ tab:3sigma ] for details ) . their most probable cause is the interaction of the satellite with soft protons ( see discussion in section [ sec : discussion ] ) which can lead to an appearance of artificial spectral features due to the uncertainties in the effective area at these energies ( c.f . indeed , the significance of the detection of these features drops below @xmath100 when the residual contamination with soft protons is modeled by an unfolded broken powerlaw ( see previous section for details ) . the search for decaying dm is a well - motivated task , having crucial importance not only for cosmology and astrophysics , but also for particle physics . x - ray astronomy has a significant potential in this respect @xcite . due to the small amount of baryons present in dwarf spheroidal galaxies , the uncertainties in dm determinations are smaller ( as compared to e.g. spiral galaxies ) and dm modeling provides tight bounds on the amount of dm . however , the above - mentioned uncertainties make willman 1 ( or other ultra - faint dwarfs ) to be bad observational targets from the point of view of searching for decaying dm . on the contrary , the `` classical '' dsphs provide excellent observational targets @xcite . a signal in fornax , sculptor ( or other `` classical '' dsphs such as draco or ursa minor not considered here because of the absence of archival observations for them ) would therefore have the potential to provide much more information about the nature of dm than would a signal from willman 1 . to improve significantly the best available bounds ( see e.g. * ) and to probe the most interesting region of parameter space , an extended ( 300 - 500 ks ) observations of the well - studied dsphs are needed . however as we are searching for the weak diffuse signal , a special care should be taken in choosing the instruments and observational times in such a way that minimize possible flare contamination ( see e.g. * ? ? ? drastic improvement in the decaying dm search is possible with a new generation of spectrometers , having large field of view and energy resolution close to several ev @xcite . we are grateful to l. chemin , e. corbelli , m. markevitch , s. molendi , j. nevalainen , m. shaposhnikov for many useful comments and discussions . di is also grateful to scientific and educational centre of the bogolyubov institute for theoretical physics in kiev , ukraine , and especially to v. shadura , for creating wonderful atmosphere for young ukrainian scientists . the work of ab and or was supported in part by the swiss national science foundation . the work of di is supported in part from the scopes project no . iz73z0_128040 of swiss national science foundation , the cosmomicrophysics programme and the program of fundamental research of the physics and astronomy division of the national academy of sciences of ukraine . the dark cosmology centre is funded by the danish national research foundation . a. , et al . , 2006 , in society of photo - optical instrumentation engineers ( spie ) conference series vol . 6270 of society of photo - optical instrumentation engineers ( spie ) conference series , ciao : chandra s data analysis system m. g. f. , et al . , 2004 , in g. hasinger & m. j. l. turner ed . , society of photo - optical instrumentation engineers ( spie ) conference series vol . 5488 of presented at the society of photo - optical instrumentation engineers ( spie ) conference , xmm - newton ( cross)-calibration .	a signal from decaying dark matter ( dm ) can be unambiguously distinguished from spectral features of astrophysical or instrumental origin by studying its spatial distribution . we demonstrate this approach by examining the recent claim of @xcite regarding the possible dm origin of the 2.5 kev line in chandra observations of the milky way satellite known as willman 1 . our conservative strategy is to adopt , among reasonable mass estimates derived here and in the literature , a relatively large dark mass for willman 1 and relatively small dark masses for the comparison objects . in light of the large uncertainty in the actual dark matter content of willman 1 , this strategy provides minimum exclusion limits on the dm origin of the reported signal . we analyze archival observations by _ xmm - newton _ of m31 and fornax dwarf spheroidal galaxy ( dsph ) and _ chandra _ observations of sculptor dsph . by performing a conservative analysis of x - ray spectra , we show the absence of a dm decay line with parameters consistent with those of @xcite . for m31 , the observations of the regions between 10 and 20 kpc from the center , where the uncertainties in the dm distribution are minimal , make a strong exclusion at the level above @xmath0 . the fornax dsph provides a @xmath1 exclusion instead of a predicted @xmath2 detection , and the sculptor dsph provides a @xmath3 exclusion instead of a predicted @xmath4 detection . the observations of the central region of m31 ( 1 - 3 kpc off - center ) are inconsistent with having a dm decay line at more than @xmath5 if one takes the most conservative among the best physically motivated models . the minimal estimate for the amount of dm in the central 40 kpc of m31 is provided by the model of @xcite , assuming the stellar disk s mass to light ratio @xmath6 and almost constant dm density within a core of 28 kpc . even in this case one gets an exclusion at @xmath7 from central region of m31 , whereas modeling _ all _ processed data from m31 and fornax produces more than @xmath8 exclusion . therefore , despite possible systematic uncertainties , we exclude the possibility that the spectral feature at @xmath9 kev found in @xcite is a dm decay line . we conclude , however , that the search for dm decay line , although demanding prolonged ( up to 1 msec ) observations of well - studied dsphs , m31 outskirts and other similar objects , is rather promising , as the nature of a possible signal can be checked . an ( expected ) non - observation of a dm decay signal in the planned observations of willman 1 should not discourage further dedicated observations .
* a weighted distribution of @xmath1-tuples of complex numbers . the intention of this paper is to provide a rigorous mathematical treatment of the theoretical physics model due to wiegmann , zabrodin , et al . , which is concerned with the behavior of a finite number of electrons ( charged fermions ) in a strong magnetic field in the context of quantum theory . in part , this was done very recently by elbau and felder . we will say more about this below . * let @xmath1 be a positive integer , and introduce the notation @xmath2}=(z_1,z_2,z_3,\ldots , z_n)\ ] ] for a point in @xmath1-dimensional complex euclidean space @xmath3 . as a real vector space , we have the identification @xmath4 . we let @xmath5 be the standard lebesgue measure on @xmath3 , so that in particular , @xmath6}\big)={{\mathrm d}}\lambda_{2}(z_1)\cdots { { \mathrm d}}\lambda_{2}(z_n).\ ] ] the _ van der monde determinant _ is the quantity @xmath7}\big)=\prod_{j , k:\,j < k}(z_j - z_k),\ ] ] where we tacitly assume that the parameters @xmath8 and @xmath9 range over the set @xmath10 . let @xmath11 be a @xmath0-smooth function , which grows so quickly that @xmath12 holds for all positive real constants @xmath13 ( with @xmath14 dependent on @xmath13 ) . let @xmath15 denote a positive real parameter . we let @xmath16 be the finite positive borel measure on @xmath17 given by @xmath18 we shall consider the probability measure @xmath19}\big)= \frac1{z_n}\,\big\|\triangle\big({{\mathbf z}}_{1,n]}\big)\big\|^2\ , { { \mathrm d}}\mu_{\beta,{\phi}}(z_1)\cdots { { \mathrm d}}\mu_{\beta,{\phi}}(z_n ) , \label{eq - distr - pi}\ ] ] where @xmath20 is the normalizing constant : @xmath21}\big)\big\|^2\ , { { \mathrm d}}\mu_{\beta,{\phi}}(z_1)\cdots { { \mathrm d}}\mu_{\beta,{\phi}}(z_n ) . \label{eq - distr - zn}\ ] ] we intend to study the typical behavior of ( [ eq - distr - pi ] ) for large values of @xmath1 . to make this more precise , let @xmath22 be a positive integer with @xmath23 , and consider the marginal distribution of @xmath24}=(z_1,z_2,z_3,\ldots , z_n),\ ] ] as @xmath25 while @xmath22 is kept fixed . @xmath22-point marginal distribution _ is given by the probability measure @xmath26})= \frac1{z_n}\,\bigg\{\int_{{{\mathbb c}}^{n - n}}\big\|\triangle({{\mathbf z}}_{[1,n]})\big\|^2\ , { { \mathrm d}}\mu_{\beta,{\phi}}(z_{n+1})\cdots { { \mathrm d}}\mu_{\beta,{\phi}}(z_{n})\bigg\}\\ \times{{\mathrm d}}\mu_{\beta,{\phi}}(z_1)\cdots{{\mathrm d}}\mu_{\beta,{\phi}}(z_n ) . \label{eq - distr - pi2}\end{gathered}\ ] ] the measure @xmath27 obtained by multiplying @xmath28 by @xmath29 is called the _ @xmath22-point correlation measure_. the distribution ( [ eq - distr - pi ] ) appears from at least two different contexts : one is the theory of eigenvalues of random normal matrices , and the other a physical configuration of @xmath1 electrons localized to a two - dimensional plane , while being exposed to a strong magnetic field ( the strength is regulated by the parameter @xmath15 ) perpendicular to the given plane . in the second instance , @xmath30 is a scalar magnetic potential , @xmath31 are the locations of the @xmath1 electrons , and one looks for time - independent solutions to the schrdinger equation . the resulting cloud of electrons is an instance of the aharonov - bohm effect , where the magnetic potential rather than the magnetic field manifests itself . for details , we ask the reader to consult the work of wiegmann , zabrodin , et al . @xcite , @xcite , @xcite , @xcite , and @xcite . in the first instance ( random normal matrices ) , we should mention first that normal @xmath32 matrices have the decomposition @xmath33 , where @xmath34 is diagonal ( consisting of the eigenvalues ) , and @xmath35 is unitary , that is , @xmath36 , where @xmath37 is the identity matrix . to focus on the eigenvalues , then , one should remove as much as possible of the unitary part ; in other words , we should mod out with respect to @xmath38 , where @xmath39 is the group of unitary @xmath32 matrices , and @xmath40 is the subgroup of unitary diagonal matrices . if this is done carefully , and a weight is introduced to get a probability measure on the normal matrices , we get a joint distribution of the eigenvalues of the type ( [ eq - distr - pi ] ) . for details of this , see @xcite and @xcite ; see also mehta s book @xcite . the standard models for the eigenvalue distribution of self - adjoint matrices are by now quite well - known ( see @xcite , @xcite ) . in the limit , we get such models from our probability measure ( [ eq - distr - pi ] ) by letting @xmath41 tend to @xmath42 on @xmath43 . for instance , the semicircle law of wigner can be obtained by taking such a limit with the appropriate interpretation . * quantum hele - shaw flow vs hele - shaw flow . we show that the @xmath22-point correlation measure @xmath44 is given in terms of a determinantal process involving reproducing kernels of polynomial subspaces of a bargmann - fock type space , and that it defines a monotonic growth process in @xmath1 ( for fixed @xmath15 and @xmath45 ) : @xmath46 for this reason , and for reasons which will be made clearer below , we call this _ quantum hele - shaw flow_. the analogy with classical hele - shaw flow was made earlier by wiegmann and zabrodin . the ratio @xmath47 seems like a natural growth parameter for the process . for mathematical reasons , however , it is more natural to use @xmath48 instead . the weighted energy of a compactly supported borel probability measure @xmath49 is ( for a given positive parameter @xmath50 ) @xmath51\left(\frac{1}{\tau}\right ) = \frac1{2\tau}\iint_{{{\mathbb c}}^2}\bigg\{2\tau\log\frac{1}{\|z - w\|}+{\phi}(z)+ { \phi}(w)\bigg\ } \,{{\mathrm d}}{\sigma}(z){{\mathrm d}}{\sigma}(w ) ; \ ] ] the unique compactly supported borel probability measure which minimizes this energy is called the _ extremal measure _ , or _ equilibrium measure _ , and denoted by @xmath52 , or simply by @xmath53 , if no misunderstanding is possible . the existence and uniqueness of the extremal measure are treated in chapter 1 of saff s and totik s book @xcite . * our first main result reads as follows . if @xmath50 , @xmath54 , is kept fixed , and @xmath55 are connected via ( [ def - tau ] ) , then , as @xmath25 , the marginal distribution has the weak - star limit @xmath56})\to { { \mathrm d}}\widehat{\sigma}(z_1)\cdots{{\mathrm d}}\widehat{\sigma}(z_n).\ ] ] [ thm - i ] this theorem is essentially as in johansson s paper @xcite , but with important additional technical details . elbau and felder @xcite were the first to realize that johansson s argument extends to the case of normal random matrices . the present version is more general elbau and felder required that the potential be infinite in a neighborhood of infinity . we discuss the extremal measure in section [ sec2 ] , and connect it to an obstacle problem in section [ sec3 ] . this leads to our second main theorem . for the formulation , we need the following standard notation : @xmath57 is the standard laplacian in the plane , while given a subset @xmath58 of the complex plane @xmath17 , @xmath59 is the characteristic function of the set @xmath58 . for each @xmath50 , @xmath54 , there exists an unbounded open set @xmath60 in @xmath17 whose complement is compact , such that the extremal measure has the form @xmath61 moreover , as @xmath50 increases , the open sets @xmath60 form a continuous decreasing chain , and the measures @xmath62 grow with @xmath50 . [ thm - ii ] this theorem justifies the claim of wiegmann , zabrodin , et al . , that in the limit one gets a monotone family of domains ( and not only measures ) in the complex plane . the proof is an application of the general theory of free boundary problems due to caffarelli et al . we should add that , if the weight function @xmath45 is real analytic , then the boundary @xmath63 consists of real analytic curves , except for finitely many cusps ( pointing outward ) or so - called contact points . the evolution of @xmath60 is quite similar to the hele - shaw flow on hyperbolic or weakly hyperbolic surfaces described in @xcite and @xcite . in fact , the present situation corresponds to hele - shaw flow on surfaces where the metric may be negative in portions of space , since @xmath64 is a kind of metric potential , with associated metric @xmath65 generally speaking , hele - shaw flow models the expansion of viscous fluid in a surrounding medium ( which we may think of as vacuum ) , as new fluid is injected at a constant rate at one or several source points . in the present setting , we have a single source point at infinity . a final remark we wish to make here is that essentially all results of the paper remain valid with little change if the @xmath0-smoothness of @xmath45 is weakened to local @xmath66-smoothness , for some @xmath67 , @xmath68 . * weighted bargmann - fock spaces and reproducing kernels . let @xmath69 be a positive finite borel measure on the complex plane @xmath17 , with @xmath70 the hilbert space @xmath71 consists of the square - integrable complex - valued functions with respect to @xmath69 , where two functions are identified if the coincide @xmath69-almost everywhere . we shall assume that @xmath69 is supported on an infinite set , so that @xmath71 becomes infinite - dimensional . the above integrability condition ensures that all ( analytic ) polynomials belong to @xmath71 . let @xmath72 denote the closure of the linear space of polynomials in @xmath71 . also , for @xmath73 , let @xmath74 be the @xmath1-dimensional linear space of polynomials of degree @xmath75 , supplied with the hilbert space structure of @xmath71 . the point evaluations in @xmath17 are linear functionals on @xmath74 , and thus given by the inner product with an element of the space : @xmath76 where , for fixed @xmath77 , the function @xmath78 is an element of @xmath74 , and so is @xmath67 . the function @xmath79 is called the _ reproducing kernel _ for the space @xmath74 . the monomials @xmath80 form a vector space basis in @xmath74 , but they need not be orthogonal to one another . to rectify this , we may apply the gram - schmidt process : let @xmath81 , and for @xmath82 , put @xmath83 where the complex scalars @xmath84 are chosen so that @xmath85 becomes perpendicular to all the vectors @xmath86 . the functions @xmath87 obtained in this fashion form an orthogonal sequence . to get an orthonormal sequence , we normalize : @xmath88 the reproducing kernel may then be conveniently expressed in terms of these orthogonal functions : @xmath89 * * matrix - valued reproducing kernels and correlation measures . for two vectors @xmath24}=(z_1,\ldots , z_n)\quad\text { and } \quad { { \mathbf w}}_{[1,n]}=(w_1,\ldots , w_n),\ ] ] we form the @xmath90 matrix @xmath91},{{\mathbf w}}_{[1,n]})= \big[k_{\mu , n}(z_j , w_k)\big]_{j , k=1}^n , \label{matrix - k}\ ] ] which we may think of as a matrix - valued reproducing kernel . * we return to the joint distribution function ( [ eq - distr - pi ] ) , in the slightly altered form @xmath92})= \frac1{z_n}\,\big\|\triangle({{\mathbf z}}_{[1,n]})\big\|^2\,{{\mathrm d}}\mu(z_1)\cdots { { \mathrm d}}\mu(z_n ) , \label{distr - pi-1}\ ] ] where @xmath93}=(z_1,\ldots , z_n)$ ] , and @xmath20 is the normalizing constant : @xmath94})\big\|^2\,{{\mathrm d}}\mu(z_1)\cdots { { \mathrm d}}\mu(z_n ) . \label{normconst}\ ] ] we fix an integer @xmath22 , @xmath23 , and split @xmath93}=({{\mathbf z}}_{[1,n]},{{\mathbf z}}_{]n , n]})$ ] , with @xmath95}=(z_1,\ldots , z_n)$ ] and @xmath96n , n]}=(z_{n+1},\ldots , z_n)$ ] . the _ @xmath22-point marginal distribution measure _ is @xmath97})=\frac1{z_n}\ , \bigg\{\int_{{{\mathbb c}}^{n - n}}\big\|\triangle({{\mathbf z}}_{[1,n]})\big\|^2\ , { { \mathrm d}}\mu(z_{n+1})\cdots{{\mathrm d}}\mu(z_{n})\bigg\}\\ \times{{\mathrm d}}\mu(z_1)\cdots{{\mathrm d}}\mu(z_n ) , \label{distr - pi-2.1}\end{gathered}\ ] ] while the _ @xmath22-point correlation measure _ is @xmath98}\big)=\frac{n!}{(n - n)!}\ , { { \mathrm d}}\pi^{(n)}_{\mu , n}({{\mathbf z}}_{[1,n]}),\qquad { { \mathbf z}}_{[1,n]}\in{{\mathbb c}}^n , \label{distr - pi-2.2}\ ] ] where the integration takes place with respect to the variables @xmath99 only . for @xmath100 , the measure @xmath101 describes the joint density of the eigenvalues . for measures @xmath69 of the form ( [ eq - mu ] ) , with @xmath102 a positive constant in a large disk centered at the origin , @xmath101 is rather similar to the majorization function of aleman , richter , and sundberg @xcite ; one major difference , though , is that here , we consider backward shift invariant subspaces ( @xmath74 ) , while in @xcite , the object of study are the forward shift invariant subspaces . we have that @xmath103},{{\mathbf z}}_{[1,n]})\big]= \frac{n!}{z_n}\,\big\|\triangle({{\mathbf z}}_{[1,n]})\big\|^2.\ ] ] [ prop - n = n ] we first apply simple row or column operations on the representation of the van der monde determinant as the determinant of the @xmath32 matrix with entries @xmath104 , for @xmath105 . the result is that the van der monde determinant may be calculated based on the entries @xmath106 instead , whence the result follows from the formula defining the determinant in terms of permutations , as in @xcite , pp . 8995 . the following `` linear algebra '' description of @xmath107 in terms of the determinant of a matrix reproducing kernel is essentially known ( see @xcite for the gaussian situation ) . we have that @xmath108})=\det \big[{\mathbf k}^n_{\mu , n}({{\mathbf z}}_{[1,n]},{{\mathbf z}}_{[1,n]})\big]\,\ , { { \mathrm d}}\mu(z_1)\cdots{{\mathrm d}}\mu(z_n).\ ] ] [ thm - lambda ] we are to show that @xmath109})\big\|^2\ , { { \mathrm d}}\mu(z_{n+1})\cdots{{\mathrm d}}\mu(z_{n } ) = z_n\,\det\big[{\mathbf k}^n_{\mu , n}({{\mathbf z}}_{[1,n]},{{\mathbf z}}_{[1,n]})\big].\ ] ] we follow the approach of @xcite , pp . 8995 , and use that by proposition [ prop - n = n ] , the identity is valid for @xmath110 , which we take as our starting point . then successive applications of theorem 5.2.1 in @xcite ( suitable modified by including complex conjugates where needed ) yields the result for general @xmath111 . the proof is complete . the @xmath22-point correlation measure @xmath112 increases monotonically with increasing @xmath1 . its total mass equals @xmath29 . that the total mass of @xmath112 is @xmath29 follows from the definition ( [ normconst ] ) of the normalizing constant @xmath20 in view of ( [ kernelexp ] ) , we have @xmath113 so that @xmath114},{{\mathbf w}}_{[1,n]})= { \mathbf k}^n_{\mu , n}({{\mathbf z}}_{[1,n]},{{\mathbf w}}_{[1,n]})+ { \boldsymbol\phi}_{\mu , n+1}({{\mathbf z}}_{[1,n]},{{\mathbf w}}_{[1,n]}),\ ] ] where @xmath115},{{\mathbf w}}_{[1,n]})= \big(\phi_{n+1}(z_j)\,\bar\phi_{n+1}(w_k)\big)_{j , k=1}^{n}.\ ] ] the matrix @xmath116},{{\mathbf z}}_{[1,n]})$ ] is positive definite , and so is @xmath117},{{\mathbf z}}_{[1,n]})$ ] . as the determinant operation is monotonic with respect to increasing positive definiteness ( see proposition [ prop - a+b ] below ) , the assertion follows from theorem [ thm - lambda ] . the proof is complete . * a linear algebra result . the following proposition is known ; for instance , it should be possible to derive it from the horn inequalities , which describe the possible eigenvalues of the sum of two hermitian matrices in terms of the eigenvalues of the summands ( see , e. g. , the survey paper @xcite ) . however , we find it convenient to supply instead a short direct proof . * let @xmath13 and @xmath118 be two positive semi - definite ( hermitian ) @xmath32 matrices . then @xmath119 [ prop - a+b ] clearly , by symmetry , it suffices to show that @xmath120 if @xmath121 , the assertion follows trivially , as the determinant of a positive semi - definite matrix is always @xmath122 . in the non - singular case @xmath123 , we note that @xmath124 is a well - defined positive semi - definite matrix , and that @xmath125 is positive semi - definite as well . we write @xmath126 where @xmath37 denotes the @xmath32 identity matrix . the assertion now follows from the fact that the eigenvalues of @xmath125 are all positive , so that the eigenvalues of @xmath127 are all greater than @xmath128 ( after all , the determinant equals the product of the eigenvalues ) . * weighted fekete points . let us analyze the probability measure ( [ eq - distr - pi ] ) , written in the form @xmath129})= \frac1{z_n}\exp\bigg\{-\beta\sum_{j=1}^n { \phi}(z_j)\bigg\}\ , \big\|\triangle({{\mathbf z}}_{[1,n]})\big\|^2\,{{\mathrm d}}\lambda_{2n}({{\mathbf z}}_{[1,n ] } ) , \label{distr - pi-3}\ ] ] where @xmath93}=(z_1,\ldots , z_n)$ ] , and @xmath20 is the normalization constant . the most likely configuration @xmath93}$ ] is the one that maximizes @xmath130})\big\|^2\ ] ] among all vectors in @xmath3 . considering that @xmath131=(n-1)\sum_{j=1}^n { \phi}(z_j),\ ] ] we may rewrite this expression : @xmath132})\big\|^2\\ = \prod_{j , k:\,j < k } \bigg\{\exp\bigg(-\frac{\beta}{n-1}\,\big[{\phi}(z_j)+{\phi}(z_k)\big ] \bigg)\,\|z_j - z_k\|^2\bigg\}. \label{eq - dens}\end{gathered}\ ] ] for positive real @xmath133 , let @xmath134 denote the function @xmath135\big ) ; \label{e - fn}\ ] ] we are interested in maximizing @xmath136 with @xmath137 let @xmath138 be this maximum raised to the power @xmath139 : @xmath140}\in{{\mathbb c}}^n}\bigg\ { \prod_{j , k:\,j < k}e_{\phi}(z_j , z_k;\theta)\bigg\}^{2/(n(n-1))}.\ ] ] the points @xmath93}={{\mathbf z}}_{[1,n]}^$ ] which achieve this maximum are called _ weighted fekete points _ ( see @xcite ) . in order that the maximum be assumed , we should add an assumption on the weight @xmath45 : @xmath141 an easy argument shows that ( [ reg - cond ] ) follows from ( [ est - below ] ) with an appropriate choice of the parameter @xmath13 . it follows from theorem 1.1 @xcite that @xmath138 _ is decreasing in _ @xmath1 , provided that @xmath45 is fixed and @xmath133 is held constant . moreover , under the same conditions , @xmath142 as @xmath25 , where @xmath143 and @xmath144 is results from a certain minimization problem , outlined below . the quantity @xmath145 can be thought of as a weighted capacity ( of the whole plane ) . * the extremal measure . let @xmath146 denote the convex `` body '' of all compactly supported borel probability measure @xmath49 on @xmath17 . for @xmath147 , we introduce the _ energy functional _ @xmath148(\theta)=2\theta\int_{{\mathbb c}}{\phi}(z)\,{{\mathrm d}}{\sigma}(z)+ \iint_{{{\mathbb c}}^2}\log\frac{1}{\|z - w\| } { { \mathrm d}}{\sigma}(z){{\mathrm d}}{\sigma}(w)\\ = \iint_{{{\mathbb c}}^2}\bigg\{\log\frac{1}{\|z - w\|}+\theta\,{\phi}(z)+ \theta\,{\phi}(w)\bigg\ } \,{{\mathrm d}}{\sigma}(z){{\mathrm d}}{\sigma}(w)\\ = \iint_{{{\mathbb c}}^2}\log \frac{1}{e_{\phi}(z , w;\theta)}\ , { { \mathrm d}}{\sigma}(z){{\mathrm d}}{\sigma}(w ) , \label{energy}\end{gathered}\ ] ] and consider the problem of minimizing @xmath149(\theta)$ ] over all @xmath49 . define @xmath150(\theta ) ; \label{cap - def}\ ] ] this quantity is related to @xmath145 via ( [ limit - eq ] ) , in view of theorem 1.3 @xcite . any @xmath147 which achieves the maximum in ( [ cap - def ] ) is called an _ extremal ( or equilibrium ) measure_. in view of ( [ reg - cond ] ) , which follows from ( [ est - below ] ) , it follows from theorem 1.3 of @xcite that there _ exists a unique extremal measure , which we will denote by _ @xmath53 , or , when it necessary to indicate the dependence on the parameter @xmath133 as well as on the weight @xmath45 , by @xmath151 . * a change of parameters . for reasons of convenience , we will now use @xmath152 in place of @xmath133 as a positive real parameter . as before , @xmath11 is @xmath0-smooth , and subject to the growth requirement ( [ est - below ] ) for every choice of the real parameter @xmath13 . in particular , ( [ reg - cond ] ) holds with @xmath153 . as we mentioned in the previous section , by theorem 1.3 @xcite , there exists a unique extremal measure @xmath154 which achieves the infimum in ( [ cap - def ] ) . for @xmath53 , the associated logarithmic potential @xmath155(z)= \int_{{\mathbb c}}\log\frac{1}{\|z - w\|}\,{{\mathrm d}}\widehat{\sigma}(w ) , \qquad z\in{{\mathbb c}},\ ] ] is locally bounded in @xmath17 ( theorem 4.3 @xcite ) . so , for instance , the extremal measure @xmath53 has no point masses . to get an understanding of this extremal measure , we make a detour to obstacle problems and hele - shaw flow . a general reference for hele - shaw flow is the recent book of gustafsson and vasiliev @xcite . * * an obstacle problem . consider the following function : @xmath156 for @xmath54 , let @xmath157 denote the collection of ( extended real - valued ) functions that are superharmonic in the whole plane @xmath17 , and which are of the form @xmath158 where @xmath159 is superharmonic in a punctured neighborhood of infinity , and bounded from below there . if we have two functions @xmath160 , then the minimum of the two @xmath161 is in @xmath157 as well . in addition , the decreasing limit of functions in @xmath157 remains in @xmath157 unless it degenerates to @xmath162 identically . _ we define @xmath163 to be the least majorant ( or lower envelope ) of @xmath164 in the class @xmath157 . _ in other words , @xmath164 is an obstacle , and @xmath163 solves the _ obstacle problem_. * * the extremal measure and the obstacle problem . the next proposition explains the relationship between the obstacle problem and the solution @xmath165 to the optimization problem ( [ cap - def ] ) with @xmath153 . for convenience of notation , we shall write @xmath52 in place of @xmath166 when we need to indicate the dependence on the parameter @xmath50 . * the logarithmic potential for the extremal measure @xmath167 is @xmath168(z)=\frac1{2\tau}\,\widehat v_{\tau}(z)+c_\tau,\ ] ] where @xmath169 is the constant @xmath170 [ prop - sigma - env ] this follows from theorem 4.1 @xcite . we recall that the function @xmath171 is assumed to be of class @xmath0 throughout @xmath17 . we shall need the class @xmath172 , which consists of continuously differentiable functions whose gradient is locally lipschitz . in other words , all second order partial derivatives of the function are locally bounded . the envelope @xmath173 is of class @xmath172 throughout @xmath17 . [ prop - c11 ] in view of proposition [ prop - sigma - env ] , the envelope function @xmath173 is harmonic off @xmath174 . since the extremal measure @xmath53 has compact support , it follows that we can find an open circular disk @xmath175 of radius @xmath176 , @xmath177 , around the origin , such that @xmath174 is compactly contained in @xmath175 . along the boundary @xmath178 of the disk , @xmath173 is harmonic , and in particular smooth . we consider the following obstacle problem . suppose @xmath179 is superharmonic on @xmath175 and @xmath180-smooth on @xmath181 , and @xmath182 the lower envelope of all such @xmath179 is denoted by @xmath183 . a simple argument shows that @xmath183 coincides with @xmath173 . indeed , we easily see that @xmath184 on @xmath175 , with equality on @xmath178 ; moreover , if we extend @xmath183 to all of @xmath17 by declaring that it should equal @xmath173 off @xmath175 , we get a superharmonic function , which belongs to @xmath157 . but by definition @xmath173 is the smallest such majorant , and we must have equality : @xmath185 . in view of the known smoothness properties of solutions of obstacle problems of the type ( [ eq - obstprob ] ) , it follows from the @xmath0-smoothness of @xmath164 that @xmath173 is of class @xmath172 on @xmath175 . in the rest of the plane , it is harmonic , and therefore automatically of class @xmath172 . the basic references on obstacle problems are chapter 1 in @xcite , or the paper @xcite by caffarelli and kinderlehrer . the proof is complete . we introduce the notation @xmath186 this is an open set which contains a punctured neighborhood of the point at infinity . we have @xmath187 let @xmath188 be a point with @xmath189 , and suppose that @xmath190 . then @xmath191 so that if we put @xmath192 , we get @xmath193 , while @xmath194 everywhere , and @xmath35 is strictly superharmonic in a neighborhood of @xmath195 . by the strong maximum principle , then , this is not possible . the result is immediate . we also need the `` harmonicity '' set of @xmath45 : @xmath196 which is a closed set in @xmath17 . if @xmath197 is a borel measurable subset of the plane @xmath17 , let us agree to say that @xmath188 is a _ pseudo - interior point for _ @xmath197 if there exists a small open disk @xmath34 around @xmath195 such that @xmath198 holds , where @xmath199 is the operation of taking the lebesgue are measure of the given set . for instance , all interior points of the set @xmath197 are pseudo - interior points of @xmath197 . we are now in a position to reformulate in precise terms our second theorem from the introduction ( theorem [ thm - ii ] ) . in the sense of measures , we have @xmath200 in particular , the support of the measure @xmath52 consists of all points of @xmath17 which are not pseudo - interior for the set @xmath201 . [ prop - basic ] we get from proposition [ prop - sigma - env ] that @xmath202(z)\,{{\mathrm d}}\lambda_2(z ) = \frac1{2\tau}\,\delta\widehat v_{\tau}(z)\,{{\mathrm d}}\lambda_2(z ) . \label{eq - aa1}\ ] ] according to @xcite , we have @xmath203 in the almost - everywhere sense . on the non - coincidence set , however , @xmath173 is harmonic : @xmath204 this follows by a standard perron process argument ( see , for instance , proposition 2.2 in @xcite ) . by proposition [ prop - c11 ] , the function @xmath205 is in @xmath206 , so that we see from ( [ eq - aa1 ] ) , ( [ eq - aa2 ] ) , and ( [ eq - aa3 ] ) that @xmath207 as asserted . the statement regarding the support of @xmath52 is an easy consequence of this identity . the proof is complete . the obstacle problem considered here share many features in common with the hele - shaw flow domains considered by hedenmalm and shimorin in @xcite however , the time parameter @xmath50 flows in the opposite direction as compared with the time parameter @xmath208 used in @xcite . this means that any cusps of the compact set @xmath209 should point outward . also , if @xmath45 is real - analytic , the smoothness analysis of based on sakai s work @xcite done in @xcite carries over to the situation treated here . the domains @xmath210 grow as @xmath50 _ decreases _ , and the `` harmonic moments '' are preserved , with the exception of the first ; this is the content of the following proposition . let @xmath211 . suppose @xmath212 is harmonic and bounded in @xmath60 , with an extension to @xmath17 that is locally of sobolev class @xmath213 . we then have the equality @xmath214 [ prop - iii ] this follows from a suitable application of green s formula , analogous to what is done in @xcite . if the domains @xmath60 are smooth and vary smoothly with @xmath50 , then , by arguing as in @xcite , pp . 188 - 189 , we see that the assertion of proposition [ prop - iii ] has the interpretation that the boundary @xmath63 propagates with velocity proportional to a weight times the gradient of the green function with singularity at infinity . the growth of the complementary sets @xmath209 as @xmath50 increases is quite an interesting process . for @xmath50 close to @xmath215 , we should expect each such complementary set to be localized close to the points where the minimum of the potential function @xmath45 is attained . if there is only one minimum point @xmath195 , and the function @xmath45 is convex in a neighborhood of @xmath195 , then we can show that a small neighborhood of @xmath195 is contained in @xmath209 , for each fixed @xmath50 , by comparing with a concave majorant . an interesting question seems to be whether @xmath216 it is easy to see that a necessary condition for this to happen is that @xmath45 be subharmonic everywhere . we do not know to what extent this condition is sufficient . * the continuous limit of eigenvalues of normal matrices . for positive real @xmath15 , let @xmath16 denote the measure @xmath217 where we recall that @xmath218 is area measure in the plane . throughout this section , we shall use as before the following scaling choice of @xmath15 as we increase @xmath1 : @xmath219 where @xmath50 is a fixed positive real parameter . also , we recall the previously used notation @xmath24}=(z_1,\ldots , z_n),\qquad { { \mathbf z}}_{[1,n]}=(z_1,\ldots , z_n).\ ] ] given this setup , we consider the @xmath22-point correlation measure @xmath107 with @xmath220 and @xmath15 given by ( [ eq - beta ] ) . we want to understand the asymptotic behavior of this measure as @xmath25 . this means that we should understand the impact of working with the @xmath221 norm in place of the @xmath222 norm in the setting of ( [ distr - pi-3 ] ) . * we recall that @xmath45 is a @xmath0-smooth real - valued function , which tends to infinity at infinity at a pace prescribed by ( [ est - below ] ) . it follows from this that there exists a compact subset @xmath223 of @xmath17 such that @xmath224 + \log\frac1{\|z - w\| } , \qquad ( z , w)\in{{\mathbb c}}^2\setminus(k\times k ) , \label{def - k}\end{gathered}\ ] ] where we recall the definition ( [ cap - def ] ) of @xmath225 . in analogy with the energy function ( [ energy ] ) , we introduce the function @xmath226}\big](\theta)$ ] , @xmath227}](\theta)= \frac2{n(n-1)}\,\log\prod_{j , k:\,j < k}\frac1{e_{\phi}(z_j , z_k;\theta)}\\ = \frac1{n(n-1)}\sum_{j , k:\,j\neq k}\log\frac1{e_{\phi}(z_j , z_k;\theta ) } , \label{energy - sharp}\end{gathered}\ ] ] where @xmath134 is given by ( [ e - fn ] ) , and @xmath153 . we recall that @xmath228}](\theta),\qquad { { \mathbf z}}_{[1,n]}\in{{\mathbb c}}^n , \label{eq - estenergy}\ ] ] with equality only at the weighted fekete points . for positive real @xmath229 , let @xmath230}\in{{\mathbb c}}^n:\,\ , { { \mathcal e}}^\sharp_{\phi}[{{\mathbf z}}_{[1,n]}](\theta)\le \log\frac1{m_{{\phi},n}(\theta)}+\epsilon \bigg\}.\ ] ] we need to know that the proportion of points in @xmath231 which stays in the compact set @xmath223 converges to @xmath128 as @xmath25 and @xmath232 . suppose @xmath93}\in{{\mathfrak a}}_n(\epsilon,\theta)$ ] , and that @xmath233 . let @xmath234 denote the number of indices @xmath8 for which @xmath235 . then , for sufficiently large @xmath1 , we have @xmath236 [ prop - eps ] let @xmath237 be the subset of all indices @xmath8 for which @xmath235 , and let @xmath238 be the complement in @xmath10 . we split the sum defining @xmath226}\big](\theta)$ ] accordingly : @xmath239}\big](\theta ) = \frac1{n(n-1 ) } \sum_{j , k\in x:\,j\neq k}\log\frac1{e_{\phi}(z_j , z_k;\theta)}\\ + \frac2{n(n-1)}\sum_{j\in x,\,k\in y:\,j\neq k } \log\frac1{e_{\phi}(z_j , z_k;\theta ) } \\ + \frac1{n(n-1 ) } \sum_{j , k\in y:\,j\neq k}\log\frac1{e_{\phi}(z_j , z_k;\theta)}.\end{gathered}\ ] ] the sum of the last two terms is estimated from below by @xmath240 while the first term of the right hand side is estimated in the following manner : @xmath241}](\theta)\\ = \frac{1}{n_k(n_k-1 ) } \sum_{j , k\in x:\,j\neq k}\log\frac1{e_{\phi}(z_j , z_k;\theta ) } , \qquad { { \mathbf z}}_{[1,n]}\in{{\mathbb c}}^n.\end{gathered}\ ] ] we now see that for @xmath93}\in{{\mathfrak a}}_n(\epsilon,\theta)$ ] , @xmath242}](\theta)\le \log\frac1{m_{{\phi},n}(\theta)}+\epsilon.\end{gathered}\ ] ] it follows that @xmath243 for large values of @xmath1 , the left hand side is positive , and hence the right hand side is , too . we quickly rule out the possibility that the first on the right hand side is negative ( which corresponds to @xmath244 ) , in which case we are left with both factors on the right hand side being positive . we get from ( [ eq - est ] ) that @xmath245 if we recall that @xmath246 . next , we note that @xmath247 as @xmath1 tends to infinity ; therefore , for sufficiently large @xmath1 , @xmath248 the assertion of the proposition now follows from ( [ eq - est2 ] ) . let @xmath147 be absolutely continuous with respect to two - dimensional lebesgue measure , @xmath249 where @xmath250 is area - summable and @xmath251 is area - summable as well . the normalization constant @xmath20 in the definition of @xmath252})$ ] then has the following bound : @xmath253(\theta ) -n\int_{{\mathbb c}}\log s(z)\,{{\mathrm d}}{\sigma}(z)\bigg\},\ ] ] with the understanding that @xmath254 vanishes off the support of @xmath49 . [ lem-1.1 ] we recall first the definition of @xmath20 : @xmath255})\big\|^2\,{{\mathrm d}}\lambda_{2n}({{\mathbf z}}_{[1,n ] } ) . \label{eq - z}\ ] ] let @xmath256 denote the set where @xmath257 ; as @xmath49 has compact support , we may assume that the set @xmath256 is bounded . we then have @xmath258})\big\|^2\,\frac1{s(z_1)\cdots s(z_n)}\ , { { \mathrm d}}{\sigma}(z_1)\cdots{{\mathrm d}}{\sigma}(z_n)\\ \ge\exp \bigg\{\int_{{\sigma}^n } \bigg(-\beta\sum_{j=1}^n { \phi}(z_j)+ 2\log\big\|\triangle({{\mathbf z}}_{[1,n]})\big\|\,-\sum_{j=1}^n \log s(z_j)\bigg)\\ \times{{\mathrm d}}{\sigma}(z_1)\cdots{{\mathrm d}}{\sigma}(z_n ) \bigg\},\end{gathered}\ ] ] where the second estimate is due to jensen s inequality . in view of ( [ eq - beta ] ) and the definition of the energy functional , we have @xmath253(\theta ) -n\int_{\sigma}\log s(z)\,{{\mathrm d}}{\sigma}(z)\bigg\},\ ] ] as claimed . for @xmath93}\in{{\mathbb c}}^n\setminus{{\mathfrak a}}_n(\epsilon,\theta)$ ] , we have @xmath239}\big](\theta)= \frac1{n(n-1)}\sum_{j , k:\,j\neq k } \log\frac1{e_{\phi}(z_j , z_k;\theta ) } > \log\frac1{m_{{\phi},n}(\theta)}+\epsilon\\ \ge \kappa_{\phi}(\theta)+ \epsilon . \ ] ] we recall the estimate ( [ est - below ] ) from below of the weight @xmath45 , which for an appropriate choice of the parameter @xmath13 gives @xmath261}\big](\theta)\ge - c+\frac1{n } \sum_{j=1}^n\log\big(1+\|z_j\|^2\big ) , \qquad { { \mathbf z}}_{[1,n]}\in{{\mathbb c}}^n , \ ] ] for a positive constant @xmath262 that only depends on @xmath133 and @xmath45 . we form a convex combination of these two estimates ( @xmath263 ) : @xmath239}\big](\theta)\ge ( 1-\gamma)\big[\kappa_{{\phi}}(\theta)+\epsilon\big]-c\gamma\\ + \frac{\gamma}{n}\sum_{j=1}^n\log\big(1+\|z_j\|^2\big),\qquad { { \mathbf z}}_{[1,n]}\in{{\mathbb c}}^n\setminus{{\mathfrak a}}_n(\epsilon,\theta ) . \label{eq - a2}\end{gathered}\ ] ] we recall the formula defining @xmath264 , while keeping in mind the identity ( [ eq - dens ] ) , @xmath265}\big)=\frac1{z_n}\bigg [ \prod_{j , k:\,j\neq k } e_{\phi}(z_j , z_k;\theta)\bigg]\,{{\mathrm d}}\lambda_{2n } \big({{\mathbf z}}_{[1,n]}\big)\\ = \frac1{z_n}\,\exp\big\{-n(n-1)\,{{\mathcal e}}^\sharp_{\phi}\big[{{\mathbf z}}_{[1,n]}\big](\theta)\big\}\ , { { \mathrm d}}\lambda_{2n}\big({{\mathbf z}}_{[1,n]}\big ) . \label{eq - pi3}\end{gathered}\ ] ] so , on the set @xmath266 , we get , in view of ( [ eq - a2 ] ) , @xmath265}\big)\le\frac1{z_n}\ , \exp\bigg\{-n(n-1)(1-\gamma)\big[\kappa_{{\phi}}(\theta)+\epsilon\big]\\ + cn(n-1)\gamma-\gamma(n-1)\sum_{j=1}^n\log\big(1+\|z_j\|^2\big ) \bigg\}\ , { { \mathrm d}}\lambda_{2n}\big({{\mathbf z}}_{[1,n]}\big ) . \ ] ] by lemma [ lem-1.1 ] , we have @xmath267 where @xmath118 is the real number @xmath268 as we combine this with ( [ eq - a2 ] ) , we arrive at @xmath265}\big)\le \exp\bigg\{-n(n-1)\big[\gamma\,\kappa_{{\phi}}(\theta)+ ( 1-\gamma)\epsilon\big]+b\,n\\ + c\gamma n(n-1)-\gamma(n-1)\sum_{j=1}^n\log\big(1+\|z_j\|^2\big ) \bigg\}\ , { { \mathrm d}}\lambda_{2n}\big({{\mathbf z}}_{[1,n]}\big)\\ = \exp\bigg\{-n(n-1)\big[\gamma\,\kappa_{{\phi}}(\theta)+ ( 1-\gamma)\epsilon\big]+b\,n+c\gamma n(n-1)\bigg\}\\ \times\prod_{j=1}^n\big(1+\|z_j\|^2\big)^{-\gamma(n-1 ) } { { \mathrm d}}\lambda_{2n}\big({{\mathbf z}}_{[1,n]}\big)\end{gathered}\ ] ] on the set @xmath266 . considering that @xmath269 provided that @xmath270 , we find that @xmath271+b\,n\bigg\ } \\ \times\bigg(\frac{\pi}{\gamma(n-1)-1}\bigg)^n.\end{gathered}\ ] ] the dominant contribution in the expression which is exponentiated is @xmath272.\ ] ] we would like to pick @xmath273 , @xmath274 , such that @xmath275 this is possible ; indeed , without loss of generality , we may assume that @xmath262 is greater than @xmath144 , in which case one such choice is @xmath276 this value of @xmath273 yields the assertion of the proposition . the proof is complete . for a point @xmath93}\in{{\mathbb c}}^n$ ] , we define the associated weighted sum of point masses @xmath277}]\in{{\mathcal p}}_c({{\mathbb c}})$ ] by the formula @xmath278}](z)=\frac1{n}\sum_{j=1}^n{{\mathrm d}}\delta_{z_j}(z ) , \qquad z\in{{\mathbb c } } , \label{eq - sigman}\ ] ] where @xmath279 means the dirac point mass at @xmath280 . also , let @xmath281 denote the space bounded complex - valued continuous functions on @xmath17 . suppose @xmath282}\big]$ ] is as above , with @xmath2}=\big(z_1,\ldots , z_n\big)\in{{\mathbb c}}^n.\ ] ] suppose , moreover , that @xmath261}\big](\theta ) \to\kappa_{\phi}(\theta)\ ] ] as @xmath25 . then @xmath283 converges to @xmath53 , the extremal measure , in the weak - star topology , as @xmath25 . in other words , for each @xmath284 , we have @xmath285 [ prop - limitmeas ] first , fix a small but positive @xmath229 . by assumption , for sufficiently big @xmath1 , @xmath2}\in{{\mathfrak a}}_n(\epsilon,\theta),\ ] ] so that by proposition [ prop - eps ] , @xmath286 . as @xmath229 can be made as small as we like , and each @xmath283 is a probability measure , we see that @xmath287 the space of all finite complex borel measures on @xmath223 is a banach space , with weak - star compact unit ball . this means that each subsequence of the sequence @xmath288 has a weak - star convergent subsequence . we shall show that any such limit coincides with @xmath53 , from which the assertion follows . without loss of generality , then , by passing to a subsequence , we may assume that the sequence @xmath289 , @xmath73 , converges weak - star itself to a limit , which we call @xmath290 . the weak - star limit of @xmath283 is then also @xmath290 , given that @xmath291 as @xmath25 . by testing with the function @xmath292 , we see that @xmath290 is a probability measure supported on @xmath223 , so that @xmath293 . _ we claim that _ @xmath294(\theta)\le\kappa_{\phi}(\theta)$ ] . once this is established , the equality @xmath295 follows from the uniqueness of the extremal measure . we use a cut - off argument : we note that @xmath261}\big](\theta)=\frac{2\theta}{n } \sum_{j=1}^n{\phi}(z_j)+\frac1{n(n-1)}\sum_{j , k:\,j\neq k } \log\frac{1}{\|z_j - z_k\|},\ ] ] so that if @xmath296 is a real parameter , @xmath239}\big](\theta)\ge \frac{2\theta}{n } \sum_{j=1}^n{\phi}(z_j ) + \frac1{n(n-1)}\sum_{j , k=1}^n \min\bigg\{\log\frac{1}{\|z_j - z_k\|},l\bigg\}\\ -\frac{l}{n-1}.\end{gathered}\ ] ] now , let @xmath25 , so that we get , in view of our assumptions , @xmath297 as we let @xmath296 tend to @xmath42 , the right hand side approaches @xmath294(\theta)$ ] , whence the claim is immediate . as in the introduction , we let @xmath28 be the @xmath22-point marginal distribution measure and @xmath27 the @xmath22-point correlation measure for the probability distribution measure @xmath264 . we keep @xmath15 connected with @xmath50 , @xmath1 , and @xmath133 via ( [ eq - beta ] ) , and fix @xmath133 ( or , if you like , @xmath50 ) while @xmath1 grows . we have @xmath300 while , as @xmath25 , @xmath301 moreover , for each @xmath302 , we have , as @xmath25 , @xmath303}\big)\ , { { \mathrm d}}\pi^{(n)}_{n,\beta,{\phi}}\big({{\mathbf z}}_{[1,n]}\big ) \to\int_{{{\mathbb c}}^n}f\big({{\mathbf z}}_{[1,n]}\big)\ , { { \mathrm d}}\widehat{\sigma}(z_1)\cdots{{\mathrm d}}\widehat{\sigma}(z_n ) . \label{eq - limitthm}\ ] ] in other words , in the weak - star topology of measures , we have , as @xmath25 , @xmath304}\big)\to { { \mathrm d}}\widehat{\sigma}(z_1)\cdots{{\mathrm d}}\widehat{\sigma}(z_n).\ ] ] here , @xmath305 is the extremal measure . [ thm - main ] as a second step , we prove ( [ eq - limitthm ] ) under the slightly more restrictive assumption @xmath306 , which means that the test function @xmath307 has compact support . let @xmath308 be a permutation of @xmath309 . then , due to the symmetry properties of @xmath264 , @xmath310}\big)\ , { { \mathrm d}}\pi^{(n)}_{n,\beta,{\phi}}\big({{\mathbf z}}_{[1,n]}\big)= \int_{{{\mathbb c}}^n}f(z_1,\ldots , z_n)\ , { { \mathrm d}}\pi_{n,\beta,{\phi}}\big({{\mathbf z}}_{[1,n]}\big)\\ = \int_{{{\mathbb c}}^n}f(z_{\varsigma(1)},\ldots , z_{\varsigma(n)})\ , { { \mathrm d}}\pi_{n,\beta,{\phi}}\big({{\mathbf z}}_{[1,n]}\big),\end{gathered}\ ] ] from which we quickly deduce that @xmath303}\big)\ , { { \mathrm d}}\pi^{(n)}_{n,\beta,{\phi}}\big({{\mathbf z}}_{[1,n]}\big ) = \frac1{n ! } \sum_\varsigma \int_{{{\mathbb c}}^n } f\big(z_{\varsigma(1)},\ldots , z_{\varsigma(n)}\big)\ , { { \mathrm d}}\pi_{n,\beta,{\phi}}\big({{\mathbf z}}_{[1,n]}\big),\ ] ] where the sum runs over all permutations of @xmath309 . we fix a small positive @xmath229 , and split the integral : @xmath310}\big)\ , { { \mathrm d}}\pi^{(n)}_{n,\beta,{\phi}}\big({{\mathbf z}}_{[1,n]}\big)\\ = \frac{1}{n!}\int_{{{\mathfrak a}}_n(\epsilon,\theta)}\sum_\varsigma f\big(z_{\varsigma(1)},\ldots , z_{\varsigma(n)}\big)\ , { { \mathrm d}}\pi_{n,\beta,{\phi}}\big({{\mathbf z}}_{[1,n]}\big)\\ + \frac{1}{n!}\sum_\varsigma \int_{{{{\mathbb c}}^n\setminus\mathfrak a}_n(\epsilon,\theta ) } f\big(z_{\varsigma(1)},\ldots , z_{\varsigma(n)}\big)\ , { { \mathrm d}}\pi_{n,\beta,{\phi}}\big({{\mathbf z}}_{[1,n]}\big ) . \label{eq - splitint}\end{gathered}\ ] ] by proposition [ prop - picompl ] , the last term is bounded in modulus by @xmath311 for large @xmath1 . to deal with the first term on the right hand side , we should understand the behavior of @xmath312}\in{{\mathfrak a}}_n(\epsilon,\theta ) . \label{eq - sum}\ ] ] let us consider the simplest case @xmath100 first . then ( [ eq - sum ] ) amounts to @xmath313}\in{{\mathfrak a}}_n(\epsilon,\theta ) . \label{eq - sum2}\ ] ] by letting @xmath229 approach @xmath215 slowly as @xmath25 , we may ensure that ( [ eq - est3 ] ) tends to @xmath215 as @xmath25 , while ( [ eq - sum2 ] ) approaches @xmath314 it is easy to check that the latter statement entails @xmath315}\big)\to \int_{{\mathbb c}}f(z)\,{{\mathrm d}}\widehat{\sigma}(z)\ ] ] as @xmath25 , if we allow @xmath229 to approach @xmath215 slowly . the remaining case @xmath316 is handled in an analogous manner : proposition [ prop - limitmeas ] should be replaced by a multidimensional analogue , which we get by iterated integration . it is useful to keep in mind that for large @xmath1 and fixed @xmath22 , the following collections of sequences are asymptotically ( as @xmath25 ) the same : @xmath317 where @xmath308 runs over all permutations of @xmath10 , and @xmath8 runs over all functions @xmath318 . for instance , @xmath319 there are a couple of assertion that remain to check . if we notice that the support of @xmath53 is actually contained in the interior of the compact set @xmath223 , then , by choosing a smooth cut - off function @xmath320 with @xmath321})\le1 $ ] everywhere , @xmath322})=1 $ ] on @xmath323 , and @xmath324 off @xmath325 , we find from ( [ eq - limitthm ] ) that @xmath326 since we are dealing with probability measures , it follows that @xmath327 as claimed . it remains to verify that ( [ eq - limitthm ] ) holds for all @xmath328 . to this end , we use a smooth cut - off function @xmath329 similar to what we defined above , and write @xmath330 . then @xmath331 , while the integral of the remainder @xmath332 tends to @xmath215 as @xmath25 , by a simple estimate . let us think of the @xmath1-tuple @xmath93}=(z_1,\ldots , z_n)$ ] as a random variable taking values in @xmath3 , with distribution measure @xmath264 . the marginal distribution of the first @xmath22 ( @xmath23 ) coordinates @xmath95}=(z_1,\ldots , z_n)$ ] is then given by the measure @xmath333 ; to stress the dependence on @xmath1 , let us write @xmath24}\|_n=\big(z_1\|_n,\ldots , z_n\|_n\big)\ ] ] for the marginal random variable . now , if @xmath15 grows with @xmath1 according to ( [ eq - beta ] ) , for some fixed positive @xmath133 , theorem [ thm - main ] says that as @xmath25 , @xmath334 become identically distributed independent random variables , each having @xmath335 as distribution measure . in particular , the covariance type matrix ( with @xmath220 ) @xmath336},{{\mathbf z}}_{[1,n]})= \big(\,k_{\mu , n}(z_j , z_k)\big)_{j , k=1}^n\ ] ] becomes asymptotically diagonal as @xmath25 . we need to give an example to illustrate the result . we consider the rather trivial case of @xmath337 , and put @xmath100 . then @xmath338 is constant . a computation shows that with @xmath339 given by ( [ eq - mu ] ) , the reproducing kernel function for the polynomial subspace is @xmath340 we then have @xmath341 while it is easy to check that @xmath342 in view of this , the content of theorem [ thm - main ] in this simple case is that , for fixed positive @xmath50 , the function @xmath343 which is real - valued , with values between @xmath215 and @xmath128 , tends to @xmath128 as @xmath25 for @xmath344 , and to @xmath215 for @xmath345 . this fact is of course well - known . in @xcite , the domain @xmath60 is computed explicitly and shown to be an ellipse in the more general case ( still with constant @xmath346 ) @xmath347 p. deift , _ orthogonal polynomials and random matrices : a riemann - hilbert approach . courant lecture notes in mathematics , * 3. new york university , courant institute of mathematical sciences , new york ; american mathematical society , providence , ri , 1999 . _ p. elbau , g. felder , _ density of eigenvalues of random normal matrices . preprint 2004 . w. fulton , _ eigenvalues , invariant factors , highest weights , and schubert calculus . ( n.s . ) 37 * ( 2000 ) , no . 3 , 209249 . _ _ i. k. kostov , i. krichever , m. mineev - weinstein , p. b. wiegmann , a. zabrodin , _ @xmath50-function for analytic curves . _ random matrix models and their applications _ , 285299 , math . publ . , * 40 * , cambridge univ . press , cambridge , 2001 . i. krichever , a. marshakov , a. zabrodin , _ integrable structure of the dirichlet boundary problem in multiply - connected domains . phys . , to appear . a. marshakov , p. wiegmann , a. zabrodin , _ integrable structure of the dirichlet boundary problem in two dimensions . * 227 * ( 2002 ) , no . 1 , 131153 . m. l. mehta , _ random matrices . second edition . academic press , inc . , boston , ma , 1991 . _ _ _ _ 3 ] p. b. wiegmann , _ aharonov - bohm effect in the quantum hall regime and laplacian growth problems . statistical field theories ( como , 2001 ) , 337349 , nato sci . ser . ii math . , * 73 * , kluwer acad . publ . , dordrecht , 2002 .	in this note , we discuss the quantum hele - shaw flow , a random measure process in the complex plane introduced by the physicists p.wiegmann , a. zabrodin , et al . this process arises in the theory of electronic droplets confined to a plane under a strong magnetic field , as well as in the theory of random normal matrices . we extend a result of elbau and felder @xcite to general external field potentials , and also show that if the potential is @xmath0-smooth , then the quantum hele - shaw flow converges , under appropriate scaling , to the classical ( weighted ) hele - shaw flow , which can be modeled in terms of an obstacle problem .
thanks to the recent works of schramm , sheffield , dubdat , miller ( see @xcite and the references therein ) , we know that many structures built using schramm s sle curves can be naturally coupled with the planar gaussian free field ( gff ) . for instance , even though the gff is not a continuous function , sle@xmath0 and the conformal loop ensemble cle@xmath0 can be viewed as `` level lines '' of the gff . in these couplings , the notion of _ local sets of the gff _ and their properties turn out to be instrumental . this general abstract concept appears already in the study of markov random field in the 70s and 80s ( see in particular @xcite ) and can be viewed as the natural generalization of stopping times for multidimensional time . more precisely , a random set @xmath1 coupled with a ( zero - boundary ) gff in a planar domain @xmath2 is said to be a local set for this gff , if the conditional distribution of the gff in @xmath3 given @xmath1 has the law of the sum of a ( conditionally ) independent gff @xmath4 in @xmath3 with a random harmonic function @xmath5 defined in @xmath3 . this harmonic function can be interpreted as the harmonic extension to @xmath3 of the values of the gff on @xmath6 . in the present paper , we are interested in a special type of local sets that we call _ bounded type thin local sets _ ( or btls ) . our definition of btls imposes three type of conditions , on top of being local sets : 1 . there exists a constant @xmath7 such that almost surely , @xmath8 in the complement of @xmath1 . 2 . the set @xmath1 is a thin local set in the sense defined in @xcite . this means that the harmonic function @xmath5 carries itself also all the information about the gff also on @xmath1 , in the sense that for any smooth test function @xmath9 , the random variable @xmath10 is almost surely equal to @xmath11 . 3 . almost surely , each connected component of @xmath1 that does not intersect @xmath12 has a neighborhood that does intersect no other connected component of @xmath1 . if the set @xmath1 is a btls with the constant @xmath7 in the first condition , we say that it is a @xmath7-btls . the first two conditions are the most important ones . the first condition is of course highly non - trivial . notice that a deterministic non - polar set is not a btls because the corresponding harmonic function is not bounded , so that a non - empty btls is necessarily random . as we shall see , the second condition can be understood as a condition on the size of the set @xmath1 ( it can not be too large ) and can be compared in the previous analogy between local sets and stopping times to requiring stopping times to be uniformly integrable . the third condition implies in particular that @xmath1 has only countably many connected components that do not intersect @xmath12 . this is somewhat restrictive as there exist totally disconnected sets cantor sets @xmath1 that are non - polar for the brownian motion ( for instance a cantor set with hausdorff dimension in @xmath13 ) , and can therefore have a non - negligible boundary effect for the gff . we however believe that this third condition is not essential and could be disposed of ( ie . all statements would still hold without this condition ) , and we comment on this at the end of the paper . we remark that the precise choice of definitions is not that important here . we see , in fact , that the combination of these three conditions implies stronger statements . for instance , the ( upper ) minkowski dimension of such a @xmath7-btls @xmath1 is necessarily bounded by some constant @xmath14 that is smaller than 2 ( and this implies the second condition ) , and that @xmath15 has almost surely only one connected component ( which implies the third condition ) . non - trivial btls do exist , and the main examples are provided by sle@xmath0 , cle@xmath0 and their variants ( for their natural coupling with the gff ) as shown in @xcite . in this paper , among other things , we prove that any btls @xmath1 is contained in a nested version of cle@xmath0 and that @xmath16 is necessarily connected . from our proofs it also follows that the cle@xmath0 and its various generalizations are deterministic functions of the gff . keeping the stopping time analogy in mind , one can compare these results with the following feature of one - dimensional brownian motion @xmath17 started from the origin : if @xmath18 is a stopping time with respect to the filtration of @xmath17 such @xmath19 almost surely , then either @xmath20 almost surely , or @xmath18 is not uniformly integrable . in other words , the cle@xmath0 and its nested versions and variants are the field analogues of one - dimensional exit times of intervals . one important general property of local sets , shown in @xcite and used extensively in @xcite , is that when @xmath1 and @xmath17 are two local sets coupled with the same gff , in such a way that @xmath1 and @xmath17 are conditionally independent given the gff , then their union @xmath21 is also a local set . it seems quite natural to expect that it should be possible to describe the harmonic function @xmath22 simply in terms of @xmath5 and @xmath23 , but deriving a general result in this direction appears to be , somewhat surprisingly , tricky . the present approach provides a way to obtain results in this direction in the case of btls : we shall see that the union of two btlss is always a btls . we plan to generalize in @xcite some features of our approach to the imaginary geometry setting , and hope that this might be of some help to enlighten some of the technical issues dealt with in @xcite . we now state more precisely some of the results that we shall derive . all of the results we state in this section hold for a simply connected domain @xmath2 . let us first briefly recall some features of the coupling between cle@xmath0 and gff in such a domain . using a branching - tree variant of sle@xmath0 introduced by sheffield in @xcite , it is possible to define a certain random conformally invariant family of marked open sets @xmath24 , where the @xmath25 s form a disjoint family of open subsets of the upper half - plane , and the marks @xmath26 belong to @xmath27 . the complement @xmath1 of @xmath28 is called a cle@xmath0 carpet ( see figure [ figcle4 ] ) and its minkowski dimension is in fact almost surely equal to @xmath29 ( see @xcite ) . as pointed out by miller and sheffield @xcite , the set @xmath1 can be coupled with the gff as a btls in a way that @xmath5 is constant and equal to @xmath30 in each @xmath25 . the set in black represents the cle@xmath0 carpet and the white holes are the open sets @xmath25 . , width=297 ] simulation by david b. wilson . by the property of local sets , conditionally on @xmath1 , the field @xmath31 consists of independent gaussian free fields @xmath32 inside each @xmath25 . we can then iterate the same construction independently for each of these gffs @xmath32 , using a cle@xmath0 in each @xmath25 . in this way , for each given @xmath33 , if we define @xmath34 to be the open set @xmath25 that contains @xmath35 ( for each fixed @xmath35 , this set almost surely exists ) and set @xmath36 , we then get a second set @xmath37 corresponding to the nested cle@xmath0 in @xmath34 and a new mark @xmath38 . iterating the procedure , we obtain for each given @xmath35 , an almost surely decreasing sequence of open sets @xmath39 and a sequence of marks @xmath40 in @xmath41 . when @xmath42 , then @xmath43 and @xmath44 , so that @xmath45 can be viewed as a constant function in @xmath39 . furthermore , for each @xmath46 , the sequence @xmath47 is a simple random walk . for each @xmath48 , the complement @xmath49 of the union of all @xmath39 with @xmath50 is then a btls and the corresponding harmonic function @xmath51 is just @xmath52 . the set @xmath49 is called the nested cle@xmath0 of level @xmath48 ( we refer to it as cle@xmath53 in the sequel ) . it is in fact easy to see that one can recover the gff @xmath54 from the knowledge of all these @xmath55 for @xmath56 ( because for each smooth test function @xmath9 , the sequence @xmath57 converges to @xmath58 in @xmath59 ) . for each integer @xmath60 , one can then define for each @xmath35 in @xmath61 , the random variable @xmath62 to be the smallest @xmath63 at which @xmath64 . as for each fixed @xmath35 , the sequence @xmath65 is just a simple random walk , @xmath62 is almost surely finite . the complement @xmath66 of the union of all these @xmath67 for @xmath35 with rational coordinates defines a @xmath68-btls : indeed , the corresponding harmonic function is then constant in each connected component @xmath69 and takes values in @xmath70 , and we explain in section [ cle4 ] ) that the minkowski dimension of @xmath66 is almost surely bounded by @xmath71 which imply the thinness . we refer to this set @xmath66 as a cle@xmath72 ( mind that @xmath73 is in superscript , as opposed to cle@xmath74 , where we just iterated @xmath63 times the cle@xmath0 ) . a special case of our results is that this @xmath75 is the only btls with harmonic function taking its values in @xmath76 . more precisely : [ cledesc ] suppose that @xmath1 is a btls coupled to the same gff as a cle@xmath72 , and such that the corresponding harmonic function @xmath5 takes values in @xmath77 . then , @xmath1 is almost surely equal to the cle@xmath72 . in particular , for @xmath78 and taking @xmath1 to be another cle@xmath0 coupled with the same gff , this means that any two @xmath79 that are coupled with the same gff as local sets are almost surely identical . this btls approach therefore provides a rather short proof of the fact the first layer cle@xmath0 and all nested cle@xmath74 are deterministic functions of the gff ( hence , that the information provided by the collection of nested labelled cle@xmath74 s is equivalent to the information provided by the gff itself ) . this fact is not new and is due to miller and sheffield @xcite , who have outlined the proof in private discussions , presentations and talks ( and a paper in preparation ) . our proof follows a slightly different strategy than the one proposed by miller and sheffield , with a different angle , although the basic ingredients are very similar ( absolute continuity properties of the gff , basic properties of local sets and the fact that sle@xmath0 itself is a deterministic function of the gff ) . let us stress that here and in subsequent propositions , the condition of being thin is necessary . indeed , if we consider the union of a cle@xmath0 with , say , the component of its complement that contains the the origin , we still obtain a local set for which the corresponding harmonic function is in @xmath77 , yet it is almost surely clearly not contained in any of the @xmath75 . we also characterize all possible btls such that the harmonic function can take only two possible prescribed values , and study their properties : [ cledesc2 ] let us consider @xmath80 . 1 . when @xmath81 , it is not possible to construct a btls @xmath1 with @xmath82 . when @xmath83 , it is possible to construct a btls @xmath84 coupled with a gff @xmath54 in such a way that @xmath85 . moreover , the sets @xmath84 are * unique in the sense that if @xmath86 is another btls coupled with the same @xmath54 , such that for all @xmath46 , @xmath87 almost surely , then @xmath88 almost surely . * measurable functions of the gff @xmath54 that they are coupled with . * monotonic in the following sense : if @xmath89 \subset [ a ' , b']$ ] and @xmath80 with @xmath90 , then almost surely , @xmath91 . further information about the sets @xmath84 and their properties ( detailed construction , dimensions ) as well as some generalizations are discussed in section [ s6 ] . another type of result that we derive using similar ideas as in the proof of proposition [ cledesc ] goes as follows : [ maxbtls ] if @xmath1 is a @xmath92-btls associated to the gff @xmath54 , @xmath1 is almost surely a subset of the @xmath93 associated to @xmath54 . notice that one would expect to conclude that @xmath94 in the proposition ( and this would mean that @xmath75 is maximal among all @xmath68-btls ) , but this seems to require some more technical work that we do not discuss in the present paper . as an immediate consequence , we see that the finite union of btls is almost surely contained in some @xmath75 for some given @xmath73 , and that it is again a btls . this type of facts helps to derive the above - mentioned fact : [ propnodisconnection ] if @xmath1 is a btls , then @xmath16 is connected . the structure of the paper is the following : we first recall some basic features about btls . then , we consider generalized level lines of the gff and discuss a useful result about their boundary hitting behaviour . thereafter , we recall features of the construction of the coupling of the gff with @xmath75 . this sets the stage for the proofs of the propositions involving cle@xmath72 . we then finally turn to proposition [ cledesc2 ] and the discussion about the dimension of the sets @xmath84 . [ s3 ] in this section , we quickly browse through basic definitions and properties of the gff and of bounded - type local sets . we only discuss items that are directly used in the current paper . for a more general discussion of local sets , thin local sets ( not necessarily of bounded type ) , we refer to @xcite . throughout this paper , the set @xmath2 denotes an open planar domain with a non - empty and non - polar boundary . by conformal invariance , we can always assume that @xmath2 is a subset of the unit disc @xmath95 . the most general case that we work with are domains @xmath2 such that the complement of @xmath2 has at most countably many connected components with only finitely many components in each compact set and no complement being a singleton ( this is a non - trivial condition as it excludes for instance sets like @xmath96 , where @xmath97 $ ] is the middle cantor set ) . recall that by a theorem of he and schramm @xcite , such domains @xmath2 are known to be conformally equivalent to a circle domain ( i.e. to @xmath96 , where @xmath7 is a union of closed disjoint discs ) . recall that the ( zero boundary ) gaussian free field ( gff ) in a domain @xmath2 can be viewed as a centered gaussian process @xmath54 ( we also sometimes write @xmath98 when we the domain needs to be specified ) indexed by the set of continuous functions with compact support in @xmath2 , with covariance given by @xmath99 = \iint_{d\times d } f(x ) g_d(x , y ) g(y ) { { d}}x { { d}}y\ ] ] where @xmath100 is the green s function ( with dirichlet boundary conditions ) in @xmath2 , normalized such that @xmath101 as @xmath102 . for this choice of normalization of @xmath103 ( and therefore of the gff ) , we set @xmath104 sometimes , other normalizations are used in the literature : if @xmath105 as @xmath102 , then @xmath106 should be taken to be @xmath107 . note that it is in fact possible and useful to define the random variable @xmath108 for any fixed borel measure @xmath109 , provided the energy @xmath110 is finite . similarly to brownian motion , the gff can be also viewed as the gaussian measure associated to the dirichlet space @xmath111 , which is the closure of the set of smooth functions of compact support in @xmath2 with respect to the dirichlet norm @xmath112 the dirichlet space is also the cameron - martin space for the gff ; in particular the law of @xmath113 is absolutely continuous with respect to the law of @xmath54 whenever @xmath114 , and the radon - nikodym derivative is given by a constant multiple of @xmath115 , see e.g. @xcite . the covariance kernel of the gff blows up on the diagonal , which makes it impossible to view @xmath54 as a random function . it can , however , be shown that the gff has a version that lives in some space of generalized functions , what justifies the notation @xmath10 for @xmath54 acting on functions @xmath9 ( see for example @xcite ) . however , it is not possible to make sense of @xmath10 when @xmath116 is a the indicator function of an arbitrary random set @xmath1 . local sets form a class of random sets where this is ( in a sense ) possible : consider a random triple @xmath117 , where @xmath54 is a gff in @xmath2 , @xmath1 is a random relatively closed subset of @xmath2 ( i.e. @xmath3 is open ) and @xmath118 a random distribution that can be viewed as a harmonic function when restricted to @xmath3 . we say that @xmath1 is a local set for @xmath54 if conditionally on @xmath1 , @xmath119 is a gff in @xmath3 . notice that being a local set can also be seen as a property of the law of the couple @xmath120 , as one can recover @xmath118 as the conditional expectation of @xmath54 given @xmath1 . let us list a few properties of local sets ( see for instance @xcite for derivations and further properties ) : [ bpls ] @xmath121 1 . any local set can be coupled in a unique way with a given gff : if two random triples @xmath117 and @xmath122 are both local couplings , then @xmath118 and @xmath123 are almost surely identical . 2 . when @xmath1 and @xmath17 are local sets coupled with the same gff @xmath54 , and that @xmath124 and @xmath125 are conditionally independent given @xmath54 , then @xmath21 is also a local set coupled with @xmath54 . we now define local sets of bounded type : [ btlscnd ] consider a random relatively closed subset @xmath1 of @xmath2 ( i.e. so that @xmath3 is open ) and @xmath54 a gff defined on the same probability space . let @xmath126 , we say that @xmath1 is a @xmath7-btls if the following four conditions are satisfied : 1 . @xmath1 is a local set of @xmath54 . 2 . almost surely , @xmath127 in @xmath3 . 3 . almost surely , each connected component of @xmath1 that does not intersect @xmath12 has a neighborhood that does intersect no other connected component of @xmath1 . 4 . @xmath1 is a thin local set in the sense defined in @xcite : for any smooth test function @xmath128 , the random variable @xmath10 is almost surely equal to @xmath11 . if @xmath1 is a @xmath7-btls for some @xmath7 , we say that it is a btls . note also that it is in fact possible to remove all isolated points from a btls ( of which there are at most countably many because of the third property ) without changing the property of being btls . indeed , the bounded harmonic function @xmath5 can be extended to those points and a gff does not see polar sets . we therefore replace the third property by : _ ( 3 ) _ : : almost surely , @xmath1 contains no isolated points and each connected component of @xmath1 that does not intersect @xmath12 has a neighborhood that does intersect no other connected component of @xmath1 . this reformulation will be handy to keep our statements simple . it is not hard to see that in the case where the harmonic function is bounded , the condition of being thin given above is equivalent to saying that for the upper dyadic approximation @xmath129_n$ ] of @xmath1 , and for any smooth test function @xmath9 , the sequence of random variables @xmath130_n})$ ] converges in probability to @xmath58 . from a borel - cantelli argument , one can moreover see that this equivalent condition is implied by ( see @xcite ) : @xmath131 : : the expected volume of the @xmath132-neighborhood of @xmath1 decays like @xmath133 as @xmath134 . note that if @xmath1 and @xmath17 are two btls that are coupled with the same gff @xmath54 , and that are conditionally independent given @xmath54 , then we know by lemma [ bpls ] that @xmath21 is a local set . hence , in order to check that @xmath21 is a btls , we need to show it is thin and give an upper bound for the harmonic function @xmath22 . it is not hard to derive the following related fact : [ local subset ] let @xmath1 be a thin local set that satisfies the condition ( 3 ) of definition [ btlscnd ] , and let @xmath86 be a @xmath7-btls coupled with the same gff in such a way that @xmath135 almost surely . then @xmath1 is a @xmath7-btls . one just has to check that @xmath136 . for any smooth @xmath137 with compact support have that @xmath138 = \gamma_a ( \rho)$ ] and @xmath139 = \gamma_{a ' } ( \rho)$ ] almost surely . now as @xmath118 is harmonic in @xmath140 , we have that for any @xmath141 and any radially symmetric @xmath142 of unit mass and with support inside @xmath140 , @xmath143 . the claim now follows from jensen s inequality for conditional expectations . after showing that any btls is contained in some cle@xmath144 and checked that cle@xmath72 satisfies indeed @xmath131 , this lemma is used to show the union of any two btls is a again a btls . the following lemma is very useful in our proof of the fact that any btls is contained in some cle@xmath72 : [ mes ] let @xmath1 and @xmath17 are two btls coupled to the same gff and suppose that almost surely , @xmath1 satisfies condition @xmath131 . then , if almost surely , for all @xmath145 , @xmath146 , we have that almost surely @xmath147 . first , let us briefly explain why the conditions on @xmath1 and @xmath17 imply that @xmath21 is a thin local set . we know that @xmath21 is local . to show that @xmath148 is thin , we use the equivalent condition mentioned above . indeed , note that @xmath149_n})$ ] can be written as the sum of @xmath150_n})$ ] and @xmath151_n\backslash [ b]_n})$ ] . the first term converges to 0 thanks to the definition and the second one thanks to condition ( @xmath152 ) . we refer to @xcite for a more detailed discussion . now we know that @xmath153 , and we know the conditional distribution of @xmath4 ( resp . @xmath154 ) given the @xmath155-field @xmath156 generated by @xmath157 ( resp . the @xmath155-field @xmath158 the fact that one can use this larger @xmath155-field instead of @xmath159 follows directly from the proof of the fact that @xmath21 is local ) . we also denote by @xmath160 ( resp . @xmath161 ) the set of points in @xmath3 where @xmath5 is non - negative ( resp . non - positive ) . then , for any open set @xmath162 , @xmath163 & = & { \mathbb e}\big [ { \mathbb e}\left [ ( h_a ( { \mathbf 1}_o ) + \gamma^a ( { \mathbf 1}_o))^2 \| { \mathcal a}\right ] { \mathbf{1}_{\left \{o \subset a_+\right \}}}\big ] \\ & = & { \mathbb e}\left [ \left ( h_a ( { \mathbf 1}_o)^2 + \int_{o \times o } g_{d \setminus a } ( x , y ) { { d}}x { { d}}y \right ) { \mathbf{1}_{\left \{o \subset a_+\right \}}}\right].\end{aligned}\ ] ] similarly , by conditioning on @xmath164 , one gets that this same quantity is equal also to @xmath165 .\ ] ] but by definition , when @xmath166 , @xmath167 ( because @xmath168 for all @xmath169 ) . hence , using the fact that @xmath170 , we conclude that for every open set , almost surely on the event @xmath166 , @xmath171 on @xmath172 . the same statement holds for @xmath161 instead of @xmath160 . therefore @xmath173 is polar in @xmath3 and condition ( 3 ) allows us to conclude . let us first recall some well - known features of the schramm - sheffield coupling of sle@xmath0 with the gff in the upper half - plane @xmath61 . consider the bounded harmonic function @xmath174 in the upper - half plane with boundary values @xmath175 on @xmath176 and @xmath177 on @xmath178 . suppose that @xmath179 is a random simple curve defined in the closed upper half - plane , such that @xmath180 , @xmath181 and the time @xmath182 is the ( possibly infinite ) smallest positive time @xmath183 at which @xmath184 . then , there exists a unique way to couple @xmath185 with a gff @xmath54 in such a way that the following property holds ( we state it in a somewhat strange way that will be easier to generalize ) : @xmath186 : : for all @xmath183 , the set @xmath187 $ ] is a btls of the gff @xmath54 , with harmonic function @xmath188 defined as follows : @xmath189 is the unique bounded harmonic function in @xmath190 $ ] with boundary values @xmath175 on the left - hand side of @xmath185 , @xmath191 on the right side of @xmath185 , and with the same boundary values as @xmath192 on @xmath176 and @xmath178 . furthermore , this curve is then a sle@xmath0 ( and @xmath193 almost surely ) , and when one couples sle@xmath0 with a gff @xmath54 in this way , then the sle@xmath0 process is in fact a measurable function of the gff ( see @xcite for all these facts ) . this statement can be generalized to more general harmonic functions @xmath194 , which gives rise to sle@xmath195-type processes . see for instance @xcite . let @xmath196 , where @xmath7 is a countable union of closed discs such that in any compact set there are only finitely many of them . consider a finite union @xmath197 of closed dyadic squares in @xmath2 with @xmath198 , and a harmonic function @xmath199 in @xmath2 such that @xmath199 extends continuously to an open neighborhood @xmath200 of @xmath201 in @xmath202 in such a way that @xmath203 on @xmath200 ( we then say that the boundary value of @xmath199 on @xmath201 is zero ) . using absolute continuity properties of the gff , it is not hard to see that the gff @xmath54 and @xmath204 are mutually absolutely continuous , when restricted to @xmath197 ( the proof of the following lemma can be found for example in @xcite ) : [ abscnty ] let @xmath54 be a ( zero boundary ) gff in @xmath2 and denote its law restricted to @xmath197 by @xmath205 . let also @xmath206 be the law of @xmath204 , restricted to @xmath197 . then @xmath205 and @xmath206 are mutually absolutely continuous with respect to each other . moreover , the radon - nikodym derivative @xmath207 is a multiple of @xmath208 where @xmath209 is the bounded harmonic function in @xmath210 that is equal to @xmath199 on the boundary of @xmath197 and to zero on the boundary of @xmath2 . furthermore , if @xmath211 is another domain as above , @xmath212 is a gff in @xmath211 and @xmath213 is at positive distance of @xmath214 , then the laws of @xmath54 and @xmath212 restricted to @xmath197 are mutually absolutely continuous . this absolute continuity property allows to change boundary conditions for local couplings away from the local sets : [ change boundary ] under the previous conditions , suppose that @xmath1 is a btls for @xmath54 an @xmath2-gff such that @xmath215 almost surely . define @xmath216 . then @xmath1 is coupled as a local set with @xmath217 ( that is a @xmath206 zero - boundary gff ) . the corresponding harmonic function @xmath218 is equal to the unique bounded harmonic function on @xmath3 such that @xmath219 goes to @xmath58 on @xmath220 , and that @xmath221 goes to @xmath58 on @xmath222 . notice that by the cameron - martin theorem for the gff ( see section [ s3 ] ) , the law of @xmath54 under @xmath223 is that of @xmath224 , where @xmath225 is a @xmath206 gff . now we have to verify that with this change of variables @xmath1 is still a local set . note that conditioned to @xmath226 we can write @xmath227 , where @xmath4 is a @xmath205-gff on @xmath140 . we have to show that under @xmath206 , we can write @xmath228 , where @xmath229 is a @xmath206-gff on @xmath140 and @xmath218 is the unique bounded harmonic function on @xmath230 such that @xmath219 goes to @xmath58 on @xmath220 , and that @xmath221 goes to @xmath58 on @xmath231 . let @xmath73 be a measurable function of the field @xmath4 , then @xmath232 & = z_a { \mathbb e}\left [ m(\gamma^a ) \exp\left ( ( \tilde h,\gamma)_{\nabla } \right ) \mid a,\gamma_a\right ] \\ & = \bar z_a { \mathbb e}\left [ m(\gamma^a ) \exp\left ( ( \tilde h,\gamma^a)_{\nabla } \right ) \mid a,\gamma_a\right ] \\ & = \bar z_a { \mathbb e}\left [ m(\gamma^a ) \exp\left ( ( \tilde h+\tilde h_a - h_a,\gamma^a)_{\nabla } \right ) \mid a,\gamma_a\right ] , \end{aligned}\ ] ] where in the last equality we use that @xmath4 is a gff and that @xmath233 is harmonic in @xmath3 . but , we know that ( under @xmath205 ) , conditionally on @xmath234 , @xmath4 is just a gff . thus using again the cameron - martin theorem and lemma [ bpls]-(i ) , we can conclude . [ change boundary2 ] let @xmath235 be another domain with the same properties . under the previous conditions , suppose that @xmath1 is a btls for a gff in @xmath2 such that @xmath236 almost surely . then , there exist an absolutely continuous probability measure @xmath206 under which @xmath1 is coupled as a local set with a @xmath206 zero - boundary gff in @xmath211 and the corresponding harmonic function @xmath218 is equal to the unique bounded harmonic function in @xmath211 such that @xmath219 goes to @xmath58 on @xmath220 , and that @xmath221 goes to @xmath58 on @xmath237 . it can be shown that @xmath1 is a local set for @xmath238 such that its harmonic function @xmath239 goes to 0 on @xmath240 and @xmath241 goes to @xmath58 on @xmath220 ( see lemma 3.9 - 1 of @xcite , or @xcite ) . we conclude using the fact that @xmath242 is independent of @xmath243 and proposition [ change boundary ] . let @xmath2 be as above and let @xmath199 be a harmonic function on @xmath2 with zero boundary conditions on some real neighbourhood @xmath201 of the origin . then we can define the generalized level line for the gff in @xmath2 with boundary conditions @xmath244 up to the first time it hits the boundary : [ generalized level line ] [ defll ] there exists a unique and measurable simple curve @xmath185 coupled with the gff in such a way that @xmath186 holds when one replaces @xmath192 by @xmath244 and considers @xmath2 instead of @xmath61 . we call @xmath185 the generalized level line of @xmath245 in @xmath2 . let @xmath246 denote the collection of regions @xmath197 such that @xmath197 is a finite union of closed dyadic squares in @xmath2 with @xmath247 and with @xmath197 simply connected . note that is enough to show that for all @xmath248 there is at most one curve satisfying @xmath186 , when one replaces @xmath192 by @xmath244 , until the time it exits @xmath197 . indeed , suppose by contradiction that @xmath249 and @xmath250 are two different curves with this property and that with positive probability they do not agree . thanks to corollary [ change boundary2 ] , we can construct two local sets @xmath251 and @xmath252 coupled with the same gff in a simply connected domain @xmath235 such that when we apply @xmath253 , the conformal transformation from @xmath211 to @xmath61 , @xmath186 holds for @xmath254 and @xmath255 until the time they exit @xmath61 . the fact that they are different with positive probability contradicts the uniqueness of the schramm - sheffield coupling proved in citedub , schsh2 . it remains to show the existence of this curve . take @xmath256 an increasing sequence of elements in @xmath257 such that their union is @xmath2 . we can define @xmath185 until the time it exits @xmath256 , by using the schramm - sheffield coupling and proposition [ change boundary2 ] . the above argument shows that they are compatible and so we can define @xmath185 until the first time it touches @xmath12 . the measurability is just a consequence of the existence and uniqueness . in @xcite lemma 3.1 , the authors show that a generalized level line in @xmath61 corresponding to @xmath258 with @xmath259 zero in a neighbourhood of the origin , @xmath260 on @xmath261 and @xmath262 elsewhere can not touch @xmath263 $ ] in finite time . combined with our previous considerations , we extend this to : [ notouch ] let @xmath185 be a generalized level line of @xmath264 in @xmath2 such that @xmath199 is harmonic in @xmath2 and has zero boundary values in a neighbourhood @xmath201 of zero . suppose @xmath265 in some neighbourhood @xmath266 of @xmath267 . let @xmath182 denote the first time at which @xmath268 . then the probability that @xmath269 and that @xmath270 converges to a point in @xmath266 as @xmath271 is equal to @xmath58 . this also holds if @xmath272 in @xmath273 and more generally for any open neighbourhood of @xmath274 on @xmath12 with similar conditions . let @xmath197 be a finite union of closed dyadic squares in the upper half - plane with @xmath275 , @xmath276 , @xmath277 and @xmath278 . from lemma [ defll ] we know that for any such @xmath197 , a generalized level line @xmath185 is measurable up to leaving @xmath197 w.r.t . the gff restricted to @xmath197 . we first consider the case @xmath279 and suppose @xmath280 . define a harmonic function @xmath259 on @xmath61 such that @xmath281 on @xmath282 , @xmath283 on @xmath284 and @xmath285 on @xmath286 $ ] . then gffs @xmath287 and @xmath264 restricted to @xmath197 are absolutely continuous by lemma [ abscnty ] . also , both of these generalized level lines are measurable w.r.t . the gff until they exit @xmath197 . now @xmath288 satisfies the conditioned mentioned above and hence almost surely the generalized level line of @xmath287 does not exit from @xmath197 on @xmath289 in finite time . by absolute continuity and measurability of both generalized level lines , the result follows . the case @xmath290 is treated similarly . to treat the case where @xmath274 is on the boundary of some component @xmath291 of @xmath7 , notice that it suffices to consider @xmath197 that do not surround @xmath291 and that for any such @xmath197 we can connect @xmath291 to @xmath202 using some curve @xmath292 such that @xmath293 is simply connected and contains @xmath197 . the previous argument and corollary [ change boundary2 ] show that @xmath185 can not leave @xmath197 through @xmath294 in finite time . in this section , we review the coupling of the gff with the nested cle@xmath0 using sheffield s sle@xmath0 exploration trees ( i.e. the branching sle@xmath296 process ) . as this section does not contain really new results , we try to be rather brief , and refer to @xcite for details . note that we also provide an alternative construction of cle@xmath0 and its coupling to the gff in section [ s6 ] . let us first recall the definition of the radial ( symmetric ) sle@xmath298 process targeted at @xmath58 ( for non - symmetric variants see @xcite ) . it is the radial loewner chain @xmath299 in the unit disc , with driving function @xmath300 defined as follows : * start with a standard real - valued brownian motion @xmath301 with @xmath302 . * then , define the continuous process @xmath303 . * finally , set @xmath304 and @xmath305 . we set @xmath306 , and denote by @xmath307 to be the conformal maps from @xmath308 onto @xmath95 normalized at the origin . notice that as the integral @xmath309 is almost surely infinite , the definition of @xmath310 can not be viewed as an usual absolutely converging integral . yet , one can for instance define first @xmath311 as the integral of @xmath312 and show that these approximating processes @xmath311 converge in @xmath59 to a continuous processes @xmath310 . see e.g. @xcite for a discussion of the chordal analogue . we now describe some properties of the hulls generated by the radial sle@xmath297 . for more details , see @xcite : 1 . the process @xmath313 describes the evolution of the marked point in the sle@xmath314 framework . in particular , sle coordinate change considerations ( see @xcite ) show that during the time - intervals during which @xmath315 is not in @xmath316 , the process behaves exactly like a chordal sle@xmath0 targeting the boundary prime end @xmath317 in the domain @xmath308 . + sketch of @xmath318 , the stretch @xmath319 $ ] . the harmonic measure from @xmath58 of the `` top part '' of @xmath320 $ ] corresponds to @xmath315 . this loop corresponds to an excursion away from @xmath316 by each of @xmath17 , @xmath321 and @xmath322 . it corresponds to an increase of @xmath323 for @xmath17 and for @xmath324 , but not for @xmath322 . the end - time of this loop is @xmath325.,width=297 ] 2 . it is therefore clear that during the open time - intervals at which @xmath17 is not in @xmath316 , the loewner chain is generated by a simple continuous curve . these excursions of @xmath17 away from @xmath316 correspond to the intervals during which the sle@xmath297 traces `` quasi - loops '' : if @xmath326 is an excursion interval away from @xmath316 , then @xmath327 is the image of a loop from @xmath328 to itself in @xmath95 via @xmath329 . when @xmath330 , then this quasi - loop does not surround the origin , when @xmath331 , it surrounds the origin clockwise or anti - clockwise . we stress that during the time - interval corresponding to excursions of @xmath17 away from @xmath316 the tip of the loewner chain does not touch its past boundary and at the end of the excursion it accumulates at the prime end corresponding to the starting point . it is also possible to see ( via the loop - soup construction of cle@xmath0 , see @xcite ) that these quasi - loops are in fact proper loops ( i.e. that @xmath332 is a proper boundary point ) , but we do not , in fact , need this fact in the present paper . in fact , it can be shown that the whole sle@xmath297 is _ generated _ by a two - dimensional curve , see @xcite . for any other @xmath333 , one can define the radial sle@xmath297 from @xmath334 to @xmath35 in @xmath95 to be the conformal image of the radial sle@xmath297 targeting the origin , by the moebius transformation of the unit disc that maps @xmath58 to @xmath35 , and @xmath334 onto itself . this is now a loewner chain growing towards @xmath35 , and it is naturally parametrized using the log - conformal radius seen from @xmath35 . moreover the radial sle@xmath296 processes for different target points can be coupled together in a nice way . this target - invariance feature was first observed in @xcite , and is closely related to the decompositions of sle@xmath297 into sle@xmath0 `` excursions '' mentioned in ( 1 ) . it says that for any @xmath35 and @xmath335 , the sle@xmath297 targeting @xmath35 and @xmath335 can be coupled in such a way that they coincide until the first time @xmath336 at which @xmath335 gets disconnected from @xmath35 for the chain targeting @xmath35 and evolve independently after that time . note that in this coupling , the natural time - parametrizations of these two processes do not coincide . however , using the previous observation about the relation between the excursions of @xmath17 and the intervals during which the sle@xmath297 traces a simple curve , we see that up to time @xmath336 the excursion intervals away from @xmath316 by the brownian motions used to generate the sle@xmath296 processes targeting @xmath35 and @xmath335 correspond to each other . the time - change between the two brownian motions can be calculated explicitly . for instance , if we define @xmath337 to be the process obtained by time - changing the brownian motion used in the construction that targets @xmath35 via the log - conformal radius of @xmath308 seen from @xmath58 ( instead of @xmath35 ) , then as long as @xmath338 , @xmath339 where by @xmath340 we denote the poisson kernel in @xmath95 seen from @xmath35 and targeted at @xmath300 ( it follows from the hadamard formula given below ) . the target - invariance property of sle@xmath296 leads to the construction of the radial sle@xmath297 branching - tree . we next summarize its definition and its main properties : * for a given countable dense collection @xmath341 of points in @xmath95 , it is possible to define on the same probability space a collection of radial sle@xmath297 processes in such a way that for any @xmath35 and @xmath335 , they coincide ( modulo time - change ) up to the first time at which @xmath35 and @xmath335 get disconnected from each other , and behave independently thereafter . * for any @xmath35 consider the first time @xmath342 at which the underlying driving function driving function exits the interval @xmath343 and define @xmath344 to be the complement of @xmath345 at that time containing @xmath35 . define @xmath346 similarly . the previous property shows that @xmath347 if and only if @xmath348 . * the cle@xmath0 carpet is then defined to be the complement of @xmath349 . in the sequel , we just write cle@xmath0 to denote this set ( i.e. we call cles to be the random fractal sets , not the collection of loops ) . notice that from the construction , it is not obvious that the law of the obtained cle@xmath0 ( and its nested versions ) does not depend on the choice of the starting point ( here @xmath334 ) on @xmath350 ( ie . of the root of the exploration tree ) . however , propositions [ cledesc ] and [ cledesc2 ] provide one possible proof of the fact that this choice of starting point does not matter ( this is also explained in @xcite , using the loop - soup construction of @xcite ) . this justify a posteriori that when one iterates the cle@xmath0 in order to construct the nested versions , one does in fact not need to specify where to continue the exploration . on the other hand , from this construction it is easy to estimate the expected area of the @xmath132-neighborhood of cle@xmath0 and to see that the upper minkowski dimension of cle@xmath0 is almost surely not greater than @xmath351 . this follow from arguments in @xcite : one wants to estimate the probability that @xmath352 . by conformal invariance it is enough to treat the case where @xmath35 is the origin in the unit dicc . but we can note that the conformal radius of @xmath308 ( in the sle@xmath297 ) is comparable to @xmath353 by koebe s @xmath354-theorem , and that the log - conformal radius of @xmath355 from the origin is just minus the exit time of @xmath343 by the brownian motion @xmath17 which is a well - understood random variable ( see @xcite for instance ) . hence , the asymptotic behavior as @xmath356 of the probability that @xmath357 can be estimated precisely . one can also prove the somewhat stronger statement that the hausdorff dimension cle@xmath0 is in fact equal to @xmath29 , by using second moment bounds ( ie . bounds on the probability that two points @xmath35 and @xmath335 are close to cle@xmath0 ) , see @xcite . the nested cle@xmath74 and cle@xmath72 can then be defined by appropriately iterating independent cle@xmath0 carpets in the respective domains @xmath344 ( starting the explorations in the nested domains at the point where one did just close the loops ) . the definition of cle@xmath72 is almost word for word the same , just replacing @xmath358 by @xmath359 which is the first time at which the underlying brownian motion @xmath17 exits the interval @xmath360 . a similar argument then shows that the upper minkowski dimension of cle@xmath72 is not greater than @xmath361 ( the @xmath362 term is then just due to brownian scaling the exit time of @xmath363 by brownian motion is equal in distribution to @xmath362 times the exit time of @xmath343 ) . to explain the coupling cle@xmath0 with the gff , we can first describe the coupling of a single radial sle@xmath296 process with a gff . the whole coupling then follows iteratively from the strong markov property and from the branching tree procedure . the proof of the coupling of the radial sle@xmath297 follows the steps of the coupling of the usual sle@xmath0 with the gff , as explained for instance in @xcite . it is based on the following observations : 1 . the hadamard formula ( see for instance @xcite ) gives the time - evolution of the green s function under loewner flow : for any two points @xmath274 and @xmath364 in @xmath95 , the green s function @xmath365 evolves until @xmath366 according to @xmath367 where by @xmath340 we denote the poisson kernel in @xmath95 seen from @xmath35 and targeted at @xmath300 as before . the cross - variation @xmath368 between the two local martingales @xmath321 and @xmath369 is equal to @xmath370 , so that @xmath371 3 . we can interpret @xmath372 ( up to @xmath373 ) as the harmonic measure in @xmath308 seen from @xmath35 , of the boundary arc between the tip of the curve and the force point @xmath310 ( the sign of @xmath337 describes which of the two arcs one considers ) . one can the first couple the gff with the radial sle@xmath297 up to the first time at which it surrounds the origin , exactly following the arguments of @xcite . this defines a btls with harmonic function in @xmath374 . for those domains where the harmonic functions are zero , one can then continue the sle@xmath297 iteration targeting another well - chosen point in that domain and proceed . this radial construction is arguably also the easiest one to explain that in fact , conditionally on the cle@xmath0 , the labels @xmath26 are i.i.d . this is then just a consequence of the time - reversal of sle@xmath0 , so that changing the orientation of the quasi - loop that traces the boundary of @xmath25 corresponds to a measure - preserving transformation of the driving brownian motion ( see for instance @xcite ) . we come back to this in paragraph [ commenting ] . we now derive proposition [ maxbtls ] , proposition [ cledesc ] and proposition [ propnodisconnection ] . in some sense , this section is the core of the present paper . here @xmath2 always denote a simply connected planar domain with non - empty boundary ( so that the previous definition of cle@xmath0 makes sense ) . let us fix a point @xmath35 , and consider the radial sle@xmath297 process targeted at @xmath35 that is coupled with the field ( and with the cle@xmath375 ) as described before . if @xmath376 , then at some rational time ( for the radial sle s parametrization ) , this radial sle is drawing some piece of @xmath377 within @xmath344 . we therefore consider a fixed rational time @xmath378 , and consider the event @xmath379 that the radial sle@xmath297 process targeting @xmath35 , stopped at @xmath380 , has the following properties : * it is tracing a loop . * its tip is in @xmath344 . * its boundary values are either @xmath68 or @xmath381 let @xmath382 denote the connected component of the complement of @xmath383 that has @xmath384 on its boundary . then , conditionally on @xmath379 , @xmath1 and @xmath385 , the evolution of @xmath185 after @xmath378 is that of a generalized level line in @xmath382 that traces a @xmath68 versus @xmath386 interface for the gff with boundary values depicted in figure [ fig : test ] , until the first time at which this level - line hits the boundary of @xmath382 again . from lemma [ notouch ] , it then follows that it can not exit @xmath382 through some point of @xmath387 because of the boundary values of the field there . hence , the domain that this generalized level line after closing itself is almost surely included inside @xmath344 . in particular , if it happens that this domain contains @xmath35 , then the connected component of the complement of @xmath388 that contains @xmath35 has @xmath389 as its outer boundary , and maybe some inner boundaries ( that are part of @xmath390 this can happen only if @xmath344 is not simply connected ) . importantly , the intersection of @xmath389 and @xmath390 is at most two points . as the maximum value of the @xmath391-circle average of the gff is of the order @xmath392 , these two intersection points can not influence the value of the harmonic function ( see @xcite for the precise statement and proof ) and thus , almost surely , the absolute value of the harmonic function corresponding to @xmath388 is bounded by @xmath393 . a similar analysis can be applied to the case where @xmath394 . note that if for all rational @xmath378 , the stopped radial sle@xmath297 is either not in @xmath344 or not in the process of tracing the loop that constitute the outer boundary of @xmath395 , then it means that @xmath396 , so that in this case too , the harmonic function corresponding to @xmath397 is necessarily bounded , in absolute value , by @xmath393 . we can therefore conclude that @xmath388 is also a @xmath393-btls . but by lemma [ mes ] ( recall that cle@xmath375 satisfies @xmath131 ) , this implies that @xmath398 and proves proposition [ maxbtls ] . in plain ; in dashed , the past of the radial exploration within @xmath344 . one discovers the dash - dotted interface as part of a loop along the radial sle@xmath297 exploration : it can not hit the dashed part without completing a cle loop at that time or before that time because of the properties of sle@xmath297 . and it can not touch the parts of @xmath390 that are away from the dashed part before completing the loop , because the gff boundary values are too small ] the proof of proposition [ cledesc ] follows along the same lines as that of proposition [ maxbtls ] . the difference lies in the fact that this time we are exploring one @xmath72 using another @xmath72 and thus the boundary conditions on @xmath399 are exactly @xmath400 . as above , we conclude from lemma [ notouch ] that the part of the radial sle@xmath297 drawing @xmath401 versus @xmath402 level - line loop can not exit @xmath344 before completing that loop ( one has to modify figure [ fig : test ] , so that the dash - dotted interface @xmath185 is now a @xmath403 vs. @xmath404 interface , and the continuous boundary data is @xmath400 ) . hence , it first follows as before that @xmath405 has to be bounded by @xmath402 , and then using lemma [ mes ] that @xmath406 almost surely . but this means that for all @xmath35 , @xmath407 again , using lemma [ mes ] we see that @xmath408 almost surely . suppose now that @xmath1 is a @xmath7-btls such that with positive probability , there exists a connected component of @xmath1 that is disconnected from @xmath12 . we choose some @xmath73 such that @xmath409 . we have just shown that almost surely , @xmath406 . on the other hand , with positive probability , @xmath3 contains an annular open region that disconnects this connected component of @xmath1 from @xmath12 . hence , there exists a deterministic such annular region @xmath162 such that with positive probability , @xmath162 is in @xmath3 and disconnects a connected component of @xmath1 from @xmath12 ( we call this event @xmath410 ) . given @xmath1 and @xmath5 , the conditional distribution of @xmath4 is a gff . it follow from lemma [ abscnty ] that on the event @xmath411 , the conditional distribution of @xmath4 restricted to @xmath162 is mutually absolutely continuous with respect to the conditional distribution of @xmath54 itself restricted to @xmath162 . it is possible to show , using corollary [ change boundary2 ] and the fact that with positive probability the radial sle@xmath296 makes a loop inside @xmath162 , that with positive probability the @xmath75 does not intersect the interior part of the complement of @xmath162 . but this contradicts the fact that @xmath94 . in this section , we describe the class of btls such that the harmonic function can only take two prescribed possible values . in particular , we construct a collection of btls that are smaller than cle@xmath0 and determine for which values such btlss exist . some aspects of the following discussion are strongly related to the @xmath412 case of boundary conformal loop ensembles ( and their nesting ) introduced and studied in @xcite ( that was written up in parallel to the present paper ) for general @xmath413 . let us first describe in some detail one specific example . consider a gff in the unit disc and fix two boundary points , say @xmath414 and @xmath415 . consider the zero - level line of this gff . this is an sle@xmath416 from @xmath414 to @xmath415 that is coupled with the gff as a btls @xcite . it is known that this is a simple path @xmath185 from @xmath414 to @xmath415 in the closed disc , and it is easy to see that its minkowski dimension is almost surely smaller than some @xmath417 . moreover , it is also known that the sle@xmath416 is a deterministic function of the gff @xcite . the harmonic function @xmath418 associated to this curve is described as follows : @xmath419 is the union of countably many connected components @xmath420 , and a component lies either to the right or to the left of the curve ( if one views the curve as going from @xmath414 to @xmath415 ) . in each component @xmath421 , the harmonic function has boundary conditions @xmath58 on @xmath422 and on @xmath423 , the boundary condition is @xmath106 or @xmath175 depending on whether @xmath421 is on the left or on the right of @xmath185 . as inside each component @xmath421 have independent gff - s with boundary conditions , we can iterate : suppose for example that the @xmath421 lies to the right of @xmath185 , then , @xmath424 is divided into to arcs , one of which is an excursion of @xmath185 away from the @xmath12 , and the other one is a counter - clockwise arc from @xmath425 to @xmath426 of @xmath12 . we now define a sle@xmath427 process from @xmath425 to @xmath426 . this is again a simple curve , and can be coupled with the gff with boundary conditions @xmath5 in @xmath421 in such a way that the boundary conditions on its left are @xmath175 and on its right are @xmath106 . we proceed in a symmetric way in the connected components that lie above @xmath185 . in this way , we obtain a new btls ( by first sampling @xmath185 , and then one sle@xmath427 process in each @xmath421 ) , and harmonic functions defined via the boundary conditions as indicated in figure [ fig : firstiteration ] . the iteration then further proceeds by defining sle@xmath428 processes in each of the remaining connected components which have part of @xmath12 on their boundary . one then defines a second layer of loops and one proceeds iteratively . we then consider the closure @xmath429 of the union of all the traced level lines . clearly , after any finite number of iterations in the previous construction , one has a @xmath106-btls , and therefore a subset of the cle@xmath430 by our previous results . hence , @xmath1 is itself a subset of the cle@xmath430 and therefore a btls . it is also easy to see that a given point @xmath46 is almost surely contained in a loop cut out after finitely many iterations , so that the harmonic function associated to @xmath1 takes its values in @xmath431 . and its boundary conditions on the left . the sle@xmath432 in the second layer on the right picture that creates loops with @xmath433 boundary conditions.,title="fig : " ] and its boundary conditions on the left . the sle@xmath432 in the second layer on the right picture that creates loops with @xmath433 boundary conditions.,title="fig : " ] * as opposed to the cle@xmath0 , the set @xmath429 is fairly skinny . indeed , it is just made out of the union of all sle@xmath0 type of paths . for instance , each excursion of @xmath185 away from @xmath350 is on the boundary between two connected components of @xmath429 ( one @xmath191 loop to its right and one @xmath175 loop to its left ) . this indicates that the minkowski dimension of @xmath429 should be @xmath434 ( we will come back to this later ) . * the set @xmath435 is the only btls with boundary values in @xmath436 . in fact , suppose that @xmath437 is another btls coupled with the same gff as @xmath429 , such that @xmath438 takes its values in @xmath431 and such that conditionally on @xmath54 , @xmath429 and @xmath437 are independent . then , one can loosely follow the strategy of the proof of proposition [ cledesc ] to conclude that @xmath439 almost surely . more precisely , consider again some @xmath440 and denote by @xmath344 the component of @xmath35 in @xmath441 . now , notice that almost surely no level line in the construction of the component of @xmath35 in @xmath442 can make an excursion inside of @xmath344 : indeed , consider for contradiction the first level line in the construction that enters @xmath344 . then one the one hand this level line can not exit @xmath344 on the boundary of @xmath437 . on the other hand it is also almost surely a simple path . thus , it can not exit @xmath344 at all and we obtain a contradiction . as this holds for a countable dense family of @xmath35 , we can again conclude from lemma [ mes ] that @xmath429 is the only @xmath106-btls with these properties . in particular its law does not depend on the arbitrary choice of the points @xmath274 and @xmath364 in the previous layered construction . furthermore , we note that , similarly to the construction of cle@xmath72 , we can iterate the construction of the set @xmath429 to construct a btls with harmonic function that takes its values in @xmath443 : indeed , we pick a countable number of dense @xmath46 and iterate the construction in the component containing @xmath35 until the boundary values are in @xmath444 . at each step in the construction , one has a @xmath445btls , and therefore in the limit , one still has a btls contained in cle@xmath430 . but from proposition [ cledesc ] it now follows that the obtained set is exactly the cle@xmath0 . this therefore provides an alternative construction of cle@xmath0 ( and of its iterated nested versions cle@xmath74 and cle@xmath72 ) that builds only on the coupling of the chordal sle@xmath446 and sle@xmath432 process with the gff . notice that in this case the measurability of the cle@xmath0 just follows from that of the respective sle@xmath447 and sle@xmath432 processes plus general properties of the local set and level lines . as we point out in section [ sdim ] , there is also a direct way to see that this set @xmath429 and its iterates are thin ( without using the relation to cle@xmath72 as we just did ) . hence , we indeed obtain a stand - alone construction of cle@xmath0 and of its properties . interestingly , we can not show that this construction gives the same law as cle@xmath0 without using the coupling with the gff . we first describe for any @xmath448 with @xmath449 , the construction of a btls @xmath450 such that the corresponding harmonic function takes its values in @xmath451 . this generalizes our previous constructions of @xmath452 and cle@xmath72 , that corresponds to @xmath453 . we then prove that there are no such btls when @xmath454 ( unless @xmath455 or @xmath456 are equal to @xmath58 , in which case one can take the empty set ) , and discuss the labels assigning to each component of @xmath457 its expected value . * _ @xmath459 or @xmath460 : _ we set @xmath461 and the corresponding harmonic function takes the value @xmath58 everywhere . * _ @xmath462 and @xmath463 , where @xmath464 and @xmath465 are positive integers : _ we can just suitably iterate independent copies of the previous set @xmath466 . we choose a countable dense set of @xmath46 and iterate in the component containing @xmath35 as long as the boundary values are in @xmath467 . * _ @xmath468 : _ set @xmath469 and repeat exactly the same construction as above , except that one now traces @xmath470-level lines ie . @xmath471 vs @xmath472 interfaces iteratively instead of @xmath175 vs. @xmath106 interfaces . exactly the same construction and the same arguments lead to the construction of a btls @xmath473 such that the corresponding harmonic function takes its values in @xmath474 . these sets are called the boundary conformal loop ensembles for @xmath412 in @xcite . notice that in these sets each interior boundary arc is shared by two components of the complement . * _ @xmath475 where @xmath476 is an integer _ : define @xmath477 such that there exists two non - negative integers @xmath478 with @xmath479 and @xmath480 . starting from @xmath473 we now iterate copies of @xmath481 ( resp . of @xmath482 ) in the connected components of the complement of @xmath483 depending on the value of the harmonic function . * _ general case with @xmath484 _ : by symmetry , we can suppose that @xmath485 . define @xmath486 and @xmath487 such that @xmath488 and @xmath489 . note that @xmath490 and that @xmath491 . now consider a sample of @xmath492 . in the connected components where the harmonic function is @xmath456 , we stop . in those components where the harmonic function is @xmath493 , we iterate copies of @xmath494 where @xmath495 . in this way , we obtain on the one hand connected components where the harmonic function is @xmath496 ( and we stop ) , and on the other hand , components where the harmonic function is @xmath497 . in the latter , we iterate independent copies of @xmath498 , and we continue in this way . at each step of the iteration , one has a btls that is contained in a given cle@xmath72 for a fixed @xmath73 . if we continue the iteration around every point @xmath35 until the iteration stops , we define a btls @xmath450 such that the corresponding harmonic function takes values in @xmath499 . conformal invariance of the gff implies the conformal invariance of @xmath84 and thus we see that @xmath500 does not depend on @xmath46 . using the fact that @xmath501 , we therefore see that for all @xmath46 , @xmath502 as one might have expected . to see that @xmath84 are the unique btls @xmath17 with @xmath503 , one can adapt the arguments that we used in the last subsection in the case where @xmath504 . thus , in order to complete the proof of part ( 2 ) of proposition [ cledesc2 ] , it remains to explain the monotonicity statement : suppose @xmath505 \subset [ -a ' , b']$ ] and @xmath80 with @xmath484 . starting with @xmath450 , and then sampling independent copies @xmath506 in all the connected components of its complement where the gff - boundary values are @xmath455 and independent copies of @xmath507 in the others , we obtain a btls with boundary values in @xmath508 . by uniqueness it follows that the obtained set is indeed @xmath509 and by construction it contains @xmath84 . let us now explain the proof of proposition [ cledesc2]-(1 ) : first , one can discard the case where @xmath496 and @xmath456 have the same sign because the mean value of the field has to remain @xmath58 . when @xmath80 and @xmath511 , suppose for contradiction that @xmath84 exists . sample @xmath512 in those connected components of the complement of @xmath450 to see that @xmath513 where the harmonic function is equal to @xmath456 . in particular , note that a connected component of @xmath3 where the corresponding harmonic function @xmath514 is @xmath496 remains a connected component of @xmath515 . such a connected component has boundary value @xmath496 , but has no boundary arc that is shared with a component of @xmath515 where the harmonic function is @xmath516 . this leads to a contradiction with an observation made above : we know from the construction of @xmath517 that all interior boundary arcs are shared by two components of the complement . we conclude this subsection with a small discussion about signs of the harmonic functions assigned to each @xmath518 of cle@xmath0 . as mentioned in section [ cle4 ] , by combining the radial sle@xmath296 construction and the reversibility of sle@xmath0 one can show that it is possible to first sample the family @xmath518 , and then the labels @xmath26 using independent fair coins . the idea is that one can define the symmetric radial sle@xmath297 using a poisson point process of sle@xmath0 bubbles , which is invariant under resampling of the orientations of the bubbles , see e.g. @xcite ) . this conditional independence between the heights in different domains feature is specific to cle@xmath0 . for instance , for cle@xmath72 the existence of a correlation between the heights in different domains is clear from the construction . for @xmath429 , the situation is actually quite reversed : conditioned on @xmath429 , one fair coin toss that decides the sign of the harmonic function at the origin is enough to determine the harmonic function in all the other connected components of the complement of @xmath429 . this follows from the fact that any two neighbouring components have different heights . in the case of @xmath473 for @xmath519 , one does not even need to toss a coin , as the asymmetry ( and conformal invariance ) makes it in fact possible to detect almost surely the sign of the harmonic function at the origin . from this point of view , it is even quite intriguing that by iterating @xmath429 , one obtains a cle@xmath0 where the signs of the heights are independent in the different components . let us first focus on the case of the set @xmath525 . fix a point @xmath333 and note that the construction of the set @xmath1 is obtained via a continuously increasing family of sets @xmath526 that correspond to the concatenation of the various chordal sle@xmath527 and sle@xmath427 processes that one iterates . all these sets @xmath528 are clearly local sets , and the value of the corresponding harmonic function @xmath529 at @xmath35 is always in @xmath530 $ ] ( as its boundary values are in @xmath531 ) . furthermore , by definition it is a local martingale , and therefore a martingale . we know that it converges to either @xmath191 or @xmath532 as @xmath533 . on the other hand , we know ( see eg . @xcite ) that when @xmath534 is a continuously increasing family of local sets then @xmath535 evolves like a constant times brownian motion when parametrized by the decrease of the log - conformal radius of @xmath536 seen from @xmath35 . this therefore implies that @xmath537 is distributed like a constant times the exit time of @xmath343 by a one - dimensional brownian motion . exactly the same argument can be applied to all @xmath450 s that we have constructed one just needs to note that in our iterative procedure , all the iterations are independent and the harmonic functions @xmath188 always remain in @xmath538 $ ] . note that this argument can be used to see that @xmath435 is indeed a btls with upper minkowski dimension almost surely not larger than @xmath434 and that cle@xmath0 obtained by iterations of @xmath435 is indeed a btls ( as it satisfies @xmath539 ) . this work was supported by the snf grant # 155922 . the authors are part of the nccr swissmap . the authors would like to thank the isaac newton institute for mathematical sciences , cambridge , as well as the clay foundation , for hospitality and support during the program _ random geometry _ where a part of this work was undertaken .	we study a certain class of local sets of the two - dimensional gaussian free field ( gff ) in a simply - connected domain , and their relation to the conformal loop ensemble cle@xmath0 and its variants . more specifically , we consider bounded type thin local sets ( btls ) , where thin means that the local set is small in size , and bounded - type means that the harmonic function describing the mean value of the field away from the local set is bounded by some deterministic constant . we show that a local set is a btls if and only if it is contained in some nested version of the cle@xmath0 carpet , and prove that all btls are necessarily connected to the boundary of the domain . we also construct all possible btls for which the corresponding harmonic function takes only two prescribed values and show that all these sets are in fact measurable functions of the gff .
the study of brown dwarfs , stellar - like objects with a mass below that required for the onset of hydrogen fusion ( e.g. burrows et al . 2001 ) , has advanced dramatically in the last decade . primarily this has been due to the discovery of a significant population of brown dwarfs by far - red and near - infrared surveys : the sloan digital sky survey ( sdss , york et al . 2000 ) , the two micron all sky survey ( 2mass , skrutskie et al . 2006 ) , the uk infrared telescope ( ukirt ) infrared deep sky survey ( ukidss , lawrence et al . 2007 ) and the canada france hawaii telescope s brown dwarf survey ( cfbds ; delorme et al . the latest spectral type and coolest dwarfs currently known are classified as t dwarfs . at the time of writing there are eight objects known with type later than t8 : ulas j003402.77 - 005206.7 ( t9 , hereafter ulas 0034 - 00 , warren et al . 2007 ) , cfbds j005910.90 - 011401.3 ( t9 , delorme et al . 2008a ) , ulas j123828.51 + 095351.3 and ulas j133553.45 + 113005.2 ( t8.5 and t9 , burningham et al . 2008 ) , wolf 940 b ( t8.5 , burningham et al . 2009 , hereafter b09 ) , ulas j130217.21 + 130851.2 ( t8.5 , burningham et al . 2010 ) , ross 458 c ( t8.5 , goldman et al . 2010 ) and ugps j072227.51 - 054031.2 ( t10 , lucas et al . these extreme objects have very low effective temperatures ( @xmath3 ) of 500 600 k ( e.g. leggett et al . 2010 ) , approaching the temperature where water clouds are expected to form in the atmosphere ( e.g. burrows et al . 2003 ) . the t dwarfs that have stellar companions are of particular interest as the primary can provide distance , metallicity and age constraints for the brown dwarf . there are ten known examples of such systems at this time . in order of increasing t dwarf sub - class they are : gl 337 cd ( t0 , t0 @xmath19 k1v @xmath19 g9v ; wilson et al . 2001 , burgasser et al . 2005 ) ; epsilon indi bab ( t1 , t6 @xmath19 k5v ; scholz et al . 2003 , mccaughrean et al . 2004 ) ; hn peg b ( t2.5 @xmath19 g0v ; luhman et al . 2007 , leggett et al . 2008 ) ; scr 1845 - 6357 b ( t6 @xmath19 m8.5 ; biller et al . 2006 , kasper et al . 2007 ) ; gl 229 b ( t7p + m1v ; nakajima et al . 1995 ; leggett et al . 2002 ) ; hd 3651 b ( t7.5 @xmath19 k0v ; mugrauer et al . 2006 , burgasser 2007 , liu et al . 2007 , luhman et al . 2007 ) ; gl 570 d ( t7.5 @xmath19 m3v @xmath19 m1.5v @xmath19 k4v ; burgasser et al . 2000 , saumon et al . 2006 , geballe et al . 2009 ) ; wolf 940 b ( t8.5 @xmath19 m4v , b09 ) ; ross 458 c ( t8.5 @xmath19 m7v @xmath19 m0.5 ; goldman et al . the degree of accuracy that can be obtained in measurements of the fundamental parameters of such t dwarfs is illustrated by the geballe et al . ( 2009 ) study of the t7.5 dwarf gl 570d . the gl 570 system has a well determined distance and metallicity , and the k4v primary constrains the age to be between 2 and 5 gyr ( geballe et al . comparison of the observed luminosity and spectral properties of the t dwarf to atmospheric and evolutionary models constrains @xmath3 to be 800 820 k and the gravity to be @xmath20 5.23 . these values constrain the age further to 3 to 5 gyr , and the mass of the t dwarf to be between 38 and 47 m@xmath18 . this paper presents an analysis of the spectral energy distribution ( sed ) of the t8.5 dwarf wolf 940 b. the dwarf was discovered in the large area survey component of the ukidss , and identified as a companion to the m4v star wolf 940 ( also known as lhs 3708 and gj 1263 ) , by b09 . b09 present near - infrared spectroscopy and photometry , as well as @xmath21 photometry , for the t dwarf . they use these data to derive the luminosity of the dwarf , and hence @xmath3 and surface gravity . leggett et al . ( 2010 ) give @xmath1 infrared array camera ( irac , fazio et al . 2004 ) [ 3.6 ] , [ 4.5 ] , [ 5.8 ] and [ 8.0 ] photometry for the t dwarf and use the infrared colors to support and further constrain the atmospheric properties determined by b09 . here we report new mid - infrared spectroscopy for wolf 940 b , obtained with the @xmath1 infrared spectrograph ( irs , houck et al . the improved accuracy of the luminosity we derive , and the additional spectroscopy , enables us to refine the b09 results . wolf 940 b is currently the coolest companion to a star with both near- and mid - infrared photometric and spectroscopic data . a more complete energy distribution for such a cool object , which has other basic data provided by its companion ( distance , age and metallicity ) , provides a good test of the model atmospheres for these almost - planetary objects . in 2 we describe the wolf 940 system . the irs data are presented in 3 . in 4 we use atmospheric and evolutionary models to calculate an accurate luminosity for wolf 940 b , and constrain its fundamental properties . our conclusions are given in 5 . the wolf 940 binary consists of a widely separated m4 and t8.5 dwarf pair ( reid et al . 1995 , b09 ) . table 1 lists the astrometric and photometric properties of the system . the primary is a high proper - motion m4 dwarf , identified in historic proper motion surveys as wolf 940 ( and lhs 3708 , gj 1263 , amongst other identifiers ) . b09 use the colors of the primary and the @xmath8:@xmath9 relationships presented by bonfils et al . ( 2005 ) to derive a metallicity [ m / h]@xmath22 for the system . however leggett et al . ( 2010 ) show that the @xmath23 and @xmath24[4.5 ] colors of the secondary imply that the system has solar or slightly super - solar metallicity . here we use the johnson & apps ( 2008 ) significant revision of the bonfils et al . calibration , together with the parallax , @xmath25 and @xmath26 values given in table 1 , to derive a metallicity of [ m / h]@xmath7 for the system . we discuss metallicity further in 4.6 . the measured radial velocity , combined with the parallax and proper motion , give uvw space motions of @xmath27 , @xmath28 and @xmath29 kms@xmath14 ( b09 and references therein ) , typical of old disk objects . b09 match the combination of the kinematics and the bonfils et al . metallicity value to the 3 5 gyr - old disk population in the galaxy population synthesis model by robin et al . ( 2003 ) . in this paper we constrain the age of the system using the geneva - copenhagen survey of age , kinematics and metallicity of a large number of stars in the solar neighbourhood ( holmberg et al . 2009 ) . based on this work , the uvw velocities and revised metallicity of the wolf 940 system implies an age of between 3 gyr and 10 gyr . we adopt this age range , taking a more conservative approach than b09 . wolf 940 a shows the h@xmath17 line in absorption , with an equivalent width of @xmath30 , typical for an m4 dwarf ( gizis et al . 2002 , their figure 6 ) . the evolution of chromospheric activity in low mass stars is a complex process , and its relationship to the strength of the h@xmath17 line is not well understood ( e.g. gizis et al . 2002 , west et al . 2008 , walkowicz & hawley 2009 ) . while it is clear that h@xmath17 in emission implies a highly active and therefore young star , stars with h@xmath17 in absorption may have no to moderate activity . given that wolf 940 a shows the line in absorption with a very typical strength for its type , we assume that it is no longer active and is therefore older than 3 gyr , which is the lower limit on the activity lifetime for an m4 dwarf ( west et al . 2008 , their figure 10 ) . again we take a more conservative approach than b09 , and do not assume that the presence of h@xmath17 absorption implies that the star is still somewhat active , and so we do not put an upper limit on the age of the system based on chromospheric activity arguments . in summary , the kinematic and h@xmath17 properties of wolf 940 a imply that the system is between 3 and 10 gyr old . a rotational velocity for wolf 940 a would be useful as it would allow a gyrochronology age to be determined . the metallicity of the system is approximately solar . we obtained @xmath1 irs spectra of wolf 940 b via the director s discretionary time program 527 . the short - low module was used in first order which provides a spectrum covering the 7.5 14.5 @xmath0 m range , with a resolution of @xmath31 . a nearby star was used in the blue peak - up array to acquire the target in the 3@xmath327 slit . standard staring mode was used with a ramp duration of 60s and 84 cycles , for an observation of duration 3.6 hours ; two such observations were obtained . each cycle places the target at two slit positions , about one - third and two - thirds of the way along the slit . thus the total on - source exposure was 5.6 hours . both observations were carried out on the 16th december 2008 . we used the basic calibrated data ( bcd ) produced by the @xmath1 pipeline version s18.7.0 . this produces a two - dimensional long - slit spectrum at each of the nod positions , so for each of our two observations there are 84 files for each nod position . the bcd files have been processed by the pipeline , which includes ramp fitting , dark subtraction , droop correction , linearity correction , flat fielding , and wavelength calibration ( see section 5.1 of the irs instrument handbook ) . we treated the two observations as independent datasets . we used the iraf data reduction tool ( tody 1993 ) to subtract each nod pair , and combined the 84 subtracted images using sigma - clipping in the imcombine routine . some structure remained in the background of the combined images ; we removed this by subtracting a smoothed version of the image , using an @xmath33 box . this residual - structure subtracted image was then fed into the @xmath1 irs data reduction package spice . the source signal was low , and there were a large number of bad and hot pixels . spice takes as input : the image file , a bad pixel mask and an uncertainty image . we used the sigma image produced by the imcombine routine as the uncertainty mask , and a bad pixel mask included with the bcd files . spice traces the appropriate module s slit profile over the input image , allowing the user to set the location and the width of the extraction window . we extracted the positive and negative spectrum using a relatively small window of 2 pixels in order to avoid adding noise . the final step in the spice application converts the signal to a flux density . we combined the two negative and two positive extracted spectra manually . the signal to noise is quite low ( s / n @xmath3410 ) , and we visually identified bad data points by comparison of the four spectra . figure 1 shows the final spectrum , with the noise spectrum which is based on the deviation between the four individual spectra . we checked the flux calibration by synthesizing the irac [ 8.0 ] photometry from the spectrum , after extending it to the short end of the filter bandpass ( @xmath35 m ) using as a template the irs second order spectrum of a bright t8 dwarf ( this extension contributed 17% of the total flux through the filter ) . this check indicated that the flux calibration is good to 5% . our analysis is based on a method we developed that uses synthetic spectra to determine a bolometric correction that is consistent with the luminosity obtained from evolution models . because the effective temperatures and gravities obtained in this fashion are obtained from the _ integrated _ observed flux and not by fitting the observed spectrum , it is quite robust and is not sensitive to the known ( and unknown ) systematic biases that arise when fitting model spectra to the observed spectrum . the solution for the @xmath36 and gravity is further refined by direct comparison of the corresponding synthetic spectra with the spectroscopic and photometric data , as well as age constraints . in this section , we first obtain @xmath36 and the gravity using solar metallicity models . the mid - infrared photometry and spectrum allow an exploration of non - equilibrium chemistry driven by vertical transport and we estimate the mixing time scale . finally , we perform the analysis again with non - solar metallicity models and by considering the uncertainty in @xmath37 . our analysis of wolf 940 b is based on the atmospheric models and non - equilibrium chemistry scheme described in saumon et al . ( 2006 , 2007 ) and the evolution sequences of saumon & marley ( 2008 ) . we use cloudless atmospheres and evolution as appropriate for a late t dwarf , and initially assume solar metallicity . the observed 1.05 2.43 @xmath0 m spectrum ( b09 ) and 7.5 14.2 @xmath0 m spectrum ( this work ) are integrated to give an observed flux at earth of @xmath38 erg@xmath39s@xmath40cm@xmath16 . the irs spectrum contributes 39% of this value . following the method described in geballe et al . ( 2001 ) and saumon et al . ( 2006 , 2007 ) , we derive a family of @xmath42 values that provide @xmath37 values that are consistent with the bolometric correction from the model spectra and the evolution . that is , the @xmath42 of the spectral model used to calculate the correction from observed to total flux , must be consistent with the @xmath42 implied by the evolutionary models for that total luminosity . these @xmath42 solutions are shown in figure 2 and table 2 , and correspond approximately to a constant @xmath37 curve in the @xmath42 plane . the bolometric correction amounts to @xmath43% of the luminosity , and the derived luminosity is @xmath44 ( for @xmath45k , see below and table 2 ) . combining the uncertainty in the flux calibration of the spectrum ( using a monte carlo sampling of a gaussian distribution of the calibration uncertainty for each segment of the spectrum ) and in the distance , the uncertainty in @xmath37 is @xmath46dex . figure 2 shows evolutionary sequences in a @xmath3 against @xmath47 plot ; the allowed set of values for wolf 940 b obtained with the method outlined above is indicated by the sequence of solid dots . the @xmath36 is thus constrained to be in the range of 500 to 640@xmath39k . this range can be narrowed from a knowledge of the age of the system , or by comparison of the data to model spectra . the @xmath42 of the models used to generate the synthetic spectra for such a data comparison must fall between the dashed red lines of figure 2 , which corresponds to @xmath48 . figure 2 also demonstrates how each ( @xmath3 , @xmath49 ) pair determines the dwarf s mass and age . table 2 gives the set of allowed parameters ( @xmath3 , @xmath49 , @xmath50 , mass , radius and age ) in steps of @xmath51k for ages ranging from 0.4 to 10 gyr and for solar metallicity . the impact of varying metallicity is discussed below in 4.6 . having constrained the allowed @xmath42 parameter range , we compare the corresponding synthetic spectra with the data . figure 3 shows such a comparison for the solutions with ages of 1 , 5.5 and 10 gyr , i.e. the @xmath52 , @xmath53 solutions ( 575 , 4.722 ) , ( 600 , 4.989 ) and ( 625 , 5.221 ) ( see table 2 ) . note that the scaling of the model fluxes to the data is not adjustable : it is fixed by the measured distance to the wolf 940 system and the radius given in table 2 , determined by evolutionary models from the values of ( @xmath3 , @xmath54 ) . here we have assumed solar metallicity and an eddy diffusion coefficient @xmath55@xmath13s@xmath14 . the eddy diffusion coefficient parameterizes the time scale of vertical mixing in the upper radiative region of the atmosphere , that drives the chemistry of carbon and nitrogen out of equilibrium ( saumon et al . 2006 , 2007 ; hubeny & burrows 2007 ) . in this analysis , we have considered @xmath56 ( chemical equilibrium ) , @xmath57 , @xmath11 and @xmath12@xmath13s@xmath14 . t dwarfs typically have values of @xmath10 between @xmath11 and @xmath12@xmath13s@xmath14 ( e.g. saumon et al . 2007 , leggett et al . 2009 , stephens et al . 2009 ) ; the effect of varying @xmath10 is discussed further below . in the deeper convection zone , the mixing time scale is determined from the mixing length formulation of convection . as shown by figure 3 , and as we have found for other late t dwarfs , the near - infrared spectra of models constrained with our method are very nearly identical ( saumon et al . 2007 , geballe et al . 2009 ) . since these models have very nearly the same @xmath37 , it is a fairly accurate statement that the near infrared spectra of low@xmath58 models are independent of gravity _ at fixed _ @xmath37 , for a given metallicity . this is because the effect of increasing the gravity is partially canceled by the accompanying increase in effective temperature . also note that the effect of vertical mixing is not significant in the near - infrared at these @xmath3 ( e.g. saumon et al . 2006 , and 4.4 below ) . the match to the near - infrared spectrum is poor , with the modelled @xmath59 and @xmath60 peaks high and the @xmath26 low ( figure 3 , top panel ) . this has been noticed before and is likely due to the known incompleteness of the molecular opacity line lists in this region ( e.g. saumon & marley 2008 ) . in contrast to the near - infrared region , differences are seen between the mid - infrared synthetic spectra generated by the approximately constant - luminosity models ( figure 3 , lower panel ) . calculation of the reduced @xmath61 between the models and the observed mid - infrared spectrum shows that the best match is the @xmath62k model ( @xmath63 ) , which is better than the 575@xmath39k ( @xmath64 ) and 625@xmath39k ( @xmath65 ) solutions . the @xmath66k solution is clearly excluded with @xmath67 . vertical transport in the convection zone drives the chemistry of nitrogen away from equilibrium and results in a typical depletion of nh@xmath68 of a factor of 8 10 ( e.g. saumon et al . this increases the overall flux level in the 9 15@xmath69 m region where absorption by nh@xmath68 dominates . a comparison of the observed spectrum and synthetic spectra computed with equilibrium chemistry gives large @xmath61 values from 6.9 ( 550@xmath39k solution ) to 11.7 ( 625@xmath39k solution ) . those higher @xmath61 values arise from the inability of the equilibrium models to match the overall flux level of the irs spectrum for @xmath70 m ( saumon et al . 2006 , see also further discussion below ) . we can conclude that there is convincing evidence for a depleted abundance of nh@xmath68 in the atmosphere of wolf 940 b , which is consistent with findings from all of the other irs spectra of late t dwarfs ( saumon et al . 2006 , 2007 ; burgasser et al . 2008 , leggett et al . 2009 ) . synthetic fluxes in the @xmath71 and irac band passes were obtained by integrating the model spectra over the filter band passes , and these are shown in figure 3 . the irac [ 3.6 ] photometry is not well matched by any of the models . this is a known model deficiency ; leggett et al . ( 2009 , 2010 ) show that the calculated [ 3.6 ] fluxes are systematically low by 20 50% . it is not clear if this is related to missing opacities in the near - infrared ( which may lead to too high near - infrared fluxes and too low mid - infrared fluxes ) , or if it is due to distortions of the pressure - temperature structure of our non - equilibrium models . the reduced @xmath61 between the synthetic and measured photometry in these five bands strongly favor the highest @xmath36 model at 625@xmath39k as well as a value of the eddy diffusion coefficient of @xmath72 . however , the @xmath61 is dominated by a large contribution from the irac [ 3.6 ] band pass ( figure 3 ) . if we ignore this flux measurement , then all three models with @xmath73k and @xmath74 match equally well and are far better than all other parameter combinations . for all four models from table 2 , the spectrum computed in chemical equilibrium ( @xmath75 ) is by far the worst . thus , the photometry strongly favors models that depart from chemical equilibrium . for a fixed metallicity , the irac [ 4.5 ] flux is a sensitive probe of non - equilibrium carbon chemistry as the filter bandpass covers the fundamental band of co centered at 4.65@xmath69 m . more vigorous vertical transport ( i.e. larger @xmath10 ) leads to an over - abundance of co in the upper atmosphere and a strong band where none should be detectable at such low @xmath36 ( noll et al . 1997 , golimowski et al . 2004 , patten et al . 2006 , geballe et al . 2009 ) . a low [ 4.5 ] flux is thus a hallmark of excess co and of a non - equilibrium co abundance . the observed [ 4.5 ] flux for wolf 940 b is well below what would be expected from chemical equilibrium and only non - equilibrium models can match its value . we find that @xmath76 gives an excellent match and figure 4 shows our best fitting non - equilibrium model spectrum , along with the corresponding equilibrium spectrum ( @xmath75 ) . a diffusion coefficient of @xmath77 @xmath13s@xmath14 corresponds to a vertical mixing time scale of @xmath78hours . note in figure 4 the large difference between the non - equilibrium and equilibrium models in [ 4.5 ] flux , and in the flux level in the 9 15@xmath69 m region where absorption by nh@xmath68 dominates . finally , we note that fitting @xmath79 independently from @xmath36 , gravity and @xmath37 affects @xmath37 by less than 0.016 dex , which is small compared to the contributions from the distance and flux calibration uncertainties of 0.047 dex . based on our solar metallicity model atmospheres , spectra and evolution , we conclude that wolf 940 b is best fitted with @xmath80k , @xmath81 and @xmath82 . if we apply the age constraint of 3 10@xmath39gyr determined by the primary , then the parameter range is further restricted to @xmath83k and @xmath84 . the derived parameters @xmath36 , @xmath54 , @xmath79 , mass , radius and age are summarized in table 3 . as shown in figure 2 , the values of @xmath36 and @xmath54 are correlated , thus wolf 940 b can not be both low-@xmath36 and high gravity . these ranges do not reflect the effect of the uncertainty in the determination of @xmath37 of 0.047 dex , which translates to @xmath85k in @xmath36 ( for a fixed gravity ) and @xmath86 dex in gravity ( for a fixed @xmath36 ) . while we have been able to obtain a reasonably good fit of all the data available ( parallax , spectra and photometry ) , systematic differences between the models and the near - infrared spectrum linger . in particular , we find that the models are much bluer in @xmath87 than the data ( figures 3 and 4 ) . the @xmath26 band modelled flux is about half of what is observed . a blue @xmath87 color in a late t dwarf is an indicator of a high - gravity or low - metallicity object ( e.g. knapp et al . 2004 ) . given all the constraints applied in our method of analysis , we can not change the gravity independently of @xmath36 ( except within the error bars ) and we have seen that this has no effect on the @xmath26 band flux ( figure 3 ) . we can , however , consider variations in metallicity and of the distance , within the range allowed by the parallax uncertainty . we have redone the complete analysis by assuming that the luminosity is increased by @xmath88 to @xmath89 ( see figure 2 ) and keeping [ m / h]@xmath900 . for consistency , this new luminosity implies that the distance is increased by @xmath91 to 13.58@xmath39pc . this moves our best fitting model of @xmath62k and @xmath92 to @xmath93k and @xmath94 . this worsens the fit slightly , but well below the uncertainties and is thus not significant . the analysis was performed again with model atmospheres , spectra and evolution of different metallicities . we find that a variation of @xmath95=\pm0.3 $ ] results in @xmath96 , which is half of the uncertainty on @xmath37 due to the photometric and parallax uncertainties . since the uncertainties in @xmath37 are effectively negligible in terms of spectrum fitting , for clarity of presentation we will make a comparison of spectra of different metallicity calculated with the best fitting @xmath97 we obtained in the solar metallicity case . spectra with [ m / h]@xmath98 and @xmath99 are shown in figure 5 . the optimal value of @xmath79 depends on [ m / h ] . to better show the role of [ m / h ] in shaping the spectrum , we keep @xmath79 fixed at @xmath100@xmath13s@xmath14 for the non - solar metallicity models . as expected , increasing the metallicity raises the modelled @xmath26 band flux and a very good agreement is obtained for [ m / h]@xmath101 , but this comes at the cost of a significant increase in the modelled flux in the @xmath102 , @xmath59 and @xmath60 bands . on the other hand , the spectrum with [ m / h]@xmath98 is an excellent match to the observed @xmath103 peaks but increases the discrepancy at @xmath26 . calculation of the reduced @xmath61 confirms the qualitative visual impression from the top panel of figure 5 that the [ m / h]@xmath98 spectrum is a better fit to the near infrared spectrum of wolf 940 b. the lower panel shows the opposite situation at mid - infrared wavelengths : the fit of the spectrum improves steadily as the metallicity increases from @xmath104 to @xmath99 . the mid - infrared photometry also favors higher metallicity if we neglect the poorly fit 3 4 @xmath0 m region ( where there is little flux ) . hence varying the metallicity from [ m / h]@xmath98 to + 0.3 has a seesaw effect on the shape of the spectrum compared to the data and provides no real improvement over the solar metallicity fit . since the extremes of that range lead to notably poor fits in the near - infrared or the mid - infrared , we can reasonably conclude that wolf 940 b has a metallicity within @xmath40.2 dex of solar , which is consistent with the value of [ m / h]@xmath7 derived for the primary star using the johnson & apps ( 2008 ) @xmath8:@xmath9 relationship ( see 2 ) . figure 5 ( lower panel ) shows that the [ 4.5 ] flux is not only sensitive to @xmath79 , it is also sensitive to [ m / h ] . the best solar metallicity solution gives @xmath105 using this datapoint , as described above in 4.4 . if [ m / h]@xmath106 then @xmath107 reproduces the observed [ 4.5 ] flux , and similarly if [ m / h]@xmath98 then @xmath108 . values of @xmath109 4 , 6 and 7 correspond to mixing time scales of 2.8 days , 1.6 hours and 14 minutes respectively . the convective mixing time scale is 0.6 min for comparison , meaning that @xmath108 implies that mixing is almost as vigorous as convection which may not be realistic . a metallicity as low as [ m / h]@xmath98 appears unlikely based on the very high value of @xmath79 required , again consistent with the solar or slightly metal - rich solution found for the primary . we have presented a 7.5 14.2 @xmath0 m low - resolution spectrum of the t8.5 dwarf wolf 940 b , which is a companion to an m4 dwarf with a projected separation of 400 au . this spectrum complements the near - infrared spectrum and @xmath21 photometry presented by b09 , and the irac photometry presented by leggett et al . ( 2010 ) . combining all these data allows a rigorous analysis of the dwarf s sed . assuming an age range of 3 10 gyr as indicated by the primary , evolutionary and atmospheric models show that for the t dwarf secondary @xmath83k and @xmath84 . gravity and temperature are correlated such that the lower gravity corresponds to the lower @xmath36 and younger age for the system , and the higher value to the higher @xmath36 and older age . the mass of the t dwarf is 24 m@xmath18 to 45 m@xmath18 for the younger to older age limits . an age younger than 1 gyr is excluded by the analysis of the t dwarf spectrum alone . these temperatures and gravities are slightly higher than those derived in b09 . our luminosity , which benefits from the inclusion of the irs spectrum , is slightly ( but not significantly ) higher than determined by b09 . our higher temperature and gravity range principally arises from our more conservative approach to the age of the primary , for which we adopt 3 10 gyr as opposed to b09 s 3 5 gyr . the irs spectrum demonstrates that vertical mixing is important in the radiative zone of the atmosphere of wolf 940 b , as is the case for many , if not all , t dwarfs . the [ 4.5 ] photometric datapoint constrains @xmath110 to 5.5 for solar metallicity . if the metallicity is higher then @xmath10 is lower , and vice versa . comparison of the synthetic sed to the observational data shows that wolf 940 b has a metallicity within @xmath40.2 dex of solar . more extreme values give poor fits to the data lower metallicity produces a poor fit at @xmath5 m while higher metallicity produces a poor fit at @xmath6 m . this is consistent with the value of [ m / h]@xmath7 derived here for the primary star using the johnson & apps ( 2008 ) @xmath8:@xmath9 relationship . it is very useful to find t dwarf companions to main sequence stars , given that the brown dwarfs are cooling and hence mass is difficult to constrain without knowing age . we look forward to more examples like wolf 940 b ( b09 ) and ross 458 c ( goldman et al . 2010 ) being found , as the ukidss sky area increases . this work is based on observations made with the _ spitzer space telescope _ , which is operated by the jet propulsion laboratory , california institute of technology under a contract with nasa . support for this work was provided by nasa through an award issued by jpl / caltech . support for this work was also provided by the spitzer space telescope theoretical research program , through nasa . skl s research is supported by the gemini observatory , which is operated by the association of universities for research in astronomy , inc . , on behalf of the international gemini partnership of argentina , australia , brazil , canada , chile , the united kingdom , and the united states of america . biller , b. a. , kasper , m. , close , l. m. , brandner , w. & kellner , s. 2006 , , 641 , l141 bonfils , x. , delfosse , x. , udry , s. , santos , n. c. , forveille , t. & segransan , d. 2005 , , 442 , 635 burgasser , a. j. et al . 2000 , , 531 l57 burgasser , a. j. , kirkpatrick , j. d. & lowrance , p. j. 2005 , , 29 , 2849 burgasser , a. j. 2007 , , 658 , 617 burgasser , a. j. , tinney , c. g. , cushing , m. c. , saumon , d. , marley , m. s. , bennett , c. s. & kirkpatrick , j. d. 2008 , , 689 , l53 burningham b. et al . 2008 , , 391 , 320 burningham , b. , et al . 2009 , , 395 , 1237 burningham , b. , et al . 2010 , , in press burrows , a. , hubbard , w. b. , lunine , j. i. & liebert , j. 2001 , reviews of modern physics , 73 , 719 burrows , a. , sudarsky , d. & lunine , j. i. 2003 , , 596 , 587 cutri , r. m. et al . 2003 , the irsa 2mass all - sky point source catalog , nasa / ipac infrared science archive dawson , p. c. & de robertis , m. m. 2005 , , 117 , 1 delorme p. et al . 2008a , , 482 , 961 delorme p. et al . 2008b , , 484 , 469 geballe , t. r. , saumon , d. , leggett , s.k . , knapp , g.r . , marley , m.s . & lodders , k. 2001 , , 556 , 373 geballe , t. r. , saumon , d. , golimowski , d. a. , leggett , s. k. , marley , m. s. & noll , k. s. 2009 , , 695 , 844 fazio , g. et al . 2004 , , 154 , 10 gizis , j. e. , reid , i. n. & hawley , s. l. 2002 , , 123 , 3356 goldman , b. , marsat , s. , henning , t. , clemens , c. & greiner , j. 2010 , , in press golimowski , d. a et al . 2004 , , 127 , 3516 harrington , r. s. & dahn , c. c. 1980 , , 85 , 454 holmberg , j. , nordstrom , b. & andersen , j. 2009 , , 501 , 941 houck j. r. et al . 2004 , , 154 , 18 hubeny , i. & burrows , a. 2007 , , 669 , 1248 johnson j. a. & apps k. 2008 , , 699 , 933 kasper , m. , biller , b. a. , burrows , a. , brandner , w. , budaj , j. & close , l. m. 2007 , , 471 , 655 knapp , g. r. et al . 2004 , , 127 , 3553 lawrence a. et al . 2007 , , 379 , 1599 leggett , s. k. , hauschildt , p. h. , allard , f. , geballe , t. r. & baron , e. 2002 , , 332 , 78 leggett , s. k. et al . 2008 , , 682 , 1256 leggett , s. k. et al . 2009 , , 695 , 1517 leggett , s. k. et al . 2010 , , 710 , 1627 liu , m. c. , leggett , s. k. & chiu , k. 2007 , , 660 , 1507 lucas , p. w. et al . 2010 , , in press luhman , k. l. et al . 2007 , , 654 , 570 mccaughrean m. j. , close , l. m. , scholz , r .- d . , lenzen , r. , biller , b. , brandner , w. , hartung , m. & lodieu , n. 2004 , , 413 , 1029 mermilliod j. c. 1997 , vizier on - line data catalog : ii/168 . originally published in : institut dastronomie , universite de lausanne ( 1991 ) mugrauer , m. , seifahrt , a. , neuhauser , r. & mazeh , t. 2006 , , 373 , l31 nakajima , t. , oppenheimer , b. r. , kulkarni , s. r. , golimowski , d. a. , matthews , k. & durrance , s. t. 1995 , nature , 378 , 463 noll , k. s. , geballe , t. r. & marley , m. s. 1997 , , 489 , l87 patten , b. m. et al . 2006 , , 651 , 502 reid , i. n , . hawley , s. l. & gizis , j. e. 1995 , , 110 , 1838 robin a. c. , reyle c. , derriere s. & picaud s. 2003 , , 409 , 523 saumon , d. & marley , m. s. 2008 , , in press . saumon , d. , marley , m. s. , cushing , m.c . , leggett , s.k . , roellig , t.l . , lodders , k. & freedman , r.s . 2006 , , 647 , 552 saumon , d. , marley , m. s. , leggett , s.k . , geballe , t.r . , stephens , d. s. , golimowski , d. a. , cushing , m. c. , fan , x. , rayner , j. t. , lodders , k. & freedman , r. s. 2007 , , 665 , 1136 scholz r .- d . , mccaughrean , m. j. , lodieu , n. & kuhlbrodt , b. 2003 , , 398 , l29 skrutskie m. f. et al . 2006 , , 131 , 1163 smart , r. l. et al . 2010 , aap , 511 , 30 stephens , d. s. , leggett , s. k. , cushing , m. c. , saumon , d. , marley , m. s. , geballe t. r. , golimowski , d. a. & fan , x. 2009 , , 702 , 154 tody , d. 1993 , `` iraf in the nineties '' in astronomical data analysis software and systems ii , a.s.p . conference ser . , vol 52 , eds . hanisch , r.j.v . brissenden , & j. barnes , 173 . walkowicz , l. m. & hawley , s. l. 2009 , , 137 , 3297 west , a. a. et al . 2008 , , 135 , 785 wilson j. c. , kirkpatrick , j. d. , gizis , j. e. , skrutskie , m. f. , monet , d. g. & houck , j. r. 2001 , , 122 , 1989 york , d. g. et al . 2000 , , 120 , 1579 for wolf 940 b. brown dwarf cooling tracks ( solid black lines ) are labeled with the mass in m@xmath111 . isochrones ( blue dotted curves ) are labeled with the age in gyr . solid dots are solutions for wolf 940 b for @xmath112 to 625@xmath39k in steps of 25@xmath39k ( see table 2 ) . a constant luminosity curve ( @xmath113 ) is plotted to guide the eye ( solid red curve ) , as well as its uncertainty ( @xmath114 , dashed red curves ) . we estimate the age of the wolf 940 b system to be between 3 and 10@xmath39gyr . the spur on the 1@xmath39gyr isochrone is caused by deuterium burning . ] k ( red ) , 600@xmath39k ( blue ) and 625@xmath39k ( green ) . see figure 2 and table 2 . photometric fluxes for these models are shown by horizontal lines . all fluxes are for cloudless models with an eddy diffusion coefficient of @xmath74 ( see text ) and are absolute , based on the measured distance and the radii given in table 2 . the data are shown in black , with photometric measurements shown as boxes with a height of @xmath115 . for clarity , the near - infrared spectrum ( b09 , upper panel ) has been smoothed with a running 7-pixel average . the irs noise spectrum is shown in the lower panel . ] k , @xmath92 , @xmath76 , blue curve ) and the same spectrum computed with chemical equilibrium abundances ( @xmath75 , red curve ) . the large non - equilibrium abundance of co is responsible for the strong band centered at 4.66@xmath69 m . the reduced , non - equilibrium abundance of nh@xmath68 results in a higher flux for @xmath116 m . see figure 3 for additional details . ] around the parameters of best fitting spectrum shown in figure 4 . all spectra shown have @xmath62k and @xmath117 . the observed spectrum of wolf 940 b is shown in black . colored curves show spectra computed with the metallicity and eddy diffusion coefficients shown in the legend . the red and blue curves overlap everywhere except in the co band centered at 4.6@xmath69 m . the parameters of the model shown in blue are almost identical to those of the best fitting spectrum of figure 4 . ] lrr ra 2000 , hhmmss.ss & 21 46 40.47 & 21 46 38.41 + dec 2000 , ddmmss.s & -00 10 25.4 & -00 10 34.6 + parallax mas & [email protected] & + pm ra mas yr@xmath14 & 765@xmath1182 & + pm dec mas yr@xmath14 & -497@xmath1182 & + radial vel . kms@xmath14 & [email protected] & + binary separation au & 400@xmath11825 & + binary pa deg . & 253@xmath1181 & + @xmath25 & [email protected] & + @xmath102 mko & & [email protected] + @xmath59 2mass / mko & [email protected] & [email protected] + @xmath60 2mass / mko & [email protected] & [email protected] + @xmath119/@xmath26 2mass / mko & [email protected] & [email protected] + @xmath21 mko & & [email protected] + 3.6 irac & & [email protected] + 4.5 irac & & [email protected] + 5.8 irac & & [email protected] + 8.0 irac & & [email protected] +	we present 7.514.2 @xmath0 m low - resolution spectroscopy , obtained with the @xmath1 infrared spectrograph , of the t8.5 dwarf wolf 940 b , which is a companion to an m4 dwarf with a projected separation of 400 au . we combine these data with previously published near - infrared spectroscopy and mid - infrared photometry , to produce the spectral energy distribution for the very low - temperature t dwarf . we use atmospheric models to derive the bolometric correction and obtain a luminosity of @xmath2 ( the observed spectra make up 47% of the total flux ) . evolutionary models are used with the luminosity to constrain the values of effective temperature ( @xmath3 ) and surface gravity , and hence mass and age for the t dwarf . we ensure that the spectral models used to determine the bolometric correction have @xmath3 and gravity values consistent with the luminosity - implied values . we further restrict the allowed range of @xmath3 and gravity using age constraints implied by the m dwarf primary , and refine the physical properties of the t dwarf by comparison of the observed and modelled spectroscopy and photometry . this comparison indicates that wolf 940 b has a metallicity within @xmath40.2 dex of solar , as more extreme values give poor fits to the data lower metallicity produces a poor fit at @xmath5 m while higher metallicity produces a poor fit at @xmath6 m . this is consistent with the independently derived value of [ m / h]@xmath7 for the primary star , using the johnson & apps ( 2008 ) @xmath8:@xmath9 relationship . we find that the t dwarf atmosphere is undergoing vigorous mixing , with an eddy diffusion coefficient @xmath10 of @xmath11 to @xmath12@xmath13s@xmath14 . we derive an effective temperature of 585 k to 625 k , and surface gravity @xmath15 to 5.22 ( cm s@xmath16 ) , for an age range of 3 gyr to 10 gyr , as implied by the kinematic and h@xmath17 properties of the m dwarf primary . gravity and temperature are correlated such that the lower gravity corresponds to the lower temperature and younger age for the system , and the higher values to the higher temperature and older age . the mass of the t dwarf is 24 m@xmath18 to 45 m@xmath18 for the younger to older age limit .
the notion of a trapped surface was introduced by penrose @xcite . in @xmath0-dimensional spacetime the spacelike boundary , @xmath1 , of a @xmath2-dimensional spatial region is called a future trapped surface if gravity is so strong there that even the future and ` outwards ' directed normal null rays starting at @xmath1 are dragged back so much that their expansion is non - positive everywhere on @xmath1 . careful analysis justified that trapped surfaces necessarily occur whenever sufficient amount of energy is concentrated in a small spacetime region @xcite . intuitively a black hole region is considered to be a part of a spacetime from which nothing can escape . therefore a black hole region is supposed to be a future set comprised by events that individually belong to some future trapped surface . the boundary of such a black hole region , referred to usually as the `` apparent horizon '' , @xmath3 , is then supposed to be comprised by marginal future trapped surfaces . as one of the most important recent results in black hole physics the existence of an `` apparent horizon '' was proved in @xcite . more specifically , it was shown there that given a strictly stable _ marginally outer trapped surface _ ( mots ) @xmath4 in a spacetime with reference foliation @xmath5 , then , there exists an open _ tube _ , @xmath6 , foliated by marginally outer trapped surfaces , with @xmath7 , through @xmath8 . let us merely mention , without getting into details here , that the applied strict stability assumption is to exclude the appearance of future trapped surfaces in the complementer of a black hole region . hawking s black hole topology theorem @xcite is proven by demonstrating that whenever the dominant energy condition ( _ dec _ ) holds a mots @xmath1 can be deformed , along the family of null geodesics transverse to the apparent horizon , yielding thereby on contrary to the fact that @xmath1 is a mots a future trapped surface in the complementer of the black hole region , unless the euler characteristic @xmath9 of @xmath1 is positive . whenever @xmath1 is a codimension two surface in a @xmath0-dimensional spacetime the euler characteristic and the `` genus '' , @xmath10 , of @xmath1 can be given , in virtue of the gauss - bonnet theorem , via the integral of the scalar curvature @xmath11 of the metric @xmath12 induced on @xmath1 as @xmath13 the main difficulty in generalising hawking s argument to the higher dimensional case originates from the fact that whenever @xmath1 is of dimension @xmath14 in an @xmath15-dimensional spacetime , the integral of the scalar curvature @xmath16 by itself is not informative , as opposed to the case of @xmath17 , therefore the notion of euler characteristic has to be replaced by the yamabe invariant . the latter is known to be a fundamental topological invariant and it is defined as follows . denote by @xmath18 $ ] the conformal class of riemannian metrics on @xmath1 determined by @xmath12 . it was conjectured by yamabe , and later proved by trudinger , aubin , and schoen that to every conformal class on any smooth compact manifold there exists a metric @xmath19 of constant scalar curvature so that @xmath20)\cdot \left(\int_{\mathscr{s}}\eeepsilon_{\tilde q}\right)^{-\frac{2}{s-2}}\,,\ ] ] where the yamabe constant @xmath21)$ ] , associated with the conformal class @xmath18 $ ] , is defined as @xmath22)= \inf_{\hat q\in[q]}\frac{\int_{\mathscr{s}}r_{\hat q}\,\eeepsilon_{\hat q } } { \left(\int_{\mathscr{s}}\eeepsilon_{\hat q}\right)^{\,\,\frac{s-2}{s } } } = \inf_{u\in c^{\infty}(\mathscr{s}),u>0}\frac{\int_{\mathscr{s}}\left [ 4\frac{s-1}{s-2}(d^au)(d_au)+r_qu^2 \right]\,\eeepsilon_{q } } { \left(\int_{\mathscr{s}}u^{\frac{2s}{s-2}}\eeepsilon_{q}\right)^{\,\,\frac{s-2}{s } } } \,.\ ] ] in the later case , the metric @xmath23 $ ] can be given as @xmath24 , and , moreover , @xmath25 and @xmath11 denote the covariant derivative operator and the scalar curvature associated with the metric @xmath12 on @xmath1 . the yamabe invariant is defined then as @xmath26}y(\mathscr{s},[q]).\ ] ] some of the recent generalisation of hawking s @xcite black hole topology theorem , and also that of gibbons @xcite and woolgar s @xcite results , proved by galloway , schoen , omurchadha and cai , that are covered by refs.@xcite may be formulated then as . [ gs ] let @xmath27 be a spacetime of dimension @xmath28 satisfying the einstein equations @xmath29 with cosmological constant @xmath30 , and with matter subject to _ dec_. suppose , furthermore , that @xmath1 is a strictly stable mots in a regular spacelike hypersurface @xmath31 . * if @xmath32 then @xmath1 is of positive yamabe type , i.e. , @xmath33 . * if @xmath34 and @xmath35 then for the `` area '' @xmath36 of @xmath1 the inequality @xmath37 holds . the significance of these results get to be transparent if one recalls that in the first case , i.e. , when @xmath33 , @xmath1 can not carry a metric of non - positive sectional curvature which immediately restricts the topological properties of @xmath1 @xcite . whereas , in the second case the a lower bound on the `` entropy '' of a black hole , that is considered to be proportional to the area @xmath38 , is provided by ( [ gib ] ) . before proceeding we would like to stress on an important conceptual point . most of the quoted investigations of black holes , see refs.@xcite , starts by assuming the existence of a reference foliation @xmath5 of the spacetime by ( partial ) cauchy surfaces @xmath39 . in this respect it is worth recalling that by a non - optimal choice of @xmath5 one might completely miss a black hole region as it follows from the results of @xcite where it was demonstrated that even in the extended schwarzschild spacetime one may find a sequence of cauchy surfaces which get arbitrarily close to the singularity such that neither of these cauchy surfaces contains a future trapped surface . hence , one of the motivations for the present work besides providing a reduction of the complexity of the proof of theorem[gs ] , and also a simultaneous widening of its range of applicability was to carry out a discussion without making use of any reference foliation . as it will be seen below the simplicity of the presented argument allows the investigation of black holes essentially in arbitrary metric theory of gravity . thereby , we do not restrict our considerations to either of the specific theories . accordingly , a spacetime is assumed to be represented by a pair @xmath27 , where @xmath40 is an @xmath15-dimensional ( @xmath41 ) , smooth , paracompact , connected , orientable manifold endowed with a smooth lorentzian metric @xmath42 of signature @xmath43 . it is assumed that @xmath27 is time orientable and that a time orientation has been chosen . the only restriction , concerning the geometry of the allowed spacetimes , is the following generalised version of _ dec_. a spacetime @xmath27 is said to satisfy the _ generalised dominant energy condition _ if there exists some smooth real function @xmath44 on @xmath40 such that for all future directed timelike vector @xmath45 the combination @xmath46 $ ] is a future directed timelike or null vector , where @xmath47 denotes the einstein tensor @xmath48 . it is straightforward to see that in einstein s theory of gravity , where @xmath42 is subject to ( [ ein ] ) , the generalised dominant energy condition , with the choice @xmath49 , is equivalent to requiring the energy - momentum tensor , @xmath50 , to satisfy _ dec _ @xcite . to restrict our considerations to black hole spacetimes , we shall also assume the existence of future trapped surfaces in @xmath27 which are defined as follows . let us consider a smooth orientable @xmath51-dimensional compact manifold @xmath1 with no boundary in @xmath40 . let @xmath52 and @xmath53 be smooth future directed null vector fields on @xmath1 scaled such that @xmath54 , and that are also normal to @xmath1 , i.e. , @xmath55 and @xmath56 for any vector field @xmath57 tangent to @xmath1 . consider then the null hypersurfaces generated by geodesics starting on @xmath1 with tangent @xmath52 and @xmath53 . these null hypersurfaces are smooth in a neighbourhood of @xmath1 , and , by making use of the associated synchronised affine parametrisations of their null generators , the vector fields @xmath52 and @xmath53 can be , respectively , extended to them . the level surfaces of the corresponding synchronised affine parametrisations do also provide foliations of these null hypersurfaces by @xmath51-dimensional compact manifolds homologous to @xmath1 . denote by @xmath58 the volume element associated with the metric , @xmath12 , induced on these @xmath51-dimensional surfaces . then the null expansions @xmath59 and @xmath60 with respect to @xmath52 and @xmath53 are defined by @xmath61 where @xmath62 and @xmath63 denotes the lie derivative with respect to the null vector fields @xmath52 and @xmath53 . the @xmath51-dimensional surface @xmath1 is called future _ trapped surface _ if both of the null expansions @xmath59 and @xmath60 are non - positive on @xmath1 . in the limiting case , i.e. , whenever either of these null expansions ( say @xmath59 ) vanishes on @xmath1 identically , @xmath1 is called future _ marginally trapped surface_. in case of a generic @xmath51-dimensional surface the quasi - local concept of outwards and inwards directions is undetermined . nevertheless , these concepts get to be well - defined for non - minimal marginally trapped surfaces . it will be said that @xmath52 and @xmath53 point _ outwards _ and _ inwards _ , respectively , provided that @xmath64 and @xmath65 , and that @xmath60 is not identically zero . ( if both @xmath59 and @xmath60 vanish identically @xmath1 is a minimal surface and the notions outwards and inwards become to be degenerate . ) to see that this , quasi - local , concept of `` outer '' direction is not counter intuitive , consider the null hypersurface @xmath66 generated by null geodesics starting at the points of @xmath1 with tangent @xmath67 ( see fig.[mots ] ) . since @xmath66 is smooth in a neighbourhood @xmath68 of @xmath1 it can be smoothly foliated by @xmath51-dimensional surfaces , @xmath69 , defined as the @xmath70 cross - sections of @xmath66 , where @xmath71 is the affine parameter along the generators of @xmath66 such that @xmath72 and @xmath73 on @xmath1 . then , it seems to be natural to consider @xmath53 as inward pointing if the `` area '' @xmath74 of the cross - sections @xmath69 is non - decreasing in the direction of @xmath75 which , in the case under consideration , follows from @xmath65 as @xmath76 accordingly , @xmath1 is called future _ marginally outer trapped surface _ ( mots ) if @xmath64 and @xmath65 on @xmath1 . in deriving our results we shall also apply a stability assumption . before formulating it let us recall first that the above imposed conditions do not uniquely determine the pair of null vector fields @xmath52 and @xmath53 on @xmath1 . in fact , together with @xmath52 and @xmath53 any pair of null vector fields @xmath77 and @xmath78 that is yielded by the boost transformation @xmath79 where @xmath80 is an arbitrary smooth function on @xmath1 , will be suitable . it is well - known , however , that the signs of @xmath59 and @xmath60 are invariant under such positive rescaling of @xmath52 and @xmath53 . suppose then that @xmath1 is a future mots with respect to a null normal @xmath52 . then , @xmath1 will be called _ strictly stably outermost _ if there exists a rescaling of the type ( [ boost ] ) so that @xmath81 for the yielded vector fields @xmath77 and @xmath78 , and also @xmath82 somewhere on @xmath1 . obviously , this definition is independent of the use of any sort of reference foliation . moreover , as it will be indicated below , it can be seen to be equivalent to the corresponding criteria applied in @xcite . the main argument of the present paper can then be given in the following simple geometric setup . we have already defined @xmath67 to be a smooth past directed null vector field on @xmath1 that is also normal to @xmath1 . similarly , the smooth null hypersurface @xmath66 , spanned by the @xmath51-parameter congruence of null geodesics starting at @xmath1 with tangent @xmath75 , and the affine parameter @xmath71 along the geodesics , synchronised such that @xmath73 on @xmath1 , have already been introduced . denote by @xmath75 the tangent field @xmath83 on the null hypersurface @xmath66 , that is foliated by the smooth @xmath70 cross - sections , @xmath69 . then , there exists a uniquely defined future directed null vector field @xmath52 on @xmath66 such that @xmath84 , and that @xmath52 is orthogonal to each @xmath69 . denote by @xmath85 the affine parameter of the null geodesics determined by @xmath52 which are synchronised such that @xmath86 on @xmath66 . since @xmath52 is , by construction , smooth on @xmath66 the null geodesics starting with tangent @xmath52 on @xmath66 do not meet within a sufficiently small open `` elementary spacetime neighbourhood '' @xmath68 of @xmath1 . extend , then , the function @xmath71 from @xmath66 onto @xmath68 by keeping its value to be constant along the geodesics with tangent @xmath52 . then the vector fields @xmath75 and @xmath52 , defined so far only on @xmath66 , do also extend onto @xmath68 such that the relations @xmath72 and @xmath87 hold there which do also imply that @xmath75 and @xmath52 commute on @xmath68 . note that @xmath68 is smoothly foliated by the @xmath88-parameter family of @xmath51-dimensional @xmath70 , @xmath89 level surfaces @xmath90 . the spacetime metric in @xmath68 can then always be given ( see , e.g. , @xcite for a justification ) as @xmath91 where @xmath92 , @xmath93 and @xmath94 are smooth fields on @xmath68 such that @xmath93 and @xmath94 are orthogonal to @xmath75 and @xmath52 . recall also that @xmath94 and the positive definite metric @xmath12 on the @xmath51-dimensional surfaces @xmath90 are related as @xmath95 this latter relation implies that @xmath96 on @xmath66 , represented by the @xmath86 hypersurface in @xmath68 , i.e. , @xmath94 is the metric on the cross - sections @xmath97 of @xmath66 . since the vector fields @xmath75 and @xmath52 are null and normal to the cross - sections @xmath97 the expansions of the associated null congruences at @xmath97 can be given as @xmath98 where @xmath99 and @xmath62 denote the lie derivative with respect to the vector fields @xmath75 and @xmath52 , respectively , and ( here and elsewhere ) all the indices are raised and lowered with the spacetime metric @xmath42 . the second equalities in ( [ exp ] ) follow from the fact that @xmath93 and @xmath94 are orthogonal to @xmath75 and @xmath52 and also that @xmath75 and @xmath52 commute in @xmath68 . by making use of ( [ metric ] ) , and the definition of the einstein tensor , it is straightforward to see that @xmath100= -\frac12 \gamma^{ef}r_{ef}\ ] ] holds on @xmath66 . then , in virtue of equation @xmath101 of @xcite , and by the coincidence of @xmath12 and @xmath94 , and also of the projectors @xmath102 , @xmath103 and @xmath104 on @xmath66 ( for the definition of @xmath104 see @xmath105 of @xcite ) we also have that on @xmath66 @xmath106\ , , \end{aligned}\ ] ] where @xmath25 and @xmath11 denote the covariant derivative operator and the scalar curvature associated with the metric @xmath96 on the @xmath51-dimensional surfaces @xmath97 on @xmath66 . by making use of the fact that the vector fields @xmath75 and @xmath52 commute in @xmath68 and that they are orthogonal to @xmath94 a direct calculation justifies then the relation @xmath107 since @xmath75 and @xmath52 commute we also have that @xmath108 in @xmath68 , where the expansion @xmath109 is defined with respect to the volume element @xmath110 associated with the metric @xmath12 on the surfaces @xmath90 as in ( [ exp ] ) . similarly , it can be verified that @xmath111 on @xmath66 , i.e. , whenever @xmath86 , where @xmath53 denotes the unique future directed null extension @xmath112\ ] ] of @xmath113 on @xmath1 onto @xmath68 which is normal to the surfaces @xmath90 and is scaled such that @xmath54 in @xmath68 . then , by making use of the vanishing of @xmath59 on @xmath1 , the above relations in particular , equations ( [ exp ] ) , ( [ gdec1 ] ) , ( [ sgdec1 ] ) and ( [ n])imply that @xmath114\,.\ ] ] since @xmath115 and @xmath52 are both future directed null vector fields on @xmath66 , and also the generalised dominant energy condition holds , i.e. , there exists a real function @xmath44 on @xmath40 such that the vector field @xmath116 $ ] is future directed and causal , the inequality @xmath117 holds on @xmath66 . finally , since @xmath1 was assumed to be a strictly stable mots , in virtue of ( [ tk ] ) , the null normals @xmath67 and @xmath52 may be assumed , without loss of generality , to be such that @xmath118 , and also that @xmath119 somewhere on @xmath1 . to see that the stability condition applied here is equivalent to the one used in @xcite note that @xmath120 and it transforms under the rescaling ( [ boost ] ) of the vector fields @xmath113 and @xmath52 on @xmath1 as @xmath121 . by making use of the notation @xmath122 and @xmath123 , it can be verified then that @xmath124\,\ ] ] holds , which is exactly the expression ` @xmath125 ' given in lemma3.1 of @xcite whenever @xmath1 is a mots and the variation vector field @xmath126 is chosen to be ` @xmath127 ' . this justifies then that the strict stability conditions applied here and in @xcite ( see , e.g. , definition 5.1 and the discussion at the end of section 5 of @xcite for more details ) are equivalent . in returning to the main stream of our argument note that whenever @xmath1 is a strictly stable mots and the generalised _ dec _ holds then , in virtue of ( [ gdec2 ] ) , @xmath128 so that the inequality is strict somewhere on @xmath1 . since @xmath12 is positive definite we also have that for any smooth function @xmath71 on @xmath1 @xmath129 thus , multiplying ( [ gdec3 ] ) by @xmath130 , where @xmath131 is arbitrary , we get , in virtue of ( [ gdec4 ] ) , that @xmath132 so that the inequality is strict somewhere on @xmath1 . to get the analogue of the first part of theorem[gs ] assume now that @xmath44 is such that @xmath133 throughout @xmath1 . then , by taking into account the inequality @xmath134 , which holds for any value of @xmath135 , we get from ( [ gdec5 ] ) that @xmath136\,\eeepsilon_{q } } { \left(\int_{\mathscr{s}}u^{\frac{2s}{s-2}}\eeepsilon_{q}\right)^{\,\,\frac{s-2}{s } } } > 0\,,\ ] ] for any smooth @xmath131 , i.e. , @xmath21)>0 $ ] , which implies that @xmath1 is of positive yamabe type . similarly , whenever the minimal value @xmath137 of @xmath44 on @xmath138 is negative , on one hand , @xmath139 to the functions @xmath140 and @xmath141 with @xmath142 , @xmath143 , we get that @xmath144^{\,\,\frac{2}{s}}\,.\ ] ] the combination of ( [ c1 ] ) and ( [ c2 ] ) , along with ( [ gdec5 ] ) , justifies then that @xmath145\,\eeepsilon_{q } } { \left(\int_{\mathscr{s}}u^{\frac{2s}{s-2}}\eeepsilon_{q}\right)^{\,\,\frac{s-2}{s } } } \geq -2\|f^{_{\mathscr{s}}}_{min}\|\left [ { \mathcal{a}}({\mathscr{s } } ) \right]^{\,\,\frac{2}{s } } \,.\ ] ] assuming finally that @xmath146 we get that , for any conformal class [ q ] , the yamabe constant @xmath21)\leq { \mathcal{y}}(\mathscr{s})<0 $ ] . this , along with ( [ c3 ] ) , implies then @xmath147)\|= -y(\mathscr{s},[q ] ) \leq 2 \|f^{_{\mathscr{s}}}_{min}\| \left [ { \mathcal{a}}({\mathscr{s } } ) \right]^{\,\,\frac{2}{s}}\,,\ ] ] which leads to the variant of the inequality ( [ gib ] ) yielded by the replacement of @xmath30 by @xmath137 . what has been proven in the previous section can be summarised as . [ gsn ] let @xmath27 be a spacetime of dimension @xmath28 in a metric theory of gravity . assume that the generalised dominant energy condition , with smooth real function @xmath148 , holds and that @xmath1 is a strictly stable mots in @xmath27 . we would like to mention that the argument of the previous section does also provide an immediate reduction of the complexity of the original proof of hawking and that of gibbons and woolgar . to see this recall that , in virtue of ( [ tipp2e ] ) , ( [ y1 ] ) and ( [ y2 ] ) , @xmath151 whenever @xmath1 is of dimension @xmath152 , and also that @xmath49 , if attention is restricted to einstein s theory with matter satisfying the dominant energy condition . thereby , as an immediate consequence of theorem[gsn ] , we have that @xmath153 whenever @xmath32 on @xmath1 , while @xmath154 whenever both @xmath9 and @xmath30 are negative . clearly the above justification of theorem[gsn ] is free of the use of any particular reference foliation of the spacetime . note also that in the topological characterisation of an @xmath51-dimensional strictly stable mots @xmath1 only the quasi - local properties of the real function @xmath148 are important . in particular , as the conditions of theorem[gsn ] do merely refer to the behaviour of @xmath44 on @xmath1 it need not to be bounded or to have a characteristic sign throughout @xmath40 . similarly , it would suffice to require the generalised dominant energy condition to be satisfied only on @xmath1 . finally , we would also like to emphasise that theorem[gsn ] provides a considerable widening of the range of applicability of the generalisation of hawking s black hole topology theorem , and also that of the results of gibbons and woolgar . as its conditions indicate , theorem[gsn ] applies to any metric theory of gravity and the only restriction concerning the spacetime metric is manifested by the generalised dominant energy condition and by the assumption requiring the existence of a strictly stable mots . accordingly , theorem[gsn ] may immediately be applied in string theory or in various other higher dimensional generalisations of general relativity .	a key result in four dimensional black hole physics , since the early 1970s , is hawking s topology theorem asserting that the cross - sections of an `` apparent horizon '' , separating the black hole region from the rest of the spacetime , are topologically two - spheres . later , during the 1990s , by applying a variant of hawking s argument , gibbons and woolgar could also show the existence of a genus dependent lower bound for the entropy of topological black holes with negative cosmological constant . recently hawking s black hole topology theorem , along with the results of gibbons and woolgar , has been generalised to the case of black holes in higher dimensions . our aim here is to give a simple self - contained proof of these generalisations which also makes their range of applicability transparent . [ section ] [ section ] [ section ] [ section ] [ section ] [ section ] [ section ] [ section ] [ section ] [ section ]
a theoretical understanding of hadronic @xmath1 decays is necessary to extract ckm matrix elements and cp - violating weak phases from data taken at the @xmath1 factories . the formalism of qcd factorization developed in @xcite has shown that two - body hadronic @xmath1 decays simplify considerably in the heavy - quark limit . in @xmath3 decays , for example , qcd factorization states that the matrix elements of the local operators @xmath4 appearing in the effective weak hamiltonian obey the factorization formula @xmath5 where @xmath6 are leading - twist light - cone distribution amplitudes , and the @xmath7-products imply integration over the light - cone momentum fractions of the constituent quarks inside the mesons . the hard - scattering kernels @xmath8 are perturbatively calculable quantities . the factorization formula is true to all orders in perturbation theory , but it receives power corrections in @xmath9 . if qcd factorization is to be used as a tool for extracting the flavor parameters of the standard model , it is important to show that these power corrections are not anomalously large . some sources of power corrections were identified and estimated in @xcite . in @xcite , the renormalon calculus was used to study power corrections due to soft gluon exchange in the decays @xmath10 , where @xmath11 is a light meson . here we generalize and extend these analyses to focus on soft gluon corrections to the amplitudes for @xmath1 decays into two light pseudoscalar mesons . the model of power corrections that we develop relies on the analysis of infrared ( ir ) renormalon singularities , which arise because the low - momentum regions in loop diagrams give increasingly important contributions at higher orders in perturbation theory @xcite , rendering perturbation theory a divergent asymptotic expansion . ir renormalons are most transparently studied through singularities in the borel transforms of perturbation series with respect to the coupling constant @xcite . the borel transform @xmath12 of the perturbation series for a physical quantity has poles ( or branch cuts ) at half - integer values of the borel variable @xmath13 , which lead to ambiguities of order @xmath14 when performing the inverse transform , a process to which we will refer as borel resummation . ( here @xmath15 is the large momentum scale in the process . ) since the final answer for a physical quantity must be unambiguous , these renormalon ambiguities should be combined with power corrections of the same order in @xmath16 . therefore , an analysis of ir renormalons may reveal non - trivial information about power corrections to various processes , which is particularly useful in cases where such corrections can not be analyzed using an operator product expansion ( see @xcite for early applications of this idea ) . we calculate the borel transforms of the hard - scattering kernels appearing in the qcd factorization formula using the so - called `` large-@xmath17 limit '' , in which the momentum flow through gluon lines is traced by inserting fermion bubbles on the propagators , replacing @xmath18 at the end of the calculation @xcite . in this way , the diagrams are effectively computed using a momentum - dependent running coupling at the vertices . this procedure gives the dominant higher - order contributions to a perturbation series in the very formal limit where @xmath19 , in which case @xmath20 with fixed number of colors . although for real qcd @xmath21 is not a large parameter , the use of the large-@xmath17 limit is motivated by the empirical observation that perturbation series are often dominated by terms of the form @xmath22 . in this paper , we calculate both the wilson coefficients and the hard - scattering kernels entering the qcd factorization formula at next - to - leading order ( nlo ) in the large-@xmath17 limit . this is necessary for obtaining renormalization - group ( rg ) invariant expressions for the physical decay amplitudes . the expansion of the wilson coefficients in the large-@xmath17 limit is a novel feature of our work , which goes beyond the approach taken in @xcite . the numerical study in section [ numerical ] helps to elucidate the main results of our work . we find that the power corrections to qcd factorization due to soft - gluon exchange are small , typically of order @xmath23 . the only exception is a first - order power correction to the hard - spectator term , which is to be expected considering that the effective scale for gluon exchange with the spectator quark is of order @xmath24 @xcite . perhaps our most important result is that the strong - interaction phases of the decay amplitudes increase significantly when going beyond one - loop order in perturbation theory . this can potentially enhance the direct cp asymmetries in decays such as @xmath25 and @xmath3 with respect to the predictions found at nlo in rg - improved perturbation theory in @xcite . whether or not such increased asymmetries are realistic depends on whether there are significant multi - loop contributions not captured in the large-@xmath17 limit . we leave a more careful analysis of such contributions for future work . [ sec - wilson ] the qcd factorization formula provides model - independent expressions for hadronic weak decay amplitudes at leading power in the heavy - quark expansion . all non - trivial strong - interaction effects , which give corrections to the so - called `` naive '' factorization ( vacuum insertion ) approximation , are contained in rg - invariant parameters @xmath26 , which are products of wilson coefficients with convolutions of hard - scattering kernels and meson light - cone distribution amplitudes . in order to obtain rg - invariant results , it is crucial that the wilson coefficients and hard - scattering kernels be treated in the same perturbative approximation scheme . since our goal is to resum the perturbation series for the hard - scattering kernels in the large-@xmath17 limit , we must first compute the wilson coefficients appearing in the effective weak hamiltonian in the same limit . this is non - trivial , since the counting scheme underlying the large-@xmath17 limit is different from the usual counting of logarithms used in the calculation of the wilson coefficients . specifically , whereas @xmath27 is considered @xmath28 in the usual rg counting , this quantity is @xmath29 in the large-@xmath17 limit . on the contrary , whereas @xmath30 is usually treated as a small parameter , it is considered @xmath28 in the large-@xmath17 limit . consider , as an example , the rare hadronic decays @xmath25 , for which the effective weak hamiltonian ( neglecting small contributions from electroweak penguins ) is @xcite @xmath31 where @xmath32 are products of ckm matrix elements . the scale dependence of the wilson coefficients is governed by the evolution equation ( in compact matrix notation ) @xmath33\ , \vec c(m ) \,,\ ] ] where the evolution matrix is given in terms of an ordered matrix exponential involving the anomalous dimension matrix @xmath34 and the @xmath35-function . the initial values @xmath36 at a scale @xmath37 are obtained by matching the standard model onto a low - energy effective theory , in which the heavy fields of the electroweak gauge bosons and the top quark are integrated out . to evaluate ( [ evolution ] ) at subleading order in the large-@xmath17 limit , we expand the qcd @xmath35-function as @xmath38 , where the rescaled coupling @xmath39 is @xmath28 in the large-@xmath17 limit . next , we change variables from @xmath40 to @xmath41 in the integral in ( [ evolution ] ) , write the anomalous dimension matrix as a function of @xmath41 , and expand in powers of @xmath42 . taking into account that factors of @xmath43 must be replaced with @xmath44 , we find that in @xmath45 block form the resulting matrix is given by @xmath46 the justification for writing the leading term in this expansion as a constant @xmath47 matrix @xmath48 times the coupling constant @xmath41 is that this corner of the anomalous dimension matrix does not pick up an additional factor of @xmath43 when going from leading order ( lo ) to next - to - leading order in perturbation theory , so that all further perturbative contributions are absorbed in the matrix @xmath49 , the form of which is irrelevant to our discussion . a similar argument applies to the submatrix in the lower left corner , which at order @xmath42 is given by a constant @xmath50 matrix @xmath51 times @xmath41 . the matrix @xmath52 governing the evolution of the current current operators @xmath53 in the large-@xmath17 limit is a non - trivial function of the coupling , which has been calculated in @xcite . the result is @xmath54 where @xmath55 is the number of colors , and @xmath56 the matrices @xmath48 and @xmath51 can be derived from the expressions for the lo anomalous dimension matrices given in @xcite : @xmath57 taking into account the textures of these matrices as well as corresponding relations between the matching conditions for the various coefficients at the scale @xmath58 , we find that the solutions for the wilson coefficients can be expanded as @xmath59 where @xmath60 up to @xmath29 , the wilson coefficients @xmath61 are thus determined in terms of two functions , given by @xmath62 the coefficient of the chromo - magnetic operator @xmath63 will be needed only at lo , where it is scale independent , i.e. , @xmath64 note that , as a consequence of the fact that the matrix @xmath48 has eigenvalues 0 and @xmath65 , the scale dependence of the penguin coefficients @xmath66 governed by the function @xmath67 is extremely simple , and given by @xmath68 . we will see below that the appearance of a logarithm is indeed required to cancel the scale dependence of the hard - scattering kernels associated with penguin contractions of the operators @xmath4 . [ cols="<,^,^,^,^,^,^,^",options="header " , ] in regards to the numerical significance of higher - order perturbative corrections , the most striking effect is the large increase of the imaginary part of the coefficient @xmath69 as one goes to higher orders in perturbation theory . an analysis using momentum distribution functions shows that the imaginary part of the hard - scattering kernel contributing to @xmath69 is dominated by contributions from scales of order a gev , indicating that it can receive large higher - order corrections @xcite . the corresponding strong - interaction phase may therefore be larger than that obtained from the nlo analysis . in the strict large-@xmath17 limit ( shown by the numbers in brackets ) , the penguin coefficients @xmath70 and @xmath71 mirror the behavior of @xmath69 . on the other hand , when the physical value @xmath72 is used as in ( [ repl ] ) , the phases are reasonably close to those at nlo and do not receive large higher - order corrections . this shows that after the replacement in ( [ repl ] ) there is a strong cancellation between the different terms in ( [ sk23 ] ) , which is maintained at higher orders . while the replacement in ( [ repl ] ) was motivated on physical grounds , it is an open and important question whether such a delicate cancellation will persist once the full set of nnlo corrections is included in the theory of qcd factorization . this section addresses the effects of renormalon ambiguities and higher - order corrections on the phenomenology of @xmath1 decays into two light pseudoscalar mesons . the results of section [ numerical ] have shown that the most notable effects are the ambiguity in the hard - spectator term and an increase in the imaginary parts of the kernels at higher orders in perturbation theory . we limit our phenomenological analysis to quantities that highlight these effects in a transparent way . the @xmath73 decay amplitudes depend strongly on @xmath69 ( see ( [ bpipi ] ) ) and are therefore a good probe of the renormalon ambiguity in the hard - spectator term . using the data in table [ tab - numai ] , and taking @xmath74 and @xmath75 , we find for the cp - averaged branching ratio @xmath76 where the numbers in parentheses refer to one - loop order , two - loop order , and borel - resummation , respectively . the nlo result obtained using qcd factorization is 4.80 . ( the central value quoted in @xcite is about 10% larger due to the inclusion of electroweak penguin contributions . ) the magnitude of the renormalon ambiguity may be compared with the error estimate given in @xcite , where an uncertainty of @xmath77 was assigned to account for the effects of parameter variations and power corrections from higher - twist distribution amplitudes . since this is the mode that is the most vulnerable to the ambiguity in @xmath69 , we are led to the conclusion that no large power corrections from soft gluon exchange are present in @xmath78 decays . the cp asymmetries of certain decays can be very sensitive to the imaginary parts of the hard - scattering kernels and show clearly the effects of borel resummation . as examples , we discuss the asymmetries for @xmath79 and @xmath80 , which to first approximation are given by @xmath81 where @xmath82 . using asymptotic distribution amplitudes , the @xmath26 parameters for @xmath3 decays are identical to those for @xmath25 . in figure [ fig : plots ] , results for the two cp asymmetries are shown both in the strict large-@xmath83 limit ( right ) and in the modified scheme discussed in section [ sec : mod ] . in both schemes , the asymmetries increase strongly at higher orders in perturbation theory , such that the borel - resummed results are almost a factor of two larger than the results found at one - loop order . the two - loop contributions appear to be the most important higher - order effects . this trend suggests that the asymmetries predicted from exact nlo calculations in @xcite will become larger upon the inclusion of higher - order terms . ( as concerns the magnitude of the cp asymmetries , we believe that the true values will be somewhere in between the two schemes . ) it is also apparent from the figure that , in all cases , higher - order perturbative corrections are much more important than the power corrections suggested by the values of the renormalon ambiguities . this provides additional motivation for calculating the imaginary part of the hard - scattering kernels at two - loop order . in fact , the two - loop imaginary parts ( combined with the one - loop real parts ) of the decay amplitudes are required for a consistent description of the cp asymmetries at nlo in rg - improved perturbation theory . in this paper we have used the renormalon calculus and the large-@xmath17 limit of qcd to develop a toy model for higher - order corrections to the qcd factorization approach as applied to @xmath1 decays into two light pseudoscalar mesons . we have presented novel results for the wilson coefficients in the large-@xmath17 limit , and have used them to show that the rg invariance of the hard - scattering kernels entering the qcd factorization formula is preserved under borel resummation . both the wilson coefficients and the coefficient functions @xmath26 appearing in the amplitudes for @xmath78 decays simplify greatly in the large-@xmath17 limit , but still allow for a realistic study of power corrections and higher - order perturbative contributions to qcd factorization . we have presented exact expressions for the borel transforms of the hard - scattering kernels in the factorization formula , from which is it possible to derive the terms of order @xmath22 in the perturbative expansions of the kernels . we have analysed the power corrections from soft gluon exchange and obtained a simple model for the first and second - order corrections to the parameters @xmath26 . the only numerically significant power correction arises from soft gluon exchange with the spectator quark , which introduces an error in the cp - averaged @xmath73 branching ratio of about ten percent . this is roughly the same as the errors associated with the hadronic uncertainties and power corrections that were estimated in @xcite . the power corrections to the real part of the penguin coefficients are of order @xmath23 , and those to the imaginary part are zero within the large-@xmath17 limit , which offers numerical evidence against models proposing dynamically enhanced penguins . a recurring result of our study is that the imaginary parts of the hard - scattering kernels may be larger than expected from previous applications of qcd factorization . a calculation of the imaginary part of the kernels at two - loop order would be highly desirable , since it would give an answer to the question of whether there are large multi - loop contributions to these decays that are not captured in the large-@xmath17 limit . m. beneke , g. buchalla , m. neubert and c.t . sachrajda , phys . * 83 * ( 1999 ) 1914 [ hep - ph/9905312 ] . m. beneke , g. buchalla , m. neubert and c.t . sachrajda , nucl . b * 591 * ( 2000 ) 313 [ hep - ph/0006124 ] . m. beneke , g. buchalla , m. neubert and c.t . sachrajda , nucl . b * 606 * ( 2001 ) 245 [ hep - ph/0104110 ] . a. khodjamirian , nucl . b * 605 * ( 2001 ) 558 [ hep - ph/0012271 ] . c. n. burrell and a. r. williamson , phys . d * 64 * ( 2001 ) 034009 + [ hep - ph/0101190 ] . t. becher , m. neubert and b.d . pecjak , nucl . b * 619 * ( 2001 ) 538 + [ hep - ph/0102219 ] . m. beneke and v.m . braun , phys . b * 348 * ( 1995 ) 513 [ hep - ph/9411229 ] . m. neubert , phys . d * 51 * ( 1995 ) 5924 [ hep - ph/9412265 ] . for a review , see : m. beneke , phys . rep . * 317 * ( 1999 ) 1 [ hep - ph/9807443 ] . g. t hooft , in : the why s of subnuclear physics , proceedings of the 15th international school on subnuclear physics , erice , sicily , 1977 , edited by a. zichichi ( plenum press , new york , 1979 ) , p. 943 . manohar and m.b . wise , phys . b * 344 * ( 1995 ) 407 [ hep - ph/9406392 ] . webber , phys . b * 339 * ( 1994 ) 148 [ hep - ph/9408222 ] ; + yu.l . dokshitzer and b.r . webber , phys . b * 352 * ( 1995 ) 451 + [ hep - ph/9504219 ] . r. akhoury and v.i . zakharov , phys . b * 357 * ( 1995 ) 646 [ hep - ph/9504248 ] ; nucl . b * 465 * ( 1996 ) 295 [ hep - ph/9507253 ] ; phys . * 76 * ( 1996 ) 2238 [ hep - ph/9512433 ] . p. nason and m.h . seymour , nucl . b * 454 * ( 1995 ) 291 [ hep - ph/9506317 ] . korchemsky and g. sterman , nucl . b * 437 * ( 1995 ) 415 [ hep - ph/9411211 ] . m. beneke and v.m . braun , nucl . b * 454 * ( 1995 ) 253 [ hep - ph/9506452 ] . g. buchalla , a.j . buras and m.e . lautenbacher , rev . * 68 * ( 1996 ) 1125 [ hep - ph/9512380 ] . n. pott , _ anomalous dimensions of four - quark operators in the large @xmath43 limit _ , preprint hep - ph/9710503 ( unpublished ) . a.g . grozin and m. neubert , phys . d * 55 * ( 1997 ) 272 [ hep - ph/9607366 ] . y. keum , h. li and a. i. sanda , phys . b * 504 * , 6 ( 2001 ) [ hep - ph/0004004 ] ; phys . d * 63 * , 054008 ( 2001 ) [ hep - ph/0004173 ] . m. ciuchini , e. franco , g. martinelli and l. silvestrini , nucl . b * 501 * , 271 ( 1997 ) [ hep - ph/9703353 ] ; m. ciuchini , e. franco , g. martinelli , m. pierini and l. silvestrini , phys . b * 515 * , 33 ( 2001 ) [ hep - ph/0104126 ] .	the renormalon calculus is used to calculate the terms of order @xmath0 in the perturbative expansions of the wilson coefficients and hard - scattering kernels entering the qcd factorization formula for hadronic @xmath1-meson decays into two light pseudoscalar mesons . the asymptotic behavior of the expansions is analyzed , and a minimal model of power corrections arising from soft `` non - factorizable '' gluon exchange to the @xmath2 decay amplitudes is obtained , which takes into account the structure of the leading and subleading infrared renormalon singularities . whereas the resulting power corrections are generally very small , some of the strong - interaction phases of the hard - scattering kernels receive sizeable two - loop corrections . the implications of these findings on cp asymmetries and branching ratios are investigated . clns 02/1777 + hep - ph/0202128 + february 2002 * higher - order corrections to qcd factorization in @xmath2 decays * matthias neubert and ben d. pecjak + _ newman laboratory of nuclear studies , cornell university + ithaca , ny 14853 , usa _ _ ( published in journal of high - energy physics , jhep02 ( 2002 ) 028 ) _
there is growing interest in hopf algebras defined from trees , permutations , and other combinatorial structures . in this paper , we provide an expository account of some of the hopf algebras that can be defined using trees , labeled trees , ordered trees and heap ordered trees . we also describe some actions of these hopf algebras on algebra of functions . we assume that the reader is familiar with the basic definitions and properties of hopf algebras , as covered , for example , in @xcite . this paper is based in part on @xcite . throughout this paper , @xmath0 will be a field of characteristic @xmath1 such as the real numbers or the complex numbers . by a _ tree _ we will mean a finite rooted tree . let @xmath2 be the set of finite rooted trees , and let @xmath3 be the @xmath0-vector space that has @xmath2 as a basis . in this section , we define an algebraic structure on @xmath3 that was introduced in @xcite . suppose that @xmath4 , @xmath5 are trees . let @xmath6 , , @xmath7 be the children of the root of @xmath4 and let @xmath8 denote the operator that operates on a tree and removes the root to produce a forest of trees : @xmath9 if @xmath10 has @xmath11 nodes ( counting the root ) , there are @xmath12 ways to attach the @xmath13 subtrees of @xmath4 that have @xmath6 , , @xmath7 as roots to the tree @xmath10 . to do this , make each @xmath14 the child of some node of @xmath10 . we let @xmath15 denote this operation that attaches a forest of trees @xmath6 , @xmath16 , @xmath7 to the nodes of a tree @xmath10 in all possible ways to produce a set of trees . the product @xmath17 of the trees @xmath4 and @xmath10 is defined to be the sum of these @xmath12 trees . we summarize this in the following definition : the product of the trees @xmath18 and @xmath5 is defined to be @xmath19 here we sum over the elements of the sets @xmath20 and @xmath21 . it is immediate that this definition can be extended by linearity from @xmath2 to the vector space @xmath3 , and that the trivial tree consisting only of the root is a right and left unit for this product . for a proof that this operation is associative , see @xcite . here are some simple examples of tree multiplication : @xmath22 we now define a coalgebra structure on the vector space @xmath3 : @xmath23 here , if @xmath24 are multisets , @xmath25 denotes the set theoretic difference . it is easily checked that the definition of a coproduct is satisfied , so that @xmath3 is a cocommutative coalgebra . this coalgebra and the algebra given in example [ shuffle - eg ] , with @xmath26 being the set of trees whose root has exactly one child , are duals of each other as graded vector spaces , with a tree with @xmath11 nodes being of degree @xmath27 , and the degree of a monomial @xmath28 being the sum of the degrees of @xmath29 , , @xmath30 . the fact that coassociativity for this coalgebra is trivial gives an easy proof of the associativity of the algebra in example [ shuffle - eg ] . another algebra of trees was described in @xcite in 1998 that is defined as follows : if @xmath31 is the set of rooted trees , form expressions @xmath32 , where @xmath33 . this is clearly a commutative algebra of monomials , which we denote @xmath34 . the coalgebra structure of @xmath34 is defined as follows : we introduce the notion of _ cut_. a cut is a set of removed edges . given a tree @xmath35 , an _ admissable cut _ of @xmath35 is one with at most one removed edge on each path from the root to any leaf . for a cut @xmath36 we let @xmath37 to be the resulting piece containing the root , and @xmath38 to be the monomial consisting of the other pieces . then define @xmath39 there is an isomorphism @xmath40 from the hopf algebra @xmath3 to the graded dual of @xmath34 satisfying @xmath41 where @xmath42 and @xmath43 . here @xmath44 is an inner product defined on @xmath34 that depends upon a factor determined by the symmetry of the arguments . for the definition of this inner product and a proof of this fact , see @xcite . it is easy to extend the definition of multiplication of trees from the vector space whose basis is the set of finite rooted trees to the vector space whose basis is the set of finite _ labeled _ rooted trees : simply label each node except the root with a label and keep the labels attached to the nodes when applying the operators @xmath21 and @xmath20 . as shown in @xcite this yields a hopf algebra . similarly , the same definition of the product and coproduct applies to the vector space whose basis is the set of finite _ ordered trees _ , as well as to the vector space whose basis is the set of finite _ ordered , labeled trees_. both produce hopf algebras . see @xcite for details . later in this paper , we will need the hopf algebra of heap ordered trees , which we define following @xcite : [ def : hot ] a standard heap ordered tree on @xmath11 nodes is a finite , rooted tree in which all nodes except the root are labeled with the numbers @xmath45 , @xmath46 , @xmath47 , , @xmath48 so that : 1 . [ hot:1 ] each label @xmath49 occurs precisely once ; 2 . [ hot:2 ] if a node labeled @xmath49 has children labeled @xmath50 , @xmath16 , @xmath51 , then @xmath52 , , @xmath53 . we denote the set of standard heap ordered trees on @xmath11 nodes by @xmath54 . let @xmath55 be the vector space over the field @xmath0 whose basis is the set of trees in @xmath54 , and let @xmath56 a heap ordered tree is a rooted tree in which every node ( including the root ) is given a different positive integer label such that condition [ hot:2 ] . of definition [ def : hot ] is satisfied . a heap ordered tree differs from a standard heap ordered tree in that the root is also labeled , and that the labels can be taken from a larger set of positive integers . the number of distinct heap ordered trees on @xmath11 nodes is @xmath57 . heap ordered trees occur naturally when studying permutations and differential operators , as we shall see below . to summarize , with the product and coproduct as defined above , the vector spaces whose bases are ( i ) the set of rooted trees , ( ii ) labeled rooted trees , ( iii ) ordered trees , ( iv ) labeled ordered trees , ( v ) heap ordered trees , and ( vi ) labeled heap ordered trees are all hopf algebras ( see @xcite ) . in section [ section : hot ] , we show that there is an isomorphism from the hopf algebra of heap ordered trees to a hopf algebra on permutations . in section [ section : psi1 ] , we show how hopf algebras of rooted labeled trees arise naturally when computing with derivations on the @xmath0-algebra of functions from @xmath58 to @xmath0 . in section [ section : psi2 ] , we show how hopf algebras of rooted , ordered , labeled trees arise naturally when computing with derivations on the @xmath0-algebra of @xmath59 functions on a @xmath59 manifold . the definitions and basic properties about shuffle algebras that we summarize in this section will be needed laer . a _ shuffle _ of the sequences @xmath60 and @xmath61 is a permutation @xmath62 of @xmath63 satisfying : if @xmath64 , @xmath65 , and @xmath66 , then @xmath67 ; if @xmath64 , @xmath68 , and @xmath69 , then @xmath67 . in other words , a shuffle of the sequences @xmath60 and @xmath61 is a permutation of @xmath70 that preserves the order of @xmath60 and of @xmath61 . we denote this shuffle by @xmath71 . [ shuffle - eg ] let @xmath26 be a set of non - commuting variables , and let @xmath72 be the vector space with basis all monomials in the variables @xmath73 . define a product on @xmath72 as follows . if @xmath74 , @xmath75 , define @xmath76 where the sum is taken over all shuffles @xmath77 . the coproduct on @xmath72 is defined in the same way as for the free non commutative algebra on the set @xmath26 , given by @xmath78 fix a field @xmath0 , and let @xmath79 denote the vector space over @xmath0 in the non - commuting variables @xmath80 , @xmath16 , @xmath81 . then the shuffle product defined above makes @xmath79 into a hopf algebra over @xmath0 . let @xmath82 denote the symmetric group on @xmath27 symbols , let @xmath83 denote the vector space over the field @xmath0 with basis @xmath82 , and let @xmath84 in this section , we define a hopf algebra structure on @xmath85 following @xcite . we begin with some notation . let @xmath86 denote the cycle in @xmath87 which sends @xmath88 to @xmath89 , @xmath89 to @xmath90 , @xmath16 , and @xmath91 to @xmath88 . every permutation is a product of disjoint cycles . if @xmath86 is a cycle , then there is a string naturally associated with the cycle that we write @xmath92 . now let @xmath93 and @xmath94 be two permutations , each written as a product of disjoint cycles . we denote the corresponding strings by @xmath95 and @xmath96 respectively . we call the elements of @xmath97 _ attachment points _ for the cycles @xmath98 , , @xmath99 on the permutation @xmath100 . we will also define @xmath101 to be the @xmath102 attachment point . also , if @xmath62 is a permutation on @xmath103 , let @xmath104 be the permutation on @xmath105 that sends @xmath106 to @xmath107 . the definition of the product in the bialgebra is simpler if we introduce the _ standard order _ of a permutation , which is defined as follows : we say that a permutation @xmath108 that is expressed as a product of cycles @xmath109 is in standard order if the cycles @xmath110 are written so that 1 . @xmath111 , @xmath112 , @xmath16 , @xmath113 2 . @xmath114 , @xmath16 , @xmath115 in other words , a product of cycles @xmath109 is written in standard order if each cycle @xmath110 starts with its smallest entry , and if the cycles @xmath98 , , @xmath99 are ordered so that their starting entries are decreasing . a permutation can always be written in standard order since disjoint cycles commute , and since a single cycle is invariant under a cyclic permutation of its string . we now define the _ heap product _ of two permutations . we define the heap product @xmath116 of @xmath108 and @xmath117 as follows : 1 . put @xmath62 and @xmath118 in standard order ; 2 . replace @xmath62 by @xmath119 ; 3 . form terms as follows : if @xmath120 is one of the cycles in @xmath62 , attach the string @xmath14 to any one of the @xmath11 attachment points ; if the attachment point is one of @xmath121 , , @xmath122 , say @xmath123 , place the string @xmath95 to the right of @xmath123 ; otherwise ( if the attachment point is @xmath101 ) we multiply the term we are constructing by @xmath120 ; 4 . the product @xmath116 is the sum of all the terms constructed in this way , taken over all the cycles in @xmath62 and over all attachment points . note that there are @xmath12 terms in @xmath116 . we now describe the coalgebra structure of @xmath85 : define a function @xmath124 from permutations to permutations as follows : let @xmath125 and let let @xmath126 be the labels ( in order ) which occur in the @xmath14 . ( if @xmath127 fixes @xmath49 , we include a 1-cycle @xmath128 as a factor in @xmath127 . ) the permutation @xmath124 is the permutation in @xmath129 gotten by replacing @xmath130 with @xmath131 in @xmath127 . for example , if @xmath132 , then @xmath133 equals @xmath134 . the coalgebra structure of @xmath85 is defined as follows . let @xmath135 , and let @xmath136 . if @xmath137 let @xmath138 . note that if @xmath139 , then @xmath140 . define @xmath141 the following theorem is proved in @xcite . with the product and coproduct defined above , @xmath85 is a hopf algebra . it turns out that this hopf algebra structure on @xmath85 is isomorphic to the hopf algebra structure on heap - ordered trees defined above @xcite . there are several other different hopf algebra structures on permutations that can be defined , including the malvenuto - reutenauer hopf algebra . the malvenuto - reutenauer is a non - commutative , non - cocommutative , self - dual , graded hopf algebra . in contrast , the hopf algebra on permutations defined above is non - commutative , cocommutative , graded hopf algebra . for descriptions of these other hopf algebras on permutations see @xcite , @xcite , @xcite . in this and the next section , we describe some actions of hopf algebras of trees on @xmath0-algebras @xmath142 of functions . in this section , we consider the case of @xmath0-algebras @xmath142 of functions on a space @xmath143 . for example , @xmath142 can be the space of polynomials , rational functions or @xmath59 functions on @xmath26 . in the next section , we consider the more general case of @xmath0-algebras of functions on a @xmath27-dimensional algebraic group @xmath26 or a @xmath27-dimensional @xmath59 manifold @xmath26 . let @xmath142 be a commutative @xmath0-algebra , and let @xmath144 be a @xmath0-bialgebra . the algebra @xmath142 is a _ left @xmath144-module algebra _ if @xmath142 is a left @xmath144-module for which @xmath145 where @xmath146 , @xmath147 , and @xmath148 , @xmath149 . consider the vector space spanned by finite rooted trees , each of whose nodes except for the root are labeled with one of the formal symbols @xmath150 , , @xmath151 . this generates an algebra using the product defined above that we denote @xmath152 or , if no confusion is possible , simply by @xmath3 . note that the same symbol may label more than one node of the tree . fix an @xmath0-algebra @xmath142 of functions and consider the situation where the @xmath153 are not formal symbols , but are differential operators of the form @xmath154 with @xmath155 . here the algebra @xmath142 is typically either the the algebra of polynomial functions @xmath157 , , @xmath158 $ ] , the algebra of rational functions @xmath159 , , @xmath160 , or the algebra of smooth functions @xmath161 . let @xmath144 denote a hopf algebra of trees labeled with derivations of @xmath142 . in this section , we define a natural action of @xmath144 on @xmath142 that turns @xmath142 into a @xmath144-module algebra . we define a map @xmath162 as follows . let @xmath163 , , @xmath164 have @xmath165 nodes , and let @xmath166 . number the non root nodes of @xmath35 from 1 to @xmath0 . we define @xmath167 as follows : for the root , form the term @xmath168 where the @xmath13 children of the root are numbered @xmath169 , , @xmath170 . for a nonroot node numbered @xmath131 and labeled with @xmath171 , form the term @xmath172 then @xmath173 the following theorem @xcite captures an important property of this @xmath174-map . fix derivations @xmath175 with @xmath176 , and @xmath177 . let @xmath178 be the free associative algebra generated by the formal symbols @xmath179 , @xmath16 , @xmath180 . let @xmath181 be the hopf algebra consisting of rooted trees labeled with the derivations @xmath182 . let @xmath183 be the higher order derivations generated by the derivations @xmath179 , @xmath16 , @xmath180 . then with the @xmath174-map as defined above makes the following diagram commute : @xmath184 the right arrow is the map induced by the fact that @xmath178 is freely generated by the @xmath153 , and @xmath153 maps to the tree with one node other that the root which is labeled with @xmath153 ; the diagonal arrow is induced by the fact that the @xmath153 are derivations . an expression such as @xmath185 corresponds to 24 trees , 18 of which cancel ( each cancellation saves @xmath186 differentiations ) . the surviving six differentiations are @xmath187 corresponding to the six trees : @xmath188 @xmath189 this example illustrates the kinds of cancellations of higher order derivations that are captured by trees . it turns out that the hopf algebra @xmath144 of trees labeled with derivations is a natural structure for computing symbolically with derivations and higher order derivations . see @xcite and @xcite . in this section , we consider what happens when @xmath26 is not @xmath58 , but instead a @xmath59 manifold . in this case , we can define @xmath142 to be the @xmath0-algebra of @xmath0-valued @xmath59 functions on the manifold . we use the connection as follows . if @xmath200 and @xmath166 , define : @xmath201 and extend using induction , the definition of tree multiplication , and a consistency requirement for subtrees of larger trees . we show in @xcite that construction defines a @xmath144-module algebra structure on @xmath142 . let @xmath142 be a commutative @xmath0-algebra , and let @xmath202 be a connection on the lie algebra @xmath190 of derivations of @xmath142 . then the construction above gives a @xmath203-module structure on @xmath142 . this module structure induces a map @xmath204 . so that the following diagram commutes : @xmath205 p.e . crouch and r. l. grossman , the explicit computation of integration algorithms and first integrals for ordinary differential equations with polynomial coefficients using trees , proceedings of the 1992 international symposium on symbolic and algebraic computation , acm press , pages 8994 .	we provide an expository account of some of the hopf algebras that can be defined using trees , labeled trees , ordered trees and heap ordered trees . we also describe some actions of these hopf algebras on algebra of functions .
most liquids become denser when cooled and more viscous when compressed . in contrast , water becomes less dense when cooled ( density , or thermodynamic , anomaly ) and its diffusivity increases upon pressurization ( diffusion , or dynamic , anomaly ) . these anomalies , which disappear at high enough temperature and pressure , are not unique to water . other liquids with local tetrahedral order ( e.g. , silica and silicon ) also exhibit thermodynamic and dynamic anomalies @xcite . a possible explanation of these anomalies is the tendency of these substances to form local open structures not present in simple liquids . however , establishing a precise and quantitative link between the microscopic structure and the dynamic and thermodynamic anomalies of tetrahedral liquids has proved elusive until recently . errington and debenedetti @xcite ( ed ) studied the relation between microscopic structure and the anomalies of liquid water by introducing two simple metrics : a translational order parameter @xmath4 @xcite , quantifying the tendency of molecule pairs to adopt preferential separations , and an orientational order parameter @xmath11 @xcite , quantifying the extent to which a molecule and its four nearest neighbors adopt a tetrahedral local structure , as in the case of hexagonal ice . a useful way of investigating structural order in liquids is to map state points into the @xmath12 plane . such a representation was introduced by torquato and coworkers @xcite , who first applied it to sphere packings and referred to it as an order map . ed used the order map to investigate structural order in water @xcite . because of the distinctive features discovered in that study , in what follows we refer to water - like order maps as the ed order map . using molecular dynamics simulation of the spc / e @xcite model , ed found that the state points accessible to liquid water define a two - dimensional region in the @xmath12 plane , meaning that in general @xmath4 and @xmath11 are independently variable in liquid water ( i.e. , equilibrium state paths exist along which one order metric varies while the other does not ) . they also found a dome - shaped region in the ( @xmath13 ) plane within which isothermal compression leads to a decrease in @xmath4 and @xmath11 . this _ decrease _ in order upon compression constitutes a structural anomaly : simple liquids , in contrast , always become more ordered upon compression . ed further found that dynamic and thermodynamic anomalies define nested domes in the ( @xmath2 ) plane : the structural anomalies dome contains the dynamic anomalies dome , which in turn contains the thermodynamic anomalies dome . this means that whenever the thermal expansion coefficient is negative , the diffusivity must necessarily increase upon isothermal compression . ed showed that all state points exhibiting structural , dynamic or thermodynamic anomalies define a line on the ( @xmath14 ) plane , meaning that when water exhibits anomalous behavior , its translational and orientational order metrics become strictly coupled . this is clear evidence of the relationship between structure and water anomalies . et al . _ subsequently found qualitatively similar behavior in molten silica s order map @xcite . however , in the case of silica , it was found that state points corresponding to anomalous behavior define a narrow stripe in the ( @xmath14 ) plane instead of a strict line . furthermore , unlike in water , the region of dynamic anomalies was found to contain that of structural anomalies . for simple spherically - symmetric liquids , including hard spheres @xcite and lennard - jones @xcite the order map was found to be a positively - sloped line in the ( @xmath14 ) plane , indicating that translational and orientational order are always strictly and positively correlated . in this case , of course , the appropriate metric for orientational order does not measure tetrahedrality ; rather , the bond - orientational order parameter introduced by steinhardt _ et al . _ @xcite was used . an important result from these studies is the fact that the order map for the lennard - jones system above its critical density is identical to that of hard spheres . furthermore , in these simple systems that do not exhibit thermodynamic or dynamic anomalies , compression always leads to an increase in the order metrics . in 1970 hemmer and stell @xcite showed that in fluids interacting via pairwise - additive , spherically - symmetric potentials consisting of a hard core plus an attractive tail , softening of the repulsive core can produce additional phase transitions . this pioneering study elicited a considerable body of work on so - called core - softened potentials @xcite . this generic term denotes continuous potentials with inflections in the repulsive core @xcite , discontinuous potentials with the core softened by shoulders or ramps @xcite , or lattice models with nearest - neighbor attraction and next - nearest neighbor repulsion @xcite . it is now well - established that such potentials can generate water - like density and diffusion anomalies @xcite . this important finding implies that strong orientational interactions , such as those that exist in water and silica , are not a necessary condition for a liquid to have thermodynamic and dynamic anomalies . the above discussion implies the existence of two well - defined classes of liquids : simple and water - like . the former which interact via spherically - symmetric non - softened potentials , do not exhibit thermodynamic nor dynamic anomalies , and their order map is a line . in water - like liquids , interactions are orientation - dependent ; these liquids exhibit dynamic and thermodynamic anomalies , and their order map is in general two - dimensional but becomes linear ( or quasi - linear ) when the liquid exhibits structural , dynamic or thermodynamic anomalies . intermediate between these well - defined extremes is the class of core - softened liquids , which interact via spherically - symmetric potentials but can also exhibit water - like thermodynamic and dynamic anomalies . two questions arise naturally from this emerging taxonomy of liquid behavior . first , is structural order in core - softened fluids hard - sphere or water - like ? second , is it possible to seamlessly connect the range of liquid behavior from hard spheres to water - like by a simple and common potential , simply by changing a physical parameter ? in a recent study @xcite we addressed the first question . we showed that a core - softened potential with two characteristic length scales not only can gives rise to water - like diffusive and density anomalies , but also to an ed water - like order map . this implies that orientational interactions are not necessary in order for a liquid to have structural anomalies . in this work we address the second question . specifically , we use the ratio of characteristic length scales as a control parameter to investigate the evolution of dynamic , thermodynamic and structural anomalies . in this manner we show that the family of tunable spherically - symmetric potentials so generated evolves continuously between water - like and hard sphere behavior . to our knowledge this is the first time that essential aspects of the wide range of liquid behavior encompassed by hard spheres and tetrahedrally - coordinated network - formers can be systematically traversed by varying a single control parameter . this paper is structured as follows . sections ii , iii , and iv provide details on the interaction potential , simulation method , and order parameters , respectively . results are presented in section v. conclusions and some suggestions for future work are provided in section vi . we perform discrete md simulations to study the equation of state , diffusion coefficient and structural order as measured by the ed order map , for a fluid whose particles interact via a pairwise - additive , spherically - symmetric potential that gives rise to both thermodynamic and dynamic water - like anomalies . the model was introduced by jagla @xcite ; the potential energy @xmath15 between a pair of particles separated by a distance @xmath16 is given by ( see fig . [ ramppotential ] ) . @xmath17 the shorter distance @xmath18 corresponds to the hard core , and the longer distance @xmath19 characterizes a softer repulsion range that can be overcome at high pressure . because of its shape , it is called the ramp potential . the constant slope of the ramp potential for @xmath20 keeps the force between particles @xmath21 constant , so the product of separation and force @xmath22 will decrease when the separation @xmath16 decreases . this satisfies the mathematical meaning of core - softening @xcite and under these conditions the thermodynamic ( density ) anomaly can be qualitatively explained by invoking the virial theorem @xcite . of interest is the ratio between the two characteristic length scales , @xmath18 and @xmath19 , @xmath23 which can vary between 0 and 1 . ref . @xcite investigated the one - scale ( @xmath24 ) and two - scale ( @xmath10 ) ramp potentials . here we investigate the full range @xmath25 , with @xmath26 . we use discrete md simulation ; details are given in ref . we use the nvt ensemble for a system composed of @xmath27 particles with periodic boundary conditions , and we control the temperature with the berendsen thermostat @xcite . however , we note that we use different units than in ref . @xcite : the distance @xmath16 , number density @xmath28 , pressure @xmath29 and temperature @xmath30 are all normalized with respect to the soft core distance @xmath19 and the potential @xmath31 at @xmath32 . we also investigate systems with different number of particles ( @xmath33 ) and confirm that the results do not depend on the number of particles . the simulation ranges of temperature and density fully cover the region where density , diffusion and structural anomalies occur . the translational order parameter @xcite is defined as , @xmath34 here @xmath35 is the radial distance @xmath16 scaled by the mean interparticle distance @xmath36 , @xmath28 is the number density , @xmath37 is the pair correlation function , and @xmath38 a numerical cutoff . we choose @xmath38 so that it corresponds to one - half the simulation box size , and we verify that our system size is always large enough so that @xmath39 at half the box size . for a completely uncorrelated system , @xmath40 , and thus @xmath41 . for systems with long - range order , the modulations in @xmath37 persist over large distances , causing @xmath4 to grow . between these limits , @xmath4 will change as a consequence of the dependence of @xmath42 upon @xmath30 and @xmath28 . an orientational order parameter based on spherical harmonic function was introduced by steinhardt _ _ @xcite and used in refs . @xcite . in this definition , all vectors connecting nearest neighbors ( i.e. , particle pairs whose separation is less than the first minimum of the radial distribution function ) are considered . each of these vectors , also called ` bond ' , defines an azimuthal and polar angle , and the corresponding spherical harmonic function is evaluated . the orientational order parameter used in refs . @xcite involves the average of each spherical harmonic function over all bonds . the orientational order parameter used for water and silica in refs . @xcite involves , first , the evaluation of the local tetrahedral order for each particle with respect to its four nearest neighbors , and then , the average of this quantity over all the molecules of the system . in the definition of the orientational order parameter used in refs . @xcite there is no such concept of ` local order ' for an individual particle . moreover , the number of bonds associated to each particle is not fixed , but instead it changes with temperature and pressure . these two differences led , in ref . @xcite , to the introduction of a slightly modified version of the original orientational order parameter introduced by steinhardt et al . in ref . the resulting order metric is based on the idea of a ` local order ' for each particle , analogous to refs . the ed maps obtained with the original ( global ) and modified ( local ) definitions of orientational order are qualitatively similar . in this work we use the same order parameter introduced in ref . we define twelve bonds connecting each particle with its twelve nearest neighbors . each bond is characterized by its azimuthal and polar angles @xmath43 and the corresponding spherical harmonic @xmath44 is computed . the orientational order parameter associated with each particle @xmath45 is @xmath46^\frac{1}{2}. \label{q6}\ ] ] here , @xmath47 denotes the average of @xmath48 over the 12 bonds associated with particle @xmath45 . for @xmath49 @xcite , @xmath50 has a large value for most crystals such as fcc , hcp and bcc @xcite . the values of @xmath51 for each molecule in the system follow a gaussian distribution . @xmath5 , the averaged value of @xmath51 over all particles @xmath45 , is used to characterize the local order of the system . this definition of order parameter is analogous to that used in water . for water , the solid at low pressure is hexagonal ice where each molecule has four neighbors . the orientational order parameter is maximum in the ice configuration and decreases as the system becomes less ice - like . for the ramp potential , the solid phase at low pressure has a fcc structure where each particle has twelve nearest neighbors . @xmath5 has a maximum value in the fcc lattice ( @xmath52 ) and decreases as the system becomes less correlated ( for uncorrelated systems , @xmath53 ) . since the pair correlation function @xmath42 is used to compute the translational order parameter @xmath4 ( eq . [ top ] ) , we first discuss the effect of density on @xmath42 . [ rdfs ] shows the effects of compression on @xmath42 at low temperature , @xmath54 , for the various values of @xmath0 considered in this study . in all cases , there is no inner peak at @xmath55 for @xmath56 and 1.21 , and only the outer peak at @xmath57 is present at these densities . the inner peak at @xmath55 , which is broad and of modest height at @xmath58 , becomes sharper and more pronounced upon further compression . interestingly , structural changes brought about by compression become progressively longer - ranged as @xmath0 increases . thus , for @xmath24 and 2/7 , the major changes in @xmath42 involve the development of structure at length scales @xmath59 associated with the growth of the inner peak at @xmath55 . however , for @xmath60 and 6/7 , structural changes upon compression occur at distances larger than @xmath19 . in particular , for @xmath61 , the effects of compression are clearly discernible at @xmath62 . these effects of density on the pair correlation function underlie the evolution of @xmath4 upon compression , shown in figs . [ tq - d](a1)-(e1 ) . consider for example the @xmath54 isotherm when @xmath63 . it can be seen that @xmath4 displays a non - monotonic dependence on density : it increases upon compression at low densities , @xmath64 , decreases over the intermediate density range @xmath65 , and increases again at high densities , @xmath66 . the initial increase at low densities is associated with the growth of @xmath67 . the emergence of structure associated with the inner ( hard ) core causes @xmath4 to decrease at intermediate densities because the initial , modest growth of @xmath68 at @xmath69 causes @xmath70 to decrease with respect to its low - density value of 1 ( see eq . [ top ] ) . upon further compression , the growth of @xmath71 above 1 eventually contributes additional area to the integral of @xmath70 , causing @xmath4 to increase . this qualitative behavior of @xmath4 is similar for @xmath72 and 4/7 , and is more pronounced at low @xmath30 . for @xmath61 , close to the hard sphere limit , the pronounced growth of the inner peak upon compression gives rise to a monotonic density dependence of @xmath4 . the case @xmath73 is clearly transitional , with non - monotonic behavior at low temperature changing to hard - sphere - like monotonic growth of t upon compression at high temperature . orientational order , as measured by @xmath5 , shows a pronounced dependence on @xmath0 , illustrated in figs . [ tq - d](a2)(e2 ) . when @xmath24 ( no hard core ) , @xmath5 increases monotonically with density for all @xmath30 . when @xmath63 , @xmath5 begins to exhibit non - monotonic behavior upon compression for this particular value of @xmath0 the trend is very mild , and is best described as a virtual insensitivity of @xmath5 to compression , except for an initial increase at low enough densities . for @xmath10 and 5/7 , orientational order exhibits a marked non - monotonic dependence on density , especially at low temperatures . when coupled with the corresponding behavior of @xmath4 , this corresponds to a water - like structural anomaly , whereby both order metrics decrease upon isothermal compression . when @xmath61 , which is close to the hard sphere value ( @xmath7 ) , orientational order increases monotonically upon compression . thus , there exists a narrow interval of @xmath0 within which the ramp fluid shows water - like structural order , whereas in the pure ramp ( @xmath24 ) and quasi - hard - sphere limits ( @xmath74 ) @xmath5 behaves conventionally upon compression . cross - plotting the order metrics against each other generates the order map , whose evolution as a function of @xmath0 is depicted in fig . [ ordermap ] . for all values of @xmath0 except 6/7 , state points fall on a two - dimensional region , signifying that @xmath4 and @xmath5 can be varied independently . as is the case for silica and water @xcite , we find , for all values of @xmath0 , an inaccessible region where no liquid state points can be found . in the pure ramp ( @xmath24 ) case , the pronounced non - monotonic dependence of @xmath4 on density gives rise to isotherms with well - characterized @xmath4-minima , the locus of which defines the boundary between the accessible and inaccessible regions of the order map . for @xmath0 = 2/7 , the barely discernible non - monotonic dependence of @xmath5 on density gives rise to loops along isotherms . the non - monotonic behavior of @xmath5 is fully developed for @xmath0 = 4/7 . this gives rise to an order map with states corresponding to structural anomalies lying on a narrow stripe of the order map adjacent to the boundary between the accessible and inaccessible regions . this behavior is strikingly analogous to that of water . the insensitivity of structural order to temperature , a distinguishing feature of hard spheres , can be clearly seen in fig . [ ordermap](e ) by the virtual collapse of all isotherms in the @xmath0 = 6/7 case . the transition from water - like to hard sphere order map occurs in the narrow interval @xmath75 . in particular , for @xmath0 = 5/7 , there is a clear evolution from water - like low-@xmath30 behavior ( @xmath76 ) to hard - sphere - like high-@xmath30 behavior ( @xmath77 ) . we now discuss the regions of the phase diagram where structural , dynamic and thermodynamic anomalies occur . in water @xcite , structural , dynamic , and thermodynamic anomalies occur as nested domes in the ( @xmath2 ) or ( @xmath3 ) planes . structural anomalies define the outer dome , within which isothermal compression results in a decrease of both translational and orientational order . dynamic anomalies define an intermediate dome , lying entirely within the structural anomalies dome , and within which isothermal compression leads to an increase in the diffusion coefficient . thermodynamic anomalies define the innermost dome , within which water expands when cooled isobarically . in silica @xcite , dynamic anomalies define the outer dome , structural anomalies the intermediate dome , and thermodynamic anomalies define the inner dome . thus , in both cases negative thermal expansion implies also diffusive and structural anomalies , but in silica diffusive anomalies occur over a broader range of densities and temperatures than structural anomalies , the opposite being true in water . [ ptphase ] shows the loci of dynamic and thermodynamic anomalies for three values of @xmath0 . similar to water and silica , in ramp fluids thermodynamic anomalies occur over a narrower temperature and density range than dynamic anomalies . in other words , if a ramp fluid is at a state point where it expands when cooled isobarically , its diffusion coefficient necessarily increases upon isothermal compression . it can be seen that upon increasing @xmath0 , the range of temperatures where anomalies occurs shrinks , whereas the upper limit of density ( or pressure ) where anomalies can occur increases . the shrinking of the temperature range where anomalies occur follows from the fact that increasing @xmath0 makes the fluid progressively hard - sphere - like , and there are no anomalies in a hard sphere fluid . [ qttmddm ] shows the relationship between the loci of dynamic , thermodynamic and structural anomalies . in water , the low - density and high - density branches of the dome of structural anomalies correspond to tetrahedrality maxima and translational order minima , respectively . for the pure ramp case ( @xmath0 = 0 ) , the orientational order increases monotonically with density over the range of temperatures explored here . accordingly , as seen in fig . [ qttmddm](a ) , the dynamic and thermodynamic anomalies domes are bounded by loci of translational order extrema ( maxima : line c ; minima : line a ) . between lines c and a , compression leads to a decrease in translational order . for @xmath0 = 4/7 and 5/7 , the locus of orientational order maxima ( b ) provides a low - density bound to the existence of thermodynamic and dynamic anomalies . thus , for these two values of @xmath0 , ramp fluids exhibit a water - like cascade of anomalies ( structural , dynamic , thermodynamic ) . for @xmath63 , @xmath5 maxima are barely discernible and the two order metrics are only weakly coupled to each other . accordingly , the locus of weak orientational order maxima ( b ) is not a relevant indicator of dynamic or thermodynamic anomalies . in this work we have investigated thermodynamic , dynamic and structural anomalies in ramp potential fluids , as a function of the ratio @xmath0 of length scales corresponding to the inner hard core , and to the outer edge of the ramp . we find that thermodynamic and dynamic anomalies exist for @xmath0 = 0 , 2/7 , 4/7 and 5/7 , but not for = 6/7 . as in water and silica , the loci of anomalies form nested domes in the ( @xmath2 ) plane , inside which the thermal expansion coefficient is negative ( inner dome ) and the diffusivity increases upon compression ( outer dome ) . the limit @xmath0 = 1 corresponds to hard spheres , and the absence of anomalies for @xmath0= 6/7 indicates approach to hard sphere behavior . the order map of this family of ramp fluids is water - like at @xmath0 = 4/7 and hard - sphere - like at @xmath0 = 6/7 . thus , by varying the ratio of characteristic length scales , the family of ramp potentials spans the range of liquid behavior from hard spheres to water - like . these findings show that orientational interactions are not necessary for the existence of thermodynamic , dynamic , or structural anomalies . instead , water - like behavior apparently emerges in this spherically - symmetric family of fluids through the existence of two competing length scales . although thermodynamic and dynamic anomalies exist almost over the entire range of the control parameter , the combination of thermodynamic and dynamic anomalies plus a water - like order map occurs over a narrow range of @xmath0 . it is interesting to note that a distinguishing feature of water is the fact that the ratio of radial distances to the first and second peaks of the oxygen - oxygen pair correlation function is not @xmath78 , as in simple liquids , but @xmath79 . this is close to 0.571 ( @xmath0 = 4/7 , the ratio of @xmath18 to @xmath19 that gives rise to water - like structural , dynamic and thermodynamic anomalies ) . in water , isothermal compression pushes molecules from the second shell towards the first shell , gradually filling the interstitial space @xcite . likewise , in the ramp potential , isothermal compression pushes molecules from the soft core ( @xmath19 ) to the hard core ( @xmath18 ) . further work is needed to establish whether a ratio of competing length scales close to 0.6 is generally associated with water - like anomalies in other core - softened potentials , for example linear combinations of gaussian @xcite potentials . in this work we used the terminology water - like to denote structural , diffusion , and density anomalies . the increase in water s isothermal compressibility upon isobaric cooling , another of this liquid s canonical anomalies , is also trivially captured by the ramp potential , because thermodynamic consistency arguments @xcite mandate that the compressibility increase upon cooling whenever there exists a negatively - sloped locus of density maxima in the ( @xmath80 ) plane . the ramp potential , when supplemented by explicit @xcite or mean - field attractions @xcite , gives rise to liquid - liquid immiscibility and a critical point distinct from the one associated with the vapor - liquid transition . a liquid - liquid transition has been observed experimentally in phosphorus @xcite , n - butanol @xcite and triphenyl phosphite @xcite , and strong experimental evidence consistent with liquid - liquid immiscibility also exists for water @xcite . computer simulations of silicon @xcite , silica @xcite , carbon @xcite and water @xcite also indicate the presence of a liquid - liquid transition . a systematic study of the effects of @xmath0 and the ratio of characteristic energies ( @xmath31 and the attractive well depth ) on the existence of a liquid - liquid transition , the positive or negative slope of the line of first - order liquid - liquid transitions in the ( @xmath80 ) plane , and the relationship , if any @xcite , between the liquid - liquid transition and density anomalies , would shed important new light on the phenomenon of liquid polyamorphism @xcite . it is generally accepted that strong orientation - dependent interactions underlie many of the distinctive properties of associating , network - forming liquids such as water . atomic liquids , on the other hand , exhibit simpler behavior , and in particular do not show structural , thermodynamic , or dynamic anomalies of the type discussed here . in this work we have shown that key properties of these apparently distinct categories of liquids can be bridged systematically by varying the ratio of two length scales in a family of spherically - symmetric potentials in which orientation - dependent interactions are absent by design . what other spherically - symmetric potentials , in addition to those possessing competing length scales , may give rise to water - like anomalies , is among the interesting questions arising from this study that we will pursue in future work . we thank nsf grants che 0096892 and che 0404699 for support and yeshiva university for cpu time . p. c. hemmer and g. stell , phys . * 24 * , 1284 ( 1970 ) . g. stell and p. c. hemmer , j. chem . phys . * 56 * , 4274 ( 1972 ) . j. m. kincaid , g. stell and c. k. hall , j. chem . phys . * 65 * , 2161 ( 1976 ) . j. m. kincaid , g. stell and e. goldmark , j. chem . phys . * 65 * 2172 ( 1976 ) . j. m. kincaid and g. stell , j. chem . phys . * 67 * , 420 ( 1977 ) . corresponds to the hard core , @xmath19 characterizes the onset of soft repulsion . when @xmath24 ( left figure ) we have a pure ramp potential ( no hard core ) . when @xmath7 ( right figure ) we have a hard sphere potential . @xmath81 is the ratio near which the system exhibits water - like structural , dynamic and thermodynamic behavior . , width=453 ] at @xmath54 for different @xmath82 values . the arrows indicate the direction of increasing density . the density values are 1.0 , 1.21 , 1.66 , 2.19 for @xmath83 and 1.0 , 1.21 , 1.66 , 1.86 for @xmath84 . the distance @xmath16 is normalized by @xmath19 , the soft core length . the curves are shifted vertically for clarity . , width=453 ] for different @xmath0 values . the solid lines are polynomial fits to the data . the lower panels ( a2)(e2 ) show the density - dependence of the orientational order parameter @xmath85 . in each panel , the different curves correspond to isotherms ( top to bottom ) @xmath86 . , width=604,height=272 ] values . the region of diffusion anomalies is defined by the loci of diffusion minima and maxima ( dm ) inside which the diffusivity increases upon isothermal compression . the thermodynamically anomalous region is defined by locus of temperatures of maximum density ( tmd ) , inside of which the density increases when the system is heated at constant pressure . , width=377,height=302 ] plane . for @xmath24 and 2/7 , the domes of dynamic and thermodynamic anomalies are bounded by loci of @xmath4 maxima ( c ) and minima ( a ) , between which isothermal compression cause a _ decrease _ in translational order . for @xmath10 and 5/7 , the domes of dynamic and thermodynamic anomalies are bounded by loci of @xmath5 maxima ( b ) and t minima ( a ) , between which isothermal compression cause a @xmath87 in both translational and orientational order ( structural anomaly ) . this cascade of anomalies is characteristic of water . , width=604,height=226 ]	we investigate the equation of state , diffusion coefficient , and structural order of a family of spherically - symmetric potentials consisting of a hard core and a linear repulsive ramp . this generic potential has two characteristic length scales : the hard and soft core diameters . the family of potentials is generated by varying their ratio , @xmath0 . we find negative thermal expansion ( thermodynamic anomaly ) and an increase of the diffusion coefficient upon isothermal compression ( dynamic anomaly ) for @xmath1 . as in water , the regions where these anomalies occur are nested domes in the ( @xmath2 ) or ( @xmath3 ) planes , with the thermodynamic anomaly dome contained entirely within the dynamic anomaly dome . we calculate translational and orientational order parameters ( @xmath4 and @xmath5 ) , and project equilibrium state points onto the ( @xmath6 ) plane , or order map . the order map evolves from water - like behavior to hard - sphere - like behavior upon varying @xmath0 between 4/7 and 6/7 . thus , we traverse the range of liquid behavior encompassed by hard spheres ( @xmath7 ) and water - like ( @xmath8 ) with a family of tunable spherically - symmetric potentials by simply varying the ratio of hard to soft - core diameters . although dynamic and thermodynamic anomalies occur almost across the entire range @xmath9 , water - like structural anomalies ( i.e. , decrease in both @xmath4 and @xmath5 upon compression and strictly correlated @xmath4 and @xmath5 in the anomalous region ) occur only around @xmath10 . water - like anomalies in structure , dynamics and thermodynamics arise solely due to the existence of two length scales , orientation - dependent interactions being absent by design .
this paper is a continuation of our work @xcite . by @xmath1 we denote a semisimple lie algebra and by @xmath9 an arbitrary reductive in @xmath1 subalgebra . in @xcite we introduced the fundamental series of generalized harish - chandra modules ( or equivalently , @xmath0-modules of finite type over @xmath9 ) and proved that any simple generalized harish - chandra module with generic minimal @xmath9-type arises as the socle of an appropriate fundamental series module . using this result we were able to show that any simple generalized harish - chandra module with generic minimal @xmath9-type can be reconstructed from its @xmath10-cohomology . this led to a classification of generalized harish - chandra modules with generic minimal @xmath9-type . in the present paper we study the fundamental series further . after recalling the necessary preliminaries from @xcite , we prove in section 2 that any fundamental series generalized harish - chandra module has finite length . in section 4 we introduce the concepts of a strongly reconstructible and a weakly reconstructible simple generalized harish - chandra module . theorem 3 from @xcite implies that any simple harish - chandra module with generic minimal @xmath9-type is strongly reconstructible ; however our aim is to study strong and weak reconstructibility of simple generalized harish - chandra modules which do not necessarily have a generic minimal @xmath9-type . from section 5 on , we concentrate on the case when the subalgebra @xmath9 of @xmath1 is isomorphic to @xmath6 , i.e. we consider generalized harish - chandra @xmath7-modules . under this assumption we are able to considerably strengthen the results of @xcite and establish new results about strong and weak reconstructibility . in particular , we prove that if @xmath3 is a simple @xmath7-module with minimal @xmath9-type @xmath11 satisfying @xmath12 ( note that @xmath13 as @xmath4 ) , @xmath14 being the maximum and submaximum eigenvalues in @xmath1 of a cartan subalgebra @xmath15 of @xmath5 , then @xmath3 is reconstructible by its @xmath10-cohomology . this yields a classification of simple @xmath7-modules @xmath3 with @xmath12 and proves that all such simple @xmath0-modules have finite type over @xmath9 . for the principal @xmath6-subalgebra the bound @xmath16 is linear in @xmath17 , while the bound established in @xcite is cubic in @xmath17 . in addition , when @xmath9 is a direct summand of a symmetric subalgebra @xmath18 of @xmath1 , we obtain new reconstruction results for harish - chandra modules . in section 7 we prove that for @xmath19 , the socle of the fundamental series module is isomorphic to @xmath20 , where @xmath21 is the simple lowest weight module associated to the data @xmath22 . the relative kazhdan - lusztig theory @xcite yields an explicit formula for the @xmath15-character of @xmath21 . in turn , the theory of the derived zuckerman functors yields an explicit formula for the @xmath9-character of the strongly reconstructible module @xmath20 . section 8 is devoted to examples . we consider six explicit pairs @xmath7 with @xmath23 and we compute the respective sharp bounds on @xmath24 which ensure that a simple @xmath7-module with minimal @xmath6-type @xmath11 is strongly reconstructible . for a principal @xmath6-subalgebra of @xmath25 this sharp bound is the one established in the present paper , in the other five cases it turns out to be the bound from @xcite . in the final section 9 we discuss the possibility that , for @xmath4 and a large enough @xmath26 , the functor @xmath27 is an equivalence between a certain category of @xmath28-finite modules @xmath29 and a category of @xmath0-modules @xmath30 . proving or disproving this statement is an open problem . we conjecture that if @xmath31 , then @xmath27 is an equivalence of categories between @xmath29 and @xmath30 . * acknowledgements . * we thank vera serganova for pointing out to us examples 1 and 2 in subsection 8.7 . both authors acknowledge partial support through dfg grant pe 980/3 - 1(spp 1388 ) . i. penkov acknowledges the hospitality and partial support of yale university , and g. zuckerman acknowledges the hospitality of jacobs university bremen . we start by recalling the setup of @xcite . the ground field is @xmath32 , and if not explicitly stated otherwise , all vector spaces and lie algebras are defined over @xmath32 . the sign @xmath33 denotes tensor product over @xmath32 . the superscript @xmath34 indicates dual space . the sign @xmath35 stands for semidirect sum of lie algebras ( if @xmath36 , @xmath37 is an ideal in @xmath38 and @xmath39 ) . @xmath40 stands for the cohomology of a lie algebra @xmath38 with coefficients in an @xmath38-module @xmath3 , and @xmath41 stands for space of @xmath38-invariants of @xmath3 . by @xmath42 we denote the center of @xmath38 . @xmath43 and @xmath44 denote respectively the exterior and symmetric algebra . if @xmath38 is a lie algebra , then @xmath45 stands for the enveloping algebra of @xmath38 and @xmath46 denotes the center of @xmath45 . we identify @xmath38-modules with @xmath45-modules . it is well known that if @xmath38 is finite dimensional and @xmath3 is a simple @xmath38-module ( or equivalently a simple @xmath45-module ) , @xmath46 acts on @xmath3 via a @xmath46-_character _ , i.e. via an algebra homomorphism @xmath47 . we say that an @xmath38-module @xmath3 is _ generated _ by a subspace @xmath48 if @xmath49 , and we say that @xmath3 is _ cogenerated _ by @xmath48 , if for any non - zero homomorphism @xmath50 , @xmath51 . by @xmath52 we denote the socle ( i.e. the unique maximal semisimple submodule ) of an @xmath38-module @xmath3 ; by @xmath53 we denote the unique maximal semisimple quotient of @xmath3 , when @xmath3 has finite length . if @xmath38 is a lie algebra , @xmath3 is an @xmath38-module , and @xmath54 , we put @xmath55 . by @xmath56 we denote the set @xmath57 . a finite _ multiset _ is a function @xmath58 from a finite set @xmath59 into @xmath60 . a _ submultiset _ of @xmath58 is a multiset @xmath61 defined on the same domain @xmath59 such that @xmath62 for any @xmath63 . for any finite multiset @xmath58 , defined on an additive monoid @xmath59 , we can put @xmath64 . if @xmath65 and @xmath66 , then @xmath3 determines the finite multiset @xmath67 which is the function @xmath68 defined on @xmath69 . let @xmath1 be a finite - dimensional semisimple lie algebra . by @xmath1-mod we denote the category of @xmath1-modules . let @xmath2 be an algebraic subalgebra which is reductive in @xmath1 . we fix a cartan subalgebra @xmath15 of @xmath9 and a cartan subalgebra @xmath70 of @xmath1 such that @xmath71 . by @xmath72 we denote the set of @xmath70-roots of @xmath1 , i.e. @xmath73 . note that , since @xmath9 is reductive in @xmath1 , @xmath1 is a @xmath15-weight module , i.e. @xmath74 . we set @xmath75 . note also that the @xmath76-span of the roots of @xmath70 in @xmath1 fixes a real structure on @xmath77 , whose projection onto @xmath78 is a well - defined real structure on @xmath78 . in what follows , we will denote by @xmath79 the real part of an element @xmath80 . we fix also a borel subalgebra @xmath81 with @xmath82 . then @xmath83 , where @xmath84 is the nilradical of @xmath85 . we set @xmath86 . the quintet @xmath87 will be fixed throughout the paper . by @xmath88 we denote the weyl group of @xmath1 , and by @xmath89 - centralizer in @xmath1 . as usual , we will parametrize the characters of @xmath90 via the harish - chandra homomorphism . more precisely , if @xmath91 is a given borel subalgebra of @xmath1 with @xmath92 ( @xmath91 will be specified below ) , the @xmath90-character corresponding to @xmath93 via the harish - chandra homomorphism defined by @xmath91 will be denoted by @xmath94 ( @xmath95 is the trivial @xmath90-character ) . by @xmath96 we denote the unique @xmath1-invariant symmetric bilinear form on @xmath97 such that @xmath98 for any long root of a simple component of @xmath1 . the form @xmath96 enables us to identify @xmath1 with @xmath97 . then @xmath70 is identified with @xmath77 , and @xmath9 is identified with @xmath99 . we will sometimes consider @xmath96 as a form on @xmath1 . the superscript @xmath100 indicates orthogonal space . note that there is a canonical @xmath9-module decomposition @xmath101 . we also set @xmath102 for any @xmath93 . we say that an element @xmath80 is @xmath0-_regular _ if @xmath103 for all @xmath104 . to any @xmath80 we associate the following parabolic subalgebra @xmath105 of @xmath1 : @xmath106 where @xmath107 . by @xmath108 and @xmath109 we denote respectively the reductive part of @xmath110 ( containing @xmath70 ) and the nilradical of @xmath110 . in particular @xmath111 , and if @xmath15 is @xmath85-dominant , then @xmath112 . we call @xmath105 a _ @xmath15-compatible parabolic subalgebra_. a @xmath15-compatible parabolic subalgebra @xmath113 ( i.e. @xmath114 for some @xmath80 ) is _ minimal _ if it does not properly contain another @xmath15-compatible parabolic subalgebra . it is an important observation that if @xmath113 is minimal , then @xmath115 . in fact , a @xmath15-compatible parabolic subalgebra @xmath110 is minimal if and only if @xmath116 equals the centralizer @xmath117 of @xmath15 in @xmath1 , or equivalently if and only if @xmath114 with @xmath118 @xmath0-regular . in this case @xmath119 . any @xmath15-compatible parabolic subalgebra @xmath114 has a well - defined opposite parabolic subalgebra @xmath120 ; clearly @xmath110 is minimal if and only if @xmath28 is minimal . if @xmath121\cap{\mathfrak{t}})\neq 0 $ ] , then @xmath105 and @xmath9 generate the lie algebra @xmath1 . since re@xmath122\cap{\mathfrak{t}})\neq 0 $ ] , there exists an @xmath6-subalgebra @xmath123 such that re@xmath124 . by definition , @xmath105 contains all @xmath125-singular vectors of @xmath1 . hence @xmath105 generates @xmath1 as a @xmath125-module , i.e. @xmath105 and @xmath125 generate @xmath1 . @xmath9-type _ is by definition a simple finite - dimensional @xmath9-module . by @xmath11 we denote a @xmath9-type with @xmath85-highest weight @xmath24 ( @xmath24 is then @xmath9-integral and @xmath85-dominant ) . let @xmath11 be a @xmath9-type such that @xmath126 is @xmath0-regular , and let @xmath113 be the minimal compatible parabolic subalgebra @xmath127 . put @xmath128 and @xmath129 . clearly @xmath130 . we define @xmath11 to be _ generic _ if the following two conditions hold : 1 . @xmath131 ; 2 . @xmath132 for every submultiset @xmath133 of @xmath134 . it is easy to show that there exists a positive constant @xmath135 depending only on @xmath136 and @xmath110 such that @xmath137 for every @xmath138 implies @xmath139 and that @xmath11 is generic . in agreement with @xcite , we define a @xmath1-module @xmath3 to be a @xmath0-_module _ if @xmath3 is isomorphic as a @xmath9-module to a direct sum of isotypic components of @xmath9-types . if @xmath3 is a @xmath0-module , we write @xmath140 $ ] for the @xmath11-isotypic component of @xmath3 , and we say that @xmath11 is a @xmath9-_type of _ @xmath3 if @xmath140\neq 0 $ ] . we say that a @xmath0-module @xmath3 is _ of finite type _ if @xmath141\neq\infty$ ] for every @xmath9-type @xmath11 . we will also refer to @xmath0-modules of finite type as _ generalized harish - chandra modules_. let @xmath142 be the discrete subgroup of @xmath143 generated by @xmath144 . by @xmath145 we denote the class of @xmath0-modules @xmath3 for which there exists a finite subset @xmath146 such that @xmath147 . if @xmath3 is a module in @xmath145 , a @xmath9-type @xmath11 of @xmath3 is _ minimal _ if the function @xmath148 defined on the set @xmath149\neq 0\}$ ] has a minimum at @xmath24 . any non - zero @xmath0-module @xmath3 in @xmath145 has a minimal @xmath9-type . this follows from the fact that the squared length of a vector has a minimum on every shifted lattice in euclidean space . recall that the _ functor of @xmath9-locally finite vectors _ @xmath150 is a well - defined left exact functor on the category of @xmath151-modules with values in @xmath0-modules , @xmath152 by @xmath153 we denote as usual the total right derived functor of @xmath150 , see @xcite and the references therein . if @xmath3 is a @xmath0-module of finite type , then @xmath154 is a well - defined @xmath0-module of finite type and @xmath155 is an involution on the category of @xmath0-modules of finite type . we put @xmath156 . there is an obvious @xmath1-invariant non - degenerate pairing @xmath157 . [ lemma1.2 ] let @xmath88 be a finite - dimensional @xmath1-module and @xmath3 be a finite length @xmath0-module of finite type over @xmath9 . then \a ) @xmath158 is a @xmath0-module of finite type . \b ) @xmath158 is a @xmath1-module of finite length . \a ) since @xmath9 is finite dimensional and reductive in @xmath1 , the class of @xmath0-modules is closed under tensor products . let @xmath11 be a @xmath9-type . since @xmath88 is finite dimensional , @xmath159 , which is finite dimensional , since @xmath160 is finite dimensional and @xmath3 has finite type over @xmath9 . \b ) since @xmath3 has finite length , @xmath3 is finitely generated over @xmath1 . note that @xmath161 , the @xmath9-finite dual of @xmath3 is a @xmath0-module of finite length and hence @xmath161 is finitely generated over @xmath1 and likewise @xmath162 is finitely generated . hence , @xmath163 is finitely generated , and satisfies the ascending chain condition . we have already seen that @xmath158 is finitely generated ; thus @xmath158 satisfies the ascending chain condition . we conclude that @xmath158 has finite length . we also introduce the following notation : if @xmath164 is a subalgebra of @xmath1 and @xmath165 is a @xmath164-module , we set @xmath166 and @xmath167 . for a finite - dimensional @xmath110- or @xmath28-module @xmath168 we set @xmath169 , @xmath170 . note that both @xmath171 and @xmath172 have simple socles , as long as @xmath168 itself is simple . the _ fundamental series _ of @xmath0-modules of finite type @xmath173 is defined as follows . let @xmath113 be a minimal compatible parabolic subalgebra , @xmath168 be a simple finite dimensional @xmath110-module on which @xmath15 acts via the weight @xmath174 , and @xmath175 where @xmath176 . set @xmath177 then the following assertions hold under the assumptions that @xmath178 and that @xmath24 is @xmath85-dominant , @xmath9-integral and yields a generic @xmath9-type @xmath11 ( theorem 2 of @xcite ) . a. @xmath173 is a @xmath0-module of finite type in the class @xmath145 , and @xmath90 acts on @xmath173 via the @xmath90-character @xmath179 where @xmath180 for some fixed borel subalgebra @xmath91 of @xmath1 with @xmath181 and @xmath182 , and where @xmath183 is the @xmath91-highest weight of @xmath168 ( note that @xmath184 ) . b. @xmath185 for @xmath186 . c. there is a @xmath9-module isomorphism @xmath187\cong{\mathbb{c}}^{\dim e}\otimes v(\mu),\ ] ] and @xmath11 is the unique minimal @xmath9-type of @xmath188 . d. let @xmath189 be the @xmath1-submodule of @xmath188 generated by @xmath190 $ ] . then @xmath189 is the unique simple submodule of @xmath188 , and moreover , @xmath188 is cogenerated by @xmath190 $ ] . this implies that @xmath191 is generated by @xmath192 $ ] , where @xmath193 is the element of maximal length in the weyl group @xmath194 of @xmath9 . e. for any non - zero @xmath1-submodule @xmath3 of @xmath188 there is an isomorphism of @xmath116-modules @xmath195 in the rest of the paper , @xmath110 is a minimal @xmath15-compatible parabolic subalgebra and @xmath168 is a simple finite - dimensional @xmath110-module . then @xmath196 and @xmath168 is a simple @xmath197-module . fix a borel subalgebra @xmath198 in @xmath116 such that @xmath199 . write @xmath200 ; then @xmath91 is a borel subalgebra of @xmath1 . set @xmath201 . [ th2.1 ] assume that @xmath178 and that @xmath24 is generic . assume in addition that @xmath171 is a simple @xmath1-module . then @xmath188 is a simple ( in particular , non - zero ) @xmath1-module . by the duality theorem from @xcite , @xmath202 for any @xmath151-module @xmath203 of finite type over @xmath15 . set @xmath204 . then @xmath203 is a @xmath151-module of finite type over @xmath15 ( see for instance @xcite ) , and ( [ eqn1 ] ) yields for @xmath205 @xmath206 we have @xmath207 thus @xmath208 moreover , @xmath171 has finite type over @xmath15 . hence , @xmath209 . by frobenius reciprocity , there is a canonical @xmath1-module homomorphism @xmath210 whose restriction to @xmath211 is the identity . as @xmath212 is simple by our assumption , @xmath213 is injective . moreover , @xmath213 must be surjective as the @xmath15-characters of @xmath212 and @xmath172 are equal . therefore there is a commutative diagram of isomorphisms @xmath214{cccc } & r^s\gamma_{{\mathfrak{k}},{\mathfrak{t}}}(n_{\mathfrak{p}}(e)^_{\mathfrak{t } } ) & \stackrel{\sim}{\rightarrow } & r^s\gamma_{{\mathfrak{k}},{\mathfrak{t}}}(n_{\bar{{\mathfrak{p}}}}(e^))\\ & \wr\downarrow&&\wr\downarrow\\ \gamma : & f^s({\mathfrak{p}},e)_{\mathfrak{k}}^ * & \stackrel{\sim}{\rightarrow}&f^s(\bar{{\mathfrak{p}}},e^ * ) . \end{array}\ ] ] the fact that @xmath11 is generic for @xmath110 implies immediately that @xmath215 is generic for @xmath28 . thus @xmath188 is cogenerated by its minimal @xmath9-isotypic component @xmath190 $ ] , and @xmath216 cogenerated by its minimal @xmath9-isotypic component . on the other hand , the isomorphism @xmath217 implies that @xmath216 is also generated by its minimal @xmath9-isotypic component as @xmath218 is generated by its minimal @xmath9-isotypic component . we conclude that @xmath219 is simple , which in turn shows that @xmath220 is simple . assume that the @xmath91-highest weight of @xmath168 is @xmath221 . set @xmath222 and @xmath223 . [ cor2.3 ] let @xmath224 be @xmath91-dominant , i.e. @xmath225 for any root @xmath217 of @xmath70 in @xmath91 . then , under the assumption that @xmath24 is generic and that @xmath178 , we have @xmath185 for @xmath226 . moreover @xmath188 is simple . thus , @xmath227 has finite length for all @xmath228 . under the hypothesis on @xmath183 , @xmath229 is simple , hence @xmath171 is simple , and the statement follows from theorem [ th2.1 ] in the rest of this section , @xmath110 is an arbitrary minimal @xmath15-compatible parabolic subalgebra . [ le2.4 ] for any @xmath230 , there exists a @xmath91-dominant integral weight @xmath231 of @xmath1 such that @xmath232 for every weight @xmath233 of @xmath15 in @xmath10 . since @xmath110 is @xmath15-compatible , there exists @xmath234 such that @xmath235 for every root @xmath217 of @xmath70 in @xmath10 . regard @xmath78 as a subspace of @xmath77 via the killing form of @xmath1 restricted to @xmath70 . then , @xmath236 for every root @xmath217 of @xmath70 in @xmath10 and @xmath237 for every root @xmath217 of @xmath70 in @xmath116 . hence , @xmath238 is a dominant integral weight of @xmath70 in @xmath1 . finally , choose a positive integer @xmath239 such that @xmath240 for every weight @xmath233 of @xmath15 in @xmath10 . then @xmath241 is a @xmath91-dominant weight of @xmath1 as required . [ pr2.5 ] suppose that @xmath224 is @xmath91-dominant . then , @xmath185 for @xmath226 , and @xmath188 has finite length . fix a constant @xmath242 . chose @xmath243 such that @xmath244 for every @xmath138 . set @xmath245 , where @xmath246 is defined in lemma [ le2.4 ] . then @xmath247 for every @xmath138 , and by possibly making @xmath135 larger , we can assume that @xmath248 is generic . in particular , @xmath249 . next , let @xmath250 , and @xmath251 be a simple finite - dimensional @xmath116-module with highest weight @xmath252 . set @xmath253 . then , by corollary [ cor2.3 ] , @xmath254 for @xmath226 , and @xmath255 is a simple @xmath1-module . furthermore , by propositions 2.6 and 2.12 in @xcite , @xmath227 is a direct summand of @xmath256 where @xmath257 stands for the finite - dimensional @xmath1-module with @xmath91-highest weight @xmath258 . lemma [ lemma1.2 ] implies the statement . * by a more refined argument with translation functors @xcite one can show using the result of @xcite that @xmath188 is simple and hence non - zero , while @xmath259 if @xmath226 . [ th2.6 ] the @xmath0-module @xmath227 has finite length for any simple @xmath110-module @xmath168 and any @xmath260 . we will assume at first that @xmath224 is a regular weight of @xmath70 in @xmath1 . then , there exists a unique element @xmath261 such that @xmath262 is dominant for @xmath70 in @xmath1 . denote by @xmath263 the length @xmath264 . we will argue by induction on @xmath263 . the theorem is true for @xmath265 by proposition [ pr2.5 ] . suppose we assume the theorem for @xmath266 , @xmath26 being fixed . if @xmath267 , we can choose a root @xmath217 of @xmath70 in @xmath1 such that @xmath268 . let @xmath59 have highest weight @xmath269 . we will show that the finiteness of the length of @xmath270 for all @xmath271 implies the finiteness of the length of @xmath227 for all @xmath271 . case i : @xmath272 . choose a translation functor @xmath273 so that @xmath274 has a central character which is singular with respect to precisely @xmath217 . let @xmath275 be the translation functor adjoint to @xmath273 . by highest weight module theory we have a short exact sequence @xmath276 this short exact sequence yields a long exact sequence @xmath277 we can rewrite this long exact sequence as @xmath278 by assumption , @xmath270 and @xmath279 have finite length . hence , by lemma [ lemma1.2 ] @xmath280 has finite length . by the long exact sequence , @xmath281 has finite length . case ii : @xmath282 . choose an integral weight @xmath283 such that if @xmath284 is dominant , @xmath285 is dominant . let @xmath59 be a finite - dimensional simple @xmath116-module such that the highest weight of @xmath59 is @xmath286 . there exists a translation functor @xmath273 such that @xmath287 . for the adjoint functor @xmath275 , @xmath288 . then @xmath289 . hence , @xmath290 , and by lemma [ lemma1.2 ] the finiteness of the length of @xmath291 implies the same for @xmath281 . let @xmath292 be a finite - dimensional @xmath293-module . then @xmath294 has finite length for all @xmath271 . induction on the length of @xmath292 as a @xmath110-module : if @xmath292 has length 1 , then @xmath295 , and we are back to theorem [ th2.6 ] . in what follows we denote by @xmath296 the full subcategory of @xmath1-modules consisting of finitely generated @xmath151-modules which are locally @xmath28-finite . [ le2.8 ] let @xmath297 . then @xmath298 has finite length , each simple constituent of @xmath298 is isomorphic to @xmath299 for some simple finite - dimensional @xmath116-module @xmath59 , and @xmath298 has finite type over @xmath15 . let @xmath300 generate @xmath298 over @xmath301 and let @xmath302 . then @xmath303 is a finite - dimensional @xmath110-module . moreover , @xmath298 is a quotient of @xmath304 , for which the lemma is well known . [ pr2.9 ] let @xmath297 . then @xmath305 has finite length for all @xmath271 . by lemma [ le2.8 ] , @xmath306 . any module in @xmath307 admits a resolution by modules of the form @xmath308 , where each @xmath309 is a finite - dimensional @xmath293-module : @xmath310 . by considering the @xmath15-finite dual of this resolution , we obtain a resolution of @xmath311 by modules of the form @xmath312 , where each @xmath313 is a finite - dimensional @xmath293-module . write this resolution as @xmath314 . we have a convergent spectral sequence of @xmath1-modules with @xmath315 , and which abuts to @xmath316 . if we fix @xmath260 , there are only finitely many terms @xmath317 with @xmath318 , since both @xmath319 and @xmath320 . hence , @xmath321 has finite length . finally , @xmath305 has finite length by @xmath322 we denote the full subcategory of @xmath1-mod consisting of @xmath0-modules which have finite type over @xmath9 and have finite length over @xmath1 . if @xmath297 and @xmath228 , then @xmath323 . the statement follows from proposition [ pr2.9 ] and from the `` finiteness statement '' of theorem 2.4 c ) in @xcite . if @xmath324 is a full abelian subcategory of @xmath1-mod , let @xmath325 be the grothendieck group of @xmath324 . if @xmath297 , let @xmath326 $ ] in @xmath327 . the fact that @xmath328 is well - defined follows from the vanishing statement of theorem 2.4 b ) in @xcite . the map @xmath329 yields a non - zero homomorphism @xmath330 . this is a well - known fact which follows from the long exact sequence of cohomology . * example. \a ) if @xmath168 is a finite - dimensional simple @xmath116-module with highest weight @xmath183 such that @xmath224 is regular and @xmath1-dominant , then @xmath331 $ ] . if @xmath24 is @xmath85-dominant and @xmath9-integral , then @xmath332 by the remark after proposition [ pr2.5 ] . \b ) @xmath333 $ ] , where @xmath334 is the weyl group of @xmath9 . indeed , in the proof of theorem 2.4 in @xcite it is shown that @xmath335 for any simple finite - dimensional @xmath9-module @xmath336 . since @xmath337 for @xmath338 , we conclude that @xmath339 . moreover , @xmath340 , where @xmath341 stands for the relative lie algebra cohomology . it is well - known that @xmath341 is the cohomology of the variety @xmath342 , @xmath343 being a connected affine real algebraic group with lie algebra @xmath9 and @xmath344 being a torus in @xmath343 with lie @xmath345 . moreover , @xmath342 is homeomorphic to the flag variety of @xmath346 and hence the euler characteristic of @xmath342 is @xmath347 , by the bruhat decomposition of the flag variety . thus , @xmath348 we start by recalling [ pz2 , proposition 1 ] in the case when @xmath9 is semisimple . in what follows , @xmath3 will denote a @xmath0-module . note that @xmath3 is automatically in the class @xmath145 . in the category of @xmath15-weight modules , there exists a bounded ( not necessarily first quadrant ) cohomology spectral sequence which converges to @xmath349 , with @xmath350 where @xmath351 runs over @xmath352 for some @xmath353 is a monotonic function on @xmath352 with values in @xmath354 such that @xmath355 and @xmath356 , @xmath357 is a @xmath15-submodule of @xmath358 for every @xmath351 , and @xmath359 . we also have @xmath360 . suppose we are interested in @xmath361 for a fixed @xmath362 . write @xmath363 then @xmath364 . [ le1 ] fix @xmath365 and @xmath362 , @xmath366 . assume that @xmath367 . then @xmath368 this follows immediately from the definition of a convergent spectral sequence of vector spaces . as a special case we have the following lemma . recall that @xmath369 , @xmath370 . [ le3.5 ] if @xmath371 for @xmath372 , then @xmath373 . the isomorphism ( [ eqn3 ] ) implies @xmath374 . as a consequence we have the following . if @xmath375 is @xmath9-dominant integral , then @xmath373 . kostant s theorem @xcite implies that if @xmath376 is @xmath9-dominant integral , @xmath377 has pure weight @xmath378 , where @xmath379 is the longest element of @xmath194 . we have as a consequence that @xmath380 if @xmath381 is @xmath9-dominant integral . since @xmath3 is a direct sum of simple finite dimensional @xmath9-modules @xmath382 for various @xmath376 , we conclude that @xmath371 . ( we have used our assumption that @xmath383 ) . then @xmath384 by lemma [ le3.5 ] . we now recall that @xmath349 is an @xmath385-module . this is established in [ v2 , ch . 5 ] in the case when @xmath9 is a symmetric subalgebra but the argument extends to the case of a general reductive in @xmath1 subalgebra @xmath9 . note that @xmath386 . the following statement is identical to [ pz2 , corollary 3 ] . [ pr3.5 ] a ) if @xmath3 is a @xmath0-module of finite type , then @xmath349 is an @xmath387-module of finite type . moreover , if @xmath3 is @xmath90-finite ( i.e. the action of @xmath90 on @xmath3 factors through a finite - dimensional quotient of @xmath90 ) then @xmath349 is @xmath388-finite . \b ) if @xmath110 is a minimal compatible parabolic subalgebra and @xmath3 is a @xmath0-module of finite type which is in addition @xmath90-finite , then @xmath349 is finite dimensional . suppose @xmath3 is simple . let @xmath11 be a minimal @xmath9-type of @xmath3 ( a priori @xmath11 is not unique ) . suppose @xmath126 is @xmath0-regular and @xmath178 . the pair @xmath389 as above is _ strongly reconstructible _ if @xmath390 is a simple @xmath116-module and there is an isomorphism of @xmath1-modules @xmath391 the isomorphism ( [ eqn4 ] ) implies via theorem 2 of @xcite that @xmath11 is the unique minimal @xmath9-type of m. moreover , @xmath392 . therefore , if the pair @xmath389 is strongly reconstructible , @xmath24 is determined by @xmath3 as the highest weight of the unique minimal @xmath9-type of @xmath3 . this allows us to simply speak of strongly reconstructible simple @xmath0-modules rather than of strongly reconstructible pairs . the first reconstruction theorem of @xcite now implies the following . if @xmath3 is a simple @xmath0-module of finite type with a generic minimal @xmath9-type @xmath11 , then @xmath3 is strongly reconstructible . below we will see ( in particular in subsection 8.2 ) that the converse to the above theorem is false . we will also see ( in subsection 8.7 ) examples of simple finite - dimensional @xmath116-modules @xmath168 such that @xmath188 has a reducible socle . a simple @xmath0-module @xmath3 of finite type over @xmath9 is weakly reconstructible if for some minimal @xmath9-compatible parabolic subalgebra @xmath110 , there exists an injective homomorphism of @xmath1-modules @xmath393 let @xmath394 be simple with central characters @xmath395 and @xmath396 respectively . assume @xmath397 and @xmath398 are dominant regular with respect to a borel subalgebra @xmath399 ; assume further that @xmath400 is dominant integral . finally assume that @xmath275 is a translation functor such that @xmath401 . then , @xmath402 is weakly reconstructible if and only if @xmath403 is weakly reconstructible . assume that @xmath403 is weakly reconstructible . then for some simple quotient @xmath251 of @xmath404 , we have an injection of @xmath1-modules @xmath405 . by assumption , @xmath275 is an equivalence of categories . hence we have an injection @xmath406 . by @xcite , we have an isomorphism @xmath407 for a simple @xmath116-module @xmath408 . moreover , we have a translation functor @xmath409 such that @xmath410 . thus , we have an injection @xmath411 . now let @xmath412 be the central character of @xmath413 for @xmath414 . let @xmath415 and @xmath416 be the respective projection functors . by assumption we have a surjection of @xmath116-modules @xmath417 . apply the translation functor @xmath409 to @xmath418 to obtain a surjection @xmath419 . by [ kv , ch . 7 ] , @xmath420 . thus , @xmath408 is a quotient of @xmath421 . finally , the injection @xmath422 yields an injection @xmath423 . hence @xmath402 is weakly reconstructible . from now on we assume that @xmath9 is isomorphic to @xmath424 . we fix a standard basis @xmath425 for @xmath9 ; the eigenvalues of @xmath426 in @xmath1 will be integers . let @xmath427 be the cartan subalgebra of @xmath9 generated by @xmath428 , and let @xmath429 where @xmath430 , @xmath431 . the subalgebra @xmath110 is automatically a minimal @xmath15-compatible parabolic subalgebra . for @xmath4 , our lemma [ le1 ] simplifies to the following . [ le2 ] fix @xmath432 and @xmath362 , @xmath433 . write @xmath434 suppose that @xmath367 . then there is an isomorphism of @xmath15-modules @xmath435 in particular , let @xmath436 and assume @xmath375 is dominant integral for @xmath9 . then @xmath437 implies @xmath438 more generally , @xmath439 dominant integral implies @xmath440 from now on we identify integral weights @xmath441 of @xmath15 with the corresponding integers , @xmath442 . let @xmath397 and @xmath398 be the maximum and submaximum weights of @xmath15 in @xmath443 ( we consider @xmath397 and @xmath398 as integers ) ; if @xmath397 has multiplicity at least two in @xmath443 , then @xmath444 . [ pr5.2 ] let @xmath24 be a nonnegative integer and let @xmath3 be a @xmath0-module with the property that @xmath445 implies @xmath446=0 $ ] . \a ) if @xmath447 , then @xmath448 \b ) if @xmath449 , then @xmath450 \a ) our hypothesis on @xmath3 implies that if @xmath446\neq 0 $ ] , then @xmath451 . since @xmath452 has weights @xmath453 with @xmath454 as above , we see that @xmath455 . hence ( 2 ) implies the inequality in a ) . \b ) it suffices to show that @xmath456 . then the statement from ( [ eqn5 ] ) , taking into account the vanishing of @xmath457 , implies ( b ) . we now check that @xmath456 . we have @xmath458 furthermore , @xmath459 @xmath460 implies @xmath461 as the weight of @xmath462 equals @xmath463 . the first term of ( [ eqn6 ] ) vanishes as the @xmath15-weights of @xmath464 are strictly positive and the smallest @xmath15-weight of @xmath465 is @xmath24 . the maximal @xmath15-weight of the second term is @xmath466 , hence the inequality @xmath467 implies the vanishing of the second term of ( [ eqn6 ] ) . our next task is to state and prove a vanishing theorem for @xmath468 , where @xmath168 is a simple finite dimensional @xmath116-module . let @xmath174 be the weight of @xmath15 in @xmath168 . [ pr5.3 ] suppose @xmath469 and @xmath470 . then @xmath471 . by definition @xmath472 we have @xmath473 and the @xmath91-highest weight of @xmath474 equals @xmath475 . on the other hand , @xmath476 for any @xmath91-dominant weight . this follows from the fact that any @xmath91-dominant weight is a non - negative linear combination of roots of @xmath91 ( see for instance @xcite , p. 686 ) . the @xmath1-module @xmath171 has a finite - dimensional submodule if and only if @xmath212 has a finite - dimensional quotient . note that @xmath477 is an integrable @xmath1-module as @xmath171 is @xmath28-locally finite and @xmath9 and @xmath28 generate @xmath1 . therefore , @xmath478 whenever @xmath171 has no finite - dimensional submodule , i.e. whenever @xmath479 is not @xmath91-dominant . the fact that @xmath480 , allows us to conclude that @xmath481 is not @xmath91-dominant , i.e. that @xmath478 . [ pr5.4 ] @xmath482 . the statement is a direct corollary of proposition 3 a ) in @xcite . note that this proof does not use genericity . [ pr5.5](cohomological frobenius reciprocity ) let @xmath470 . if @xmath3 is a @xmath0-module such that @xmath349 is finite dimensional , then we have a natural isomorphism @xmath483 this follows from the existence of a ( not necessarily first quadrant ) spectral sequence with @xmath408-term @xmath484 converging to @xmath485 , see proposition 6 of @xcite . by assumption @xmath349 is finite dimensional . choose @xmath486 to be the least possible integer with @xmath487 . by the same argument as in the proof of theorem 2 , b ) in @xcite , we conclude that @xmath488 . thus , @xmath489 and @xmath490 for @xmath491 . consequently , @xmath492 and we deduce that @xmath493 . hence , @xmath494 and the spectral sequence is a first quadrant spectral sequence , with corner isomorphism @xmath495 . [ cor5.6 ] suppose @xmath13 . \a ) let @xmath203 be any @xmath1-submodule of @xmath496 . then @xmath168 is a quotient of @xmath497 . in particular , if @xmath203 is simple , then @xmath203 is weakly reconstructible . \b ) let @xmath3 be a simple @xmath0-module such that @xmath349 is finite dimensional and @xmath168 is isomorphic to a quotient of @xmath498 . then @xmath3 is isomorphic to a submodule of @xmath499 . in particular @xmath3 is weakly reconstructible and @xmath3 has finite type over @xmath9 . fix a central character @xmath500 . the set of isomorphism classes of simple @xmath0-modules @xmath3 with central character @xmath500 such that @xmath501 and @xmath502 for some @xmath503 is finite . a @xmath1-module @xmath3 as in the corollary is isomorphic by corollary [ cor5.6 ] ( b ) to a @xmath1-submodule of @xmath504 , where @xmath505 runs over finitely many simple finite - dimensional @xmath110-modules . since @xmath504 has finite length for each @xmath505 by theorem [ th2.6 ] , the statement follows . [ th5.8 ] suppose @xmath9 is regular in @xmath1 . let @xmath3 be a simple @xmath0-module , not necessarily of finite type over @xmath9 , with lowest @xmath9-type @xmath11 for @xmath470 . \a ) if @xmath506 , there exists a 1-dimensional simple quotient @xmath168 of @xmath498 . for any such @xmath168 we have an injective homomorphism @xmath507 . hence @xmath3 is weakly reconstructible and @xmath3 is of finite type over @xmath9 . \b ) if @xmath3 is of infinite type over @xmath9 , then @xmath508 . \a ) since @xmath3 is simple , proposition [ pr3.5 ] b ) implies that @xmath349 is finite dimensional . note that the regularity of @xmath9 in @xmath1 implies that @xmath116 is a cartan subalgebra of @xmath1 . hence there exists a 1-dimensional simple @xmath116-quotient @xmath168 of @xmath498 . proposition [ pr5.5 ] implies now that any such @xmath168 induces an injective homomorphism of @xmath3 into @xmath509 . in particular , @xmath3 is weakly reconstructible and is of finite type over @xmath9 . \b ) follows from a ) . in the rest of the paper @xmath510 are as in section 1 . we start with the following result on the of @xmath0-modules . [ th6.1 ] let @xmath19 . then @xmath511 and @xmath512 is simple . in particular , @xmath513 . let @xmath203 be a non - zero submodule of @xmath496 . since @xmath496 is a @xmath0-module of finite type and is @xmath514-finite , proposition [ pr3.5 ] b ) and proposition [ pr5.5 ] apply to @xmath203 , yielding a surjective homomorphism of @xmath116-modules @xmath515 . hence @xmath516 . next , by @xcite , @xmath517 therefore @xmath518 by the left exactness of @xmath519 . finally , by proposition [ pr5.2 ] a ) , @xmath520 . combining these inequalities we see that @xmath521 hence @xmath522=f^1({\mathfrak{p}},e ) [ \mu ] $ ] , or equivalently @xmath523 . since in this way @xmath512 is contained in any non - zero submodule of @xmath496 , @xmath512 is simple and @xmath524 . [ cor6.2 ] under the assumptions of theorem [ th6.1 ] , let @xmath525 be a @xmath1-submodule of @xmath496 . then the minimal @xmath9-type of @xmath203 is @xmath11 , @xmath526 , and there is an isomorphism of @xmath116-modules @xmath527 . the statement was established in the proof of theorem [ th6.1 ] . [ cor6.3 ] let @xmath3 be a simple @xmath0-module whose minimal @xmath9-type @xmath11 satisfies @xmath19 . then , if @xmath349 is finite - dimensional and @xmath506 , @xmath3 is strongly reconstructible . let @xmath505 be a simple quotient of the @xmath116-module @xmath498 . by proposition [ pr5.5 ] , @xmath3 is a simple submodule of @xmath504 , hence by theorem [ th6.1 ] @xmath528 [ cor6.4 ] let @xmath3 be a simple @xmath0-module of finite type such that its minimal @xmath9-type @xmath11 satisfies @xmath529 . then @xmath349 is finite - dimensional , and @xmath506 , hence @xmath3 is strongly reconstructible by corollary [ cor6.3 ] . the statement follows from corollary [ cor6.4 ] via proposition [ pr5.2 ] b ) . [ cor6.5 ] the correspondences @xmath530 @xmath531 induce mutually inverse bijections between the set of isomorphism classes of simple @xmath0-modules of finite type @xmath3 whose minimal @xmath9-type @xmath11 satisfies @xmath529 and the set of isomorphism classes of finite dimensional @xmath116-modules on which @xmath15 acts via @xmath532 , where @xmath533 . [ cor6.6 ] suppose @xmath9 is regular in @xmath1 ( i.e. let @xmath15 contain an element regular in @xmath1 ) . suppose @xmath3 is a simple @xmath0-module ( not necessarily of finite type over @xmath9 ) with lowest @xmath9-type @xmath11 . if @xmath529 , then @xmath534 is a 1-dimensional @xmath116-module and @xmath535 . thus @xmath3 is strongly reconstructible . in particular , @xmath3 has finite type over @xmath9 . we apply theorem [ th5.8 ] a ) and proposition [ pr5.2 ] b ) to conclude that , for any 1-dimensional quotient @xmath168 of @xmath498 , we have an injection @xmath507 . since @xmath529 , there are isomorphisms @xmath536 and @xmath537 ( the latter is an isomorphism of @xmath116-modules ) . example . * let @xmath1 be a classical simple lie algebra of rank @xmath26 and @xmath538 be a principal subalgebra . then the claim of corollary [ cor6.5 ] is proved in @xcite under the assumption that @xmath539 , where @xmath540 , @xmath541 being the simple roots of @xmath91 . it is well - known that @xmath542 grows cubically with the growth of @xmath26 , while the value @xmath16 has linear growth in @xmath26 . therefore , for large @xmath26 , the result of corollary [ cor6.5 ] strengthens considerably theorem 3 of @xcite for @xmath9 being a principal @xmath6-subalgebra . on the other hand , we will see in section 8 that for @xmath543 , there are cases where the bound @xmath544 is lower that @xmath16 . set now @xmath545 and note that @xmath18 is a reductive in @xmath1 subalgebra . recall that @xmath546 . moreover , @xmath547 . if @xmath548 , then for any simple @xmath549-module @xmath3 of finite type over @xmath18 , @xmath349 is finite dimensional . by proposition [ pr3.5 ] a ) , @xmath349 is an @xmath550-module of finite type as @xmath110 is @xmath18-compatible ( see also [ v2 , corollary 5.2.4 ] ) . but @xmath551 . hence @xmath349 is an integrable @xmath116-module . finally , [ pz2 , corollary 3 a ) ] implies now that @xmath349 is finite dimensional . we conclude this section with some applications to the case when @xmath18 is a symmetric subalgebra . [ pr6.8 ] assume that @xmath1 is simple and @xmath18 is symmetric . \a ) if @xmath1 is classical with @xmath552 , the only case of a symmetric pair of the form @xmath549 for which @xmath553 is not equal to @xmath554 is the series @xmath555 , where @xmath556 . \b ) if @xmath1 is exceptional , then @xmath548 . in fact , @xmath18 is symmetric if and only if @xmath9 is conjugate to the @xmath6-subalgebra of a highest root of @xmath1 . follows from the classification of symmetric pairs . if @xmath557 and @xmath18 is symmetric , then @xmath548 . follows from proposition [ pr6.8 ] , but a more elegant proof is based on borel - de siebenthal @xcite . assume that @xmath9 is symmetric and @xmath548 . let @xmath3 be a simple @xmath549-module . \a ) if @xmath558 , then @xmath3 is strongly reconstructible as a @xmath0-module ; in particular @xmath3 has finite type over @xmath9 . \b ) if @xmath12 , then @xmath3 is strongly reconstructible as a @xmath0-module ; in particular @xmath3 has finite type over @xmath9 . assume @xmath559 . set @xmath560 and recall that @xmath21 is simple . also note that @xmath171 and @xmath21 are objects of @xmath296 . denote by @xmath59 a variable simple finite - dimensional @xmath110-module on which @xmath15 acts via @xmath561 . non - negative integers @xmath562 are determined from the equality @xmath563 = \sum m(e , d)[l_{\mathfrak{p}}(d)]$ ] in the grothendieck group @xmath564 . we arrange the integers @xmath562 into a matrix @xmath565 with rows indexed by all possible @xmath168 and columns indexed by all possible @xmath59 ; the rows and columns of @xmath565 are finitary , i.e. each row and each column have finitely many non - zero entries . the algorithm for computing the integers @xmath562 is discussed in @xcite . [ le7.1 ] suppose @xmath59 is not isomorphic to @xmath168 . then @xmath566 implies @xmath567 . we claim that the minimum @xmath15-weight of @xmath171 is @xmath568 . to see this it suffices to note that @xmath569 , as the maximum weight of @xmath212 is @xmath570 ( note that @xmath571 ) . thus , if @xmath572 is a composition factor of @xmath171 and @xmath573 , we have @xmath574 , or equivalently , @xmath575 . [ pr7.2 ] let @xmath470 . then @xmath576 = \sum_d m(e , d ) [ r^1 \gamma_{{\mathfrak{k}},{\mathfrak{t } } } ( l_{\mathfrak{p}}(d))]$ ] . taking into account that @xmath577 is a homomorphism , it suffices to prove : \a ) @xmath578 for @xmath579 and @xmath470 ; \b ) @xmath580 for @xmath581 . part a ) follows from proposition [ pr5.3 ] and proposition [ pr5.4 ] . to prove part b ) , note that @xmath566 implies @xmath582 by lemma [ le7.1 ] . then @xmath583 implies @xmath584 as @xmath585 . to see that @xmath586 , note that , by the duality theorem in @xcite , @xmath587 the @xmath1-module @xmath588 is @xmath110-locally finite , simple and infinite dimensional . as @xmath110 and @xmath9 generate @xmath1 , @xmath589 would imply @xmath590 . as the latter is false , @xmath591 . part b ) is proved . [ pr7.3 ] suppose @xmath470 . \a ) then @xmath592 , and the lowest @xmath9-type of @xmath20 is @xmath11 of multiplicity @xmath593 . \b ) we have the following inclusions of @xmath0-modules : @xmath594 \c ) if @xmath171 is reducible then the submodule @xmath20 of @xmath496 is proper and non - zero , and hence @xmath496 is reducible . \a ) by proposition [ pr7.2 ] , we have under our hypothesis @xmath595=\sum_d r^1\gamma_{{\mathfrak{k}},{\mathfrak{t}}}(l_{\mathfrak{p}}(d))[\mu].\ ] ] by theorem 2 from @xcite , @xmath596\cong(\dim e)v(\mu)$ ] . next , apply proposition [ pr7.2 ] to @xmath59 . we see that @xmath597 $ ] is a summand of @xmath598 $ ] . thus , @xmath599)\leq\dim ( r^1\gamma_{{\mathfrak{k}},{\mathfrak{t}}}(n_{\mathfrak{p}}(d))[\mu])$ ] . assume now that @xmath566 and @xmath600 . then @xmath575 by lemma [ le7.1 ] . applying theorem 2 from @xcite a second time , we have @xmath601=0 $ ] . we conclude that if @xmath566 and @xmath573 , @xmath602=0 $ ] . so , from formula ( [ eqn7 ] ) above , we deduce that @xmath603=r^1\gamma_{{\mathfrak{k}},{\mathfrak{t}}}(l_{\mathfrak{p}}(e))[\mu].\ ] ] this proves part a ) . \b ) by the vanishing theorems a ) and b ) in the proof of proposition [ pr7.2 ] , the inclusion of @xmath21 into @xmath171 yields an injection of @xmath20 into @xmath496 ; part b ) follows immediately from ( [ eqn8 ] ) and the definition of @xmath512 . \c ) if @xmath171 is reducible then for some @xmath59 with @xmath604 , @xmath605 . so , part c ) follows from proposition [ pr7.2 ] and proposition [ pr7.3 ] a ) . [ conj7.4 ] if @xmath470 , then @xmath20 is a semisimple @xmath1-module . if true , this conjecture would imply that all simple constituents of @xmath20 are weakly reconstructible for @xmath470 . this would follow from corollary [ cor5.6 ] a ) . see subsection 8.7 , examples 1 and 2 for cases when @xmath20 is reducible . [ th7.4 ] assume @xmath606 . then @xmath20 is a simple ( in particular , non - zero ) submodule of @xmath496 and @xmath607 . by proposition [ pr7.3 ] , we have @xmath608 . by the duality theorem , @xmath609 . let @xmath610 be the submodule of @xmath611 consisting of vectors which are orthogonal to @xmath20 via the above duality . by proposition [ pr7.3 ] , @xmath612=r^1\gamma_{{\mathfrak{k}},{\mathfrak{t}}}(l_{\mathfrak{p}}(e))[\mu]$ ] . hence , @xmath613=0 $ ] . by construction , @xmath610 is a submodule of @xmath614 . it follows from the proof of corollary [ cor6.2 ] that @xmath615 and hence @xmath616 . the statement of theorem [ th6.1 ] implies the remainder of the proof of theorem [ th7.4 ] . [ cor7.5 ] if @xmath617 , then : \a ) @xmath576 = \sum m(e , d)[\bar{f}^1({\mathfrak{p}},d)]$ ] . \b ) if @xmath171 is irreducible , then @xmath618 , and @xmath496 is irreducible . b ) is a strengthening of theorem [ th2.1 ] under the assumption that @xmath538 . ] \c ) @xmath612 = f^1({\mathfrak{p}},e)[\mu](\cong { \mathbb{c}}^{\dim e}v(\mu))$ ] . \d ) @xmath619 = \sum p(e , d)[f^1({\mathfrak{p}},d)]$ ] , where ( p(e , d ) ) is the matrix inverse to @xmath565 . \e ) @xmath620 . ( see @xcite for a formula for @xmath621 . ) \f ) @xmath622 ; in particular , the @xmath1-module @xmath20 determines @xmath168 up to isomorphism . \a ) apply proposition [ pr7.2 ] and theorem [ th7.4 ] . \b ) if @xmath171 is irreducible , then @xmath623 for @xmath573 and @xmath624 . now apply corollary [ cor7.5 ] a ) . \c ) combine formula ( [ eqn8 ] ) with theorem [ th7.4 ] . \d ) follows from a ) and the definition of the matrix @xmath625 . \e ) follows from c ) . \f ) apply corollary [ cor6.2 ] and theorem [ th7.4 ] . let @xmath626 and let @xmath627 be the full subcategory of @xmath296 consisting of @xmath151-modules @xmath298 whose weight spaces @xmath628 satisfy @xmath629 and @xmath630 . let @xmath30 be the full subcategory of @xmath322 consisting of @xmath0-modules @xmath3 with minimal @xmath9-type @xmath11 for @xmath631 . assume @xmath298 is a non - zero object in @xmath632 . [ lemma7.7 ] @xmath633 for @xmath634 and @xmath635 ; @xmath636 . @xmath298 has a finite composition series with simple subquotients @xmath572 in @xmath632 . we know that @xmath580 for @xmath634 and @xmath635 . therefore our claim follows from the long exact sequence for right derived functors . [ pr7.8 ] the restriction of @xmath637 to the full subcategory @xmath632 is a faithful exact functor . the exactness follows from lemma [ lemma7.7 ] . every map in @xmath632 factors into a composition of surjection followed by an injection . lemma [ lemma7.7 ] implies that @xmath637 maps a nonzero surjection to a nonzero surjection and a nonzero injection to a nonzero injection . suppose @xmath3 is a @xmath0-module . then @xmath638 for @xmath228 . apply the frobenius reciprocity spectral sequence in ch . 6 of @xcite , then quote lemma [ lemma7.7 ] . \a ) if @xmath3 is a finite dimensional @xmath1-module , then @xmath638 . \b ) suppose @xmath639 and @xmath640 are objects in @xmath632 . then , @xmath641 as finite - dimensional vector spaces.thus , @xmath642 . assume again that @xmath18 is symmetric . let @xmath168 be a simple finite dimensional @xmath116-module . then @xmath643 is a @xmath0-module of finite type , and hence a @xmath549-module of finite type over @xmath18 , i. e. a harish - chandra module . by the comparison principle [ pz4,proposition 2.6 ] , we have a @xmath1-module isomorphism @xmath644 . the @xmath0-module @xmath645 , denoted by @xmath646 , has been studied extensively in the harish - chandra module literature . ( see for example @xcite . ) if @xmath19 , the harish - chandra module @xmath646 has a simple socle , and @xmath647 . in this section we consider six different pairs @xmath0 such that @xmath23 and @xmath4 . we start by recalling the construction of the algebraic principal series of @xmath648-modules for a symmetric subalgebra @xmath649 , @xcite . we use this construction in subsections 8.3 - 8.6 below . let @xmath649 be a symmetric subalgebra of @xmath1 . denote by @xmath650 a maximal toral subalgebra of @xmath651 . if @xmath652 is proper , @xmath650 is non - zero . let @xmath653 be a cartan subalgebra of @xmath1 such that @xmath654 . choose an element @xmath655 such that the eigenvalues of @xmath351 on @xmath1 are real and @xmath656 . let @xmath657 . the following results are proved in @xcite . \a ) @xmath658 ; @xmath659 . \b ) if @xmath660 is a borel subalgebra of @xmath661 such that @xmath662 , then @xmath663 is a borel subalgebra of @xmath1 . hence , @xmath664 is a parabolic subalgebra of @xmath1 . \c ) if @xmath665 such that @xmath666 , then @xmath667 is conjugate to @xmath664 under the connected algebraic subgroup @xmath668 whose lie algebra is @xmath652 . we define an element @xmath669 to be _ nondegenerate _ if @xmath670 is a toral subalgebra of @xmath651 . moreover , an _ iwasawa parabolic subalgebra for the pair _ @xmath648 is any subalgebra of the form @xmath664 for some nondegenerate element @xmath669 , such that ad@xmath351 has real eigenvalues in @xmath1 . fix an iwasawa parabolic subalgebra @xmath671 . let @xmath672 be a finite - dimensional simple module over @xmath661 . endow @xmath672 with a @xmath673-module structure by setting @xmath674 . \a ) the _ iwasawa principal series module corresponding to the pair _ @xmath675 is the @xmath648-module @xmath676 \b ) a degenerate principal series module corresponding to the pair @xmath675 is the @xmath648-module @xmath677 where @xmath164 is a subalgebra containing @xmath673 and the @xmath673-module structure of @xmath672 extends to a @xmath164-module structure . there is an isomorphism of @xmath652-modules @xmath678 in particular , @xmath679 is a @xmath648-module of finite type ; if @xmath336 is a simple finite - dimensional @xmath652-module , then @xmath680 , hence @xmath681 a similar statement holds for @xmath682 . ( harish - chandra s subquotient theorem ) let @xmath3 be a simple @xmath648-module . then there exists a simple finite - dimensional @xmath116-module @xmath672 such that @xmath3 is a subquotient of @xmath679 . for any simple @xmath648-module @xmath3 and for any @xmath652-type @xmath336 , @xmath683 let @xmath685 and @xmath9 be the diagonal @xmath6-subalgebra of @xmath1 . the subalgebra @xmath9 is regular in @xmath1 . the pair @xmath0 is symmetric and its harish - chandra modules have been studied for over half century , see @xcite , @xcite and @xcite . the parabolic subalgebra @xmath110 is a borel subalgebra , and @xmath686 . we have @xmath687 , hence a minimal @xmath9-type @xmath11 is generic if @xmath688 and there is a bijection between the 1-dimensional complex family @xmath689 and the set of isomorphism classes of @xmath0-modules with minimal @xmath9-type @xmath11 . hence any simple @xmath0-module @xmath3 with minimal @xmath9-type @xmath690 is strongly reconstructible by theorem 4 in [ pz2 ] . on the other hand , corollary [ cor6.5 ] above , implies this fact under the stronger assumption @xmath691 . note that for each @xmath24 , there exists a 1-dimensional complex family of simple @xmath0-modules with minimal @xmath9-type @xmath11 . these modules are multiplicity - free ; a self - contained purely algebraic description of these modules in given in @xcite . we now consider the case @xmath692 . [ pr8.5 ] for any infinite - dimensional simple @xmath0-module @xmath3 with minimal @xmath9-type @xmath693 ( i. e. spherical simple @xmath0-module ) , there exists an @xmath70-module @xmath168 such that @xmath694 as a @xmath9-module , @xmath3 is isomorphic to @xmath695 , and there is no finite - dimensional simple @xmath1-module with the same central character as @xmath3 ( see for instance @xcite ) . now choose a 1-dimensional @xmath70-module @xmath168 ( in the case we consider , @xmath696 ) such that @xmath697 and @xmath496 has the same central character as @xmath3 . then @xmath471 , since otherwise @xmath468 would be a finite - dimensional @xmath0-module with the same central character as @xmath3 . by an application of the euler characteristic principle [ pz1 , theorem 11 ] , @xmath496 is isomorphic as a @xmath9-module to @xmath695 . therefore , @xmath496 has the same central character and the same @xmath9-character as @xmath3 , i.e. @xmath698 . note that modules @xmath3 as in proposition [ pr8.5 ] are not strongly reconstructible . indeed , if the central character of @xmath3 is regular , it is not difficult to show that there are precisely two 1-dimensional modules @xmath251 and @xmath408 such that @xmath699 for @xmath700 . in addition , in this case @xmath701 . in the case of a singular central character @xmath702 is a non - trivial self - extension of the unique 1-dimensional @xmath116-module @xmath168 such that @xmath703 . finally , it is true that any simple @xmath0-module for the pair considered is weakly reconstructible . we also remark that for any @xmath470 , @xmath496 can be either irreducible or reducible . let @xmath704 and @xmath9 be the @xmath6-subalgebra of @xmath1 generated by the root spaces @xmath705 . the subalgebra @xmath9 is regular in @xmath1 and @xmath706 is a symmetric subalgebra of @xmath1 isomorphic to @xmath707 . the parabolic subalgebra @xmath110 is a borel subalgebra with roots @xmath708 and @xmath709 . hence @xmath710 , and any simple @xmath0-module @xmath3 of finite type over @xmath9 is strongly reconstructible for @xmath711 by theorem 4 in @xcite . on the other hand , @xmath712 , hence corollary [ cor6.3 ] above implies the strong reconstructibility under the stronger assumption @xmath713 . for completeness we note that for a @xmath9-type @xmath714 of @xmath18 , a necessary , but not sufficient condition for @xmath714 to be generic is that @xmath715 . next , @xmath716 and hence @xmath717 . fix @xmath718 . as @xmath9 is regular in @xmath1 , by theorem 4 of @xcite there exists a bijection between isomorphism classes of simple @xmath0-modules with lowest @xmath9-type @xmath24 and @xmath70-weights @xmath183 such that @xmath719 . if @xmath720 is a generator of @xmath721 , observe that @xmath722 is a free continuous parameter of @xmath183 . an iwasawa parabolic subalgebra @xmath671 relative to @xmath723 is a borel subalgebra of @xmath1 . hence a finite - dimensional simple @xmath673-module @xmath672 is 1-dimensional . write @xmath724 , and @xmath725 for @xmath726 . [ pr8.6 ] the principal series module @xmath727 has finite type over @xmath9 and has lowest @xmath9-type @xmath693 . by frobenius reciprocity for the principal series , @xmath728 the right hand side can be computed explicitly . write @xmath729 , @xmath730 being a cartan subalgebra of @xmath18 with basis @xmath428 and @xmath720 . let @xmath731 satisfy @xmath732 . then @xmath733 is the basis of @xmath734 dual to the basis @xmath735 of @xmath730 . if @xmath714 is a @xmath18-type we can now write @xmath736 , with @xmath737 and @xmath738 . next , @xmath739 is a toral subalgebra of @xmath730 and is spanned over @xmath32 by @xmath740 . the eigenvalues of @xmath741 in @xmath714 are @xmath742 for @xmath743 , @xmath744 , all of multiplicity one . the single eigenvalue of @xmath741 in @xmath745 is @xmath746 . hence , @xmath747 precisely when there exists @xmath748 with @xmath744 such that @xmath749 . thus , by frobenius reciprocity , @xmath714 is a @xmath18-type of @xmath727 iff there exists @xmath743 with @xmath744 such that @xmath750 . as a consequence , if @xmath714 is a @xmath18-type of @xmath727 , then @xmath751 . if we restrict the action on @xmath727 from @xmath18 to @xmath9 we see that the multiplicity of @xmath752 in @xmath727 is @xmath753 . in figure 1 we indicate the convex hull of the @xmath18-support of @xmath727 . in general , @xmath754 , but in the figure we take @xmath755 . [ pr8.7 ] there exists an open dense subset @xmath756 such that @xmath727 is simple and not weakly reconstructible for every @xmath757 . the existence of an open dense subset @xmath758 such that @xmath727 is simple for @xmath759 is established in @xcite . moreover , this implies the claim as the set of weakly reconstructible modules depends on one complex and one integer parameters , while the set of irreducible principal series modules depends on two complex parameters . the bound @xmath690 is sharp relative to weak ( and also strong ) reconstruction for @xmath0-modules of finite type over @xmath9 . for a classification of simple @xmath549-modules , see @xcite . let @xmath704 and @xmath556 , the principal @xmath6-subalgebra of @xmath1 . the subalgebra @xmath9 is regular in @xmath1 and it is a symmetric subalgebra of @xmath1 . the parabolic subalgebra @xmath110 is a borel subalgebra and has positive roots @xmath709 , @xmath760 and @xmath761 . hence , @xmath762 . moreover , it is easy to check that @xmath763 , which shows that @xmath764 is generic if @xmath765 . on the other hand , @xmath766 , so the condition @xmath12 is equivalent to the same inequality : @xmath767 . furthermore , we have @xmath768 . fix @xmath769 . by theorem 4 of @xcite , there exists a bijection between isomorphism classes of simple @xmath0-modules with lowest @xmath9-type @xmath24 and @xmath70-weights @xmath183 such that @xmath770 . if @xmath720 is a generator of @xmath771 , observe that @xmath722 is a free continuous parameter of @xmath183 . let @xmath671 be an iwasawa parabolic subalgebra relative to @xmath9 . the principal series @xmath727 has two free complex parameters . let @xmath772 be the sum in @xmath727 of the @xmath9-types @xmath773 . since @xmath774 , it is easy to see that @xmath772 is a @xmath1-submodule of @xmath727 ( see for instance [ v2 , ch . a much deeper fact is that @xmath772 splits as a direct summand of two submodules @xmath775 and @xmath776 , where the lowest @xmath9-type of @xmath775 is @xmath777 and the lowest @xmath9-type of @xmath776 is @xmath635 . furthermore , we have the following . there exists an open dense subset @xmath756 such that @xmath776 , for @xmath757 , is simple and not weakly reconstructible . the existence of an open dense subset @xmath778 , such that the modules @xmath775 and @xmath776 are simple for @xmath759 , is established in [ v2 , ch . 8 ] . this implies the claim as ( similarly to the proof of proposition [ pr8.7 ] ) the set of weakly reconstructible modules depends on one complex and one integer parameters , while the set of irreducible principal series modules depends on two complex parameters . the bound @xmath779 is sharp relative to weak ( and also strong ) reconstruction for @xmath0-modules of finite type over @xmath9 . let @xmath780 . the @xmath70-roots of @xmath1 are @xmath781 . let @xmath9 be the @xmath6-subalgebra generated by @xmath782 . the nilradical of @xmath110 has roots @xmath783 . hence , @xmath784 , @xmath785 , and a weight @xmath470 is generic if @xmath688 . on the other hand , @xmath786 , so the condition @xmath787 is stronger than the genericity condition . note that @xmath717 . finally , @xmath788 , where @xmath721 is the @xmath6-subalgebra generated by @xmath789 . fix @xmath718 . by theorem 3 of @xcite , we have a bijection between the following sets : \a ) isomorphism classes of simple @xmath0-modules @xmath3 having finite type over @xmath9 and lowest @xmath9-type @xmath24 ; \b ) isomorphism classes of simple , finite - dimensional @xmath116-modules @xmath168 such that the highest weight @xmath183 of @xmath168 satisfies @xmath719 . moreover , if @xmath720 is a generator of @xmath790 is a free but discrete parameter of @xmath183 ; in this the fundamental series of @xmath0-modules depends on two discrete parameters . let @xmath671 be an iwasawa parabolic subalgebra relative to the symmetric subalgebra @xmath706 . we can choose a levi decomposition @xmath791 such that @xmath792 is the diagonal @xmath6-subalgebra in @xmath18 . let @xmath745 be a simple finite - dimensional @xmath661-module with highest weight @xmath726 , where @xmath653 is a cartan subalgebra of @xmath661 . the weight @xmath793 has one discrete and one continuous parameter . the principal series module @xmath727 has finite type over @xmath9 and has lowest @xmath9-type @xmath777 . it is completely analogous to the proof of proposition [ pr8.6 ] , and figure 2 is the analogue of figure 1 . every simple @xmath549-module has finite type over @xmath9 . if @xmath793 is nonintegral as a weight of @xmath1 , then @xmath727 is simple and not weakly reconstructible . see @xcite , theorem 2.3.1 . the bound @xmath690 is sharp relative to weak ( and also strong ) reconstruction for @xmath0-modules of finite type over @xmath9 . the set of simple modules of the form @xmath727 is not countable while the set of simple weakly reconstructible @xmath0-modules is countable . this implies the claim . let @xmath780 and @xmath9 be generated by @xmath705 . then @xmath794 . the nilradical of the parabolic subalgebra @xmath110 has roots @xmath795 and @xmath796 . hence , @xmath797 , @xmath798 , and @xmath13 is generic if @xmath799 . on the other hand , @xmath800 , so the condition @xmath801 is equivalent to being generic . note that @xmath768 . finally , @xmath788 , where @xmath721 is the @xmath6-subalgebra generated by @xmath802 . fix @xmath803 . theorem 3 of @xcite , or equivalently corollary [ cor6.5 ] , implies that we have a bijection between the following sets : \a ) isomorphism classes of simple @xmath0-modules @xmath3 having finite type over @xmath9 and lowest @xmath9-type @xmath24 ; \b ) isomorphism classes of simple finite - dimensional @xmath116-modules @xmath168 such that the highest weight @xmath183 of @xmath168 satisfies @xmath770 . if @xmath804 is a generator of @xmath805 $ ] , observe that @xmath806 is a free but discrete parameter for @xmath183 . we will exhibit a simple @xmath0-module @xmath3 of finite type over @xmath9 such that @xmath3 has lowest @xmath9-type @xmath807 but @xmath3 is not weakly reconstructible . let @xmath808 . let @xmath673 be an iwasawa parabolic subalgebra of @xmath1 relative to @xmath18 . this is a borel subalgebra . let @xmath164 be a maximal parabolic subalgebra of @xmath1 such that @xmath809 ( there are two choices for @xmath164 . ) write @xmath810 , where @xmath38 is a reductive part of @xmath164 and @xmath811 is the nilradical of @xmath164 . note that we can choose @xmath38 so that @xmath812 and @xmath813 is a 1-dimensional toral subalgebra of @xmath38 . next , let @xmath745 be a simple finite - dimensional @xmath38-module with @xmath653-highest weight @xmath793 . write @xmath814 for the degenerate principal series module @xmath815 . since @xmath816 as a @xmath9-module , @xmath814 is a direct sum of two submodules , @xmath817 and @xmath818 corresponding to even highest weights of @xmath9 and odd highest weights of @xmath9 , respectively . \a ) @xmath814 is a @xmath0-module of finite type over @xmath9 . \b ) the lowest @xmath9-type of @xmath817 is @xmath693 ; the lowest @xmath9-type of @xmath818 is @xmath819 . \c ) we can choose @xmath793 so that the central character of @xmath814 is not equal to the central character of a fundamental series module for @xmath0 . straightforward calculation . [ pr8.14 ] the bound @xmath691 is sharp relative to weak reconstruction for @xmath0-modules of finite type over @xmath9 . we take @xmath3 to be a simple quotient of @xmath820)$ ] and we chose @xmath793 so that @xmath3 does not have the central character of a fundamental series module . let @xmath780 and let @xmath9 be a principal @xmath6-subalgebra . here it is easy to check that @xmath821 , and the results of @xcite ( see formula ( 16 ) in @xcite ) imply that any simple @xmath0-module with minimal @xmath9-type @xmath11 for @xmath822 is strongly reconstructible ( and in particular is of finite type ) . the same follows from corollary [ cor6.6 ] under the weaker assumption that @xmath823 . since @xmath824 , corollary [ cor6.6 ] implies that , for any @xmath825 , we have a bijection between the set @xmath826 and the set of isomorphism classes of @xmath0-modules with minimal @xmath9-type @xmath11 . the bound @xmath823 is sharp relative to the theorem on strong reconstruction for @xmath0-modules of finite type over @xmath9 . in @xcite a simple multiplicity - free @xmath0-module @xmath827 with @xmath9-character @xmath828 and central character @xmath829 is exhibited : see equation 6.2 in @xcite . on the other hand , there are 8 fundamental series modules of the form @xmath830 such that @xmath831 . a non - difficult computation shows that their respective minimal @xmath9-types are @xmath832 and @xmath833 . this shows that @xmath827 is not strongly reconstructible . we do not know whether the bound @xmath823 is sharp relative to weak reconstruction . the following examples demonstrate that theorem [ th7.4 ] does not extend to the case @xmath834 . * example 1 . * there is a unique 1-dimensional @xmath91-module @xmath835 such that @xmath836 and such that the minimal @xmath9-type of @xmath837 is @xmath833 . by direct computation , @xmath838 is multiplicity free over @xmath9 . by comparison with the simple multiplicity free modules discussed in @xcite , we conclude that @xmath839 is the sum of two simple characters . by the duality theorem and the fact that @xmath840 is multiplicity free , we conclude that @xmath840 is the direct sum of two simple submodules with lowest @xmath9-types @xmath833 and @xmath841 respectively . this decomposition is consistent with conjecture [ conj7.4 ] . moreover , the proper inclusions of @xmath0-modules @xmath842 demonstrate that the inclusions discussed in proposition [ pr7.3 ] b ) are generally proper . * example 2 . * there is a unique 1-dimensional @xmath91-module @xmath251 such that @xmath843 and the minimal @xmath9-type of @xmath844 is @xmath819 . as in example 1 , we find that @xmath845 is a direct sum of two simple multiplicity free @xmath0-modules with lowest @xmath9-types @xmath819 and @xmath846 respectively . recall the categories @xmath627 and @xmath30 introduced in section 7 . proposition [ pr7.2 ] and proposition [ pr7.8 ] imply that @xmath27 is a well - defined faithful and exact functor between @xmath29 and @xmath30 for @xmath847 . theorem [ th7.4 ] implies that if @xmath848 , the simple objects @xmath849 of the category @xmath29 are being mapped by @xmath27 into simple objects of @xmath30 , and corollary [ cor6.4 ] ensures that , under the stronger condition @xmath850 , @xmath27 induces a bijection on the isomorphism classes of simple objects of @xmath29 and @xmath30 . conjecture [ con9.1 ] implies the existence of an isomorphism @xmath851 for any simple objects @xmath852 of @xmath29 where @xmath31 . we have checked the existence of such an isomorphism by direct computations in the cases of subsections 8.2 and 8.4 . in conclusion we note that it is easy to check that conjecture [ con9.1 ] holds for the case when @xmath853 . in this case @xmath854 and @xmath27 is an equivalence of the categories @xmath29 and @xmath30 for any @xmath847 . 20 v. bargmann , irreducible unitary representations of the lorentz group , ann . of math . ( 2 ) * 48 * ( 1947 ) , 586 - 640 . a. borel , j. de siebenthal , les sous - groups ferms de rang maximum de groupes de lie clos , comm . helv . * 23(1949 ) , no . 1 , 200 - 221 . j. bernstein , s. gelfand , tensor products of finite and infinite - dimensional representations of semisimple lie algebras , comp . math . 41 * ( 1980 ) , 245 - 285 . d. collingwood , representations of rank one lie groups , research notes in mathematics * 137 * , pitman , london , 1985 . l. casian , d. collingwood , the kazhdan - lusztig conjecture for generalized verma modules , math . z. * 195 * ( 1987 ) , 581 - 600 . j. dixmier , enveloping algebras , american math . soc . , 1996 . enright , n.r . wallach , notes on homological algebra and representations of lie algebras , duke math . j. * 47 * ( 1980 ) , 1 - 15 . i. m. gelfand , m. a. naimark , unitary representations of the lorentz group , izvestiya akad . nauk sssr , ser . mat . * 11 * ( 1947 ) , 411 - 504 . harish - chandra , infinite irreducible representations of the lorentz group , proc . london , ser . a * 189 * ( 1947 ) , 372 - 401 . a. knapp , lie groups : beyond an introduction , birkhauser , boston , 2002 . b. kostant , lie algebra cohomology and the generalized borel - weil theorem , ann . of math . * 74 * ( 1961 ) , 329 - 387 . h. kraljevi , representations of the universal covering group of the group su(n,1 ) , glasnik mat . iii * 28 * ( 1973 ) , 23 - 72 . a. knapp , d. vogan , cohomological induction and unitary representations , princeton mathematical series , princeton university press , 1995 . i. penkov , v. serganova , bounded simple @xmath855-modules for rk@xmath856 , journal of lie theory * 20 * ( 2010 ) , 581 - 615 . i. penkov , g. zuckerman , generalized harish - chandra modules : a new direction of the structure theory of representations , acta applicandae mathematicae * 81(2004 ) , 311 - 326 . i. penkov , g. zuckerman , generalized harish - chandra modules with generic minimal @xmath9-type , asian j. of math . 8 * ( 2004 ) , 795 - 812 . i. penkov , g. zuckerman , a construction of generalized harish - chandra modules with arbitrary minimal @xmath9-type , canadian mathematical bulletin * 50(2007 ) , 603 - 609 . i. penkov , g. zuckerman , a construction of generalized harish - chandra modules for locally reductive lie algebras , transformation groups 13 * ( 2008 ) , 799 - 817 . d. vogan , the algebraic structure of the representations of semisimple lie groups i , ann . of math . * 109*(1979 ) , 1 - 60 . d. vogan , representations of real reductive lie groups , progress in math . 15 , birkhauser , boston,1981 . g. zuckerman , generalized harish - chandra modules , progress in math . 295 , birkhauser , boston , 2011 , to appear .	we continue the study of the fundamental series of generalized harish - chandra modules initiated in @xcite . generalized harish - chandra modules are @xmath0-modules of finite type where @xmath1 is a semisimple lie algebra and @xmath2 is a reductive in @xmath1 subalgebra . a first result of the present paper is that a fundamental series module is a @xmath1-module of finite length . we then define the notions of strongly and weakly reconstructible simple @xmath0-modules @xmath3 which reflect to what extent @xmath3 can be determined via its appearance in the socle of a fundamental series module . in the second part of the paper we concentrate on the case @xmath4 and prove a sufficient condition for strong reconstructibility . this strengthens our main result from @xcite for the case @xmath5 . we also compute the @xmath6-characters of all simple strongly reconstructible ( and some weakly reconstructible ) @xmath7-modules . we conclude the paper by discussing a functor between a generalization of the category @xmath8 and a category of @xmath7-modules , and we conjecture that this functor is an equivalence of categories . * mathematics subject classification ( 2000 ) . * primary 17b10 , 17b55 .
gravitational collapse with formation of black holes is one of the main problems of classical general relativity . a very important question here is how formation of a black hole depends on initial data , for example , which initial data evolve into black holes and which do not . it is concerned with many important problems such as cosmic censorship @xcite . christodoulou @xcite rigorously analyzed such a problem in a relatively simple system , namely , spherically symmetric scalar field collapse . he proved that the field disperses at later times if the initial field is sufficiently weak , and also found a sufficient condition for black hole formation . however , the behaviour of the field and the space - time for initial data sets around the threshold of the black hole formation remained open , and he posed a question whether the phase transition between space - times with and without black holes is of the first or of the second order , i.e. , whether the mass of the black hole formed is a continuous function at the threshold . choptuik @xcite numerically analyzed the system and obtained a result which strongly suggests the latter . furthermore , he discovered a `` critical behaviour '' of the gravitational collapse of the system . his result can be summarized as follows : let the initial distribution of the scalar field be parametrized smoothly by a parameter @xmath7 , such that the solutions with the initial data @xmath8 contain a black hole while those with @xmath9 do not . for several one - parameter families investigated , near - critical solutions ( @xmath10 ) satisfy the following : ( 1 ) the critical solution ( i.e. @xmath11 ) is _ universal _ in the sense that it approaches the identical space - time for all families , ( 2 ) the critical solution has a discrete self - similarity , and ( 3 ) for supercritical solutions ( @xmath12 ) the black hole mass satisfies @xmath13 and the critical exponent @xmath0 , which is about 0.37 , is universal for all families . abrahams and evans @xcite found similar phenomena in axisymmetric collapse of gravitational wave and obtained @xmath14 . evans and coleman @xcite found similar phenomena with @xmath15 in spherically symmetric collapse of radiation fluid , in which case the self - similarity is not discrete but _ continuous_. employing a self - similar ansatz they also found a numerical solution which fits the inner region of the near - critical solutions very well . since the above values of @xmath0 were close to each other , some people considered that there is a universality over many systems . for a deeper understanding of these phenomena , an important direction is to pursue the analogy not only phenomenologically but also theoretically between the `` critical phenomena '' in gravitational collapse and those in phase transitions in statistical mechanics . argyres @xcite reviewed the above numerical results in terms of renormalization group ( rg ) . independently , the present authors @xcite showed that both qualitative and quantitative understanding is possible by introducing renormalization group ideas . they showed that _ uniqueness _ of the relevant ( unstable ) eigenmode of the linear perturbations around the critical self - similar space - time is essential for the universality to be observed and that the critical exponent @xmath0 is given by @xmath16 , where @xmath1 is the only eigenvalue with positive real part . emphasized there was that the standard renormalization group argument implies that the linear perturbation analysis is sufficient to obtain the _ exact _ value of @xmath0 . they confirmed that their scenario holds in the case of radiation fluid and obtained a very accurate value of the exponent , @xmath17 . the difference from those obtained for scalar field collapse ( 0.37 ) and gravitational wave collapse ( 0.38 ) was beyond the possible numerical errors in their simulations . this showed that there is no universality between radiation fluid collapse and other systems . maison @xcite applied the method to the systems of perfect fluids with pressure proportional to density and showed how @xmath0 depends on the equation of state , under the assumption that the same phenomena occur and the same scenario holds as the radiation fluid . gundlach @xcite confirmed @xmath18 in scalar field collapse . some new and intriguing phenomena in the critical behaviour have been found such as bifurcation @xcite and coexistence of first and second order phase transitions @xcite . however , in this paper we will pursue a slightly different direction , namely concentrate on providing deeper understanding of basic aspects of the critical behaviour . this paper is the fully expanded version of ref . ( some of the materials here have already been treated in @xcite . ) the aim of the paper is twofold . first , we present a general , mathematically rigorous , framework in which critical behaviour observed in various models so far can be analyzed and understood in a unified way , based on a renormalization group picture . now that there seems to be some consensus on the qualitative picture on the simplest case , we in this paper focus our attention to two problems which remain to be more clarified . ( 1 ) we give a sufficient condition under which the critical exponent is given _ exactly _ by @xmath2 . in the above , @xmath3 is a scaling exponent which appears in calculating the lyapunov exponent ( see sec . [ sec - rgscenario ] for details ) . note that in @xcite we only considered the radiation fluid collapse , where one can take @xmath19 . ( 2 ) we also consider the problem of universality , in the wider sense of the word that details of the definition of the model is irrelevant for the critical behaviour , and we present a sufficient condition under which the universality of this kind occurs . both conditions are sufficiently weak that they are expected to be applicable to most systems for which the critical behaviour has been observed . second , we present a more thorough analysis of the perfect fluid collapse , with various adiabatic index @xmath4 . in addition to presenting the details of our analysis on the radiation fluid collapse reported in @xcite , we here report a result of a new method of analysis , the so called _ lyapunov analysis _ , which was performed to further confirm the uniqueness of the relevant mode around the critical solution . a lyapunov analysis and the shooting method for an ordinary differential equation adopted in @xcite are complementary to each other in the following sense : ( 1 ) the former extracts eigenmodes in the descending order of its real part , whereas the latter affords information only of a finite region of complex @xmath1 plane . ( 2 ) the lyapunov analysis is useful even if the eigenmodes do not form a complete set . ( 3 ) it is easier in the latter than in the former to numerically obtain accurate values of the eigenvalues @xmath1 hence the critical exponent @xmath0 . our analysis in the second part is sufficiently convincing for physicists , we believe . but mathematically it is unsatisfactory in the sense that it does not provide rigorous theorems . rigorous full analysis of the system is a subject of future studies . this paper is organized as follows . first , we present our general scenario in sec . [ sec - rgscenario ] . we define renormalization group transformations acting on the phase space ( sec . [ sec - rg ] ) , and see how all the aspects of critical phenomena for radiation fluid collapse , etc . are deduced from the behaviour of the rg flow near a fixed point ( sec . [ sec - scenario ] ) . we present a sufficient condition for the exactness of the relation @xmath2 in sec . [ sub - rgflow ] . we then consider more general cases , i.e. discrete self - similarity ( sec . [ sec-discr.ss ] ) and problem of universality for different systems ( sec . [ sec-univ.class ] ) . then we turn our attention to detailed study of perfect fluid collapse . after reviewing the equations of motion in sec . [ sec - eom ] , we first study self - similar solution in sec . [ sec - ss ] . then we study perturbations around the self - similar solution and confirm our picture to an extent in sec . [ sec - pert ] by numerical study . in sec . [ sec - lyapunov ] we establish uniqueness of the relevant mode by the lyapunov analysis . the uniqueness of the relevant mode implies that that self - similar solution is responsible for the critical behaviour . so we list in sec . [ sec - results ] the value of the critical exponent @xmath3 for various @xmath4 . [ conc ] is for conclusions and discussions . in this section we present a scenario of critical behaviour based on renormalization group ideas , which gives clear understanding of every aspect of the phenomena . we first introduce renormalization group transformations acting on the phase space of partial differential equations ( pdes ) in sec . [ sec - rg ] . in sec . [ sec - scenario ] we give a ( nonrigorous ) scenario for the simplest case where the system has scale invariance and a ( continuously ) self - similar solution plays an important role . even in this simplest case we can argue most of the essential features of the critical behaviour . this case covers a perfect fluid collapse treated in detail in the subsequent sections . we then proceed in sec . [ sub - rgflow ] to make our scenario into a rigorous mathematical framework which starts from basic assumptions , based on renormalization group ideas . the case where discrete self - similar solution becomes relevant such as scalar field collapse is understood essentially in the same way , as is explained in sec . [ sec-discr.ss ] . in general cases where the system is not scale invariant , the renormalization group transformation drives the equations of motion to a fixed point . renormalization group method then tells us which systems can exhibit the same critical behaviour , as will be explained in sec . [ sec-univ.class ] . as examples , we consider perfect fluid with a modified equation of state , and also show that scalar fields with any potential term ( which depends on the scalar field only ) should show the same critical behaviour . currently , there seem to be two kinds of renormalization group type analysis of pdes , @xcite and @xcite . although these two are certainly philosophically related , they differ in details . we here adapt the approach of @xcite , and extend their work to cases where the critical behaviour is observed . it would also be very interesting to consider applications of the method of @xcite . we give a general formalism which deals with time evolution of initial data considered as a flow of a renormalization group transformation . the argument is general but the notations are so chosen in order to suit the case considered in the subsequent sections . in particular , we explicitly consider only the case of spherical symmetry , although our framework can in principle be applied to more general cases in @xmath20-space dimensions with suitable changes : the independent space variable @xmath21 is increased to @xmath22 , and the differential operators @xmath23 and @xmath24 become partial differential operators on a space of functions of @xmath20 variables . we are interested in the time evolution of @xmath20 unknown functions @xmath25 , which are _ real - valued _ functions of time @xmath26 and spatial coordinate @xmath21 , and which satisfy a system of partial differential equations ( pdes ) : @xmath27 this real formulation does not restrict applicable systems in any way , because any pdes for complex unknown functions can be rewritten in the real form by doubling the number of components . we employ this real form , because this makes the dimensional counting ( e.g. number of relevant modes ) simpler and clearer . see a remark after claim 1 of sec . [ sec-sen.critical ] . ( however , in practical problems , e.g. complex scalar fields , one could proceed without employing the real formulation , as long as some caution is taken in counting dimensions . ) the time evolution of this system is determined once one specifies the values of @xmath28 at the initial time @xmath29 . we call the space of functions @xmath28 ( at fixed time @xmath26 ) the _ phase space _ @xmath30 , which is considered as a real vector space . we leave the definition of @xmath30 ( e.g. how smooth functions in @xmath30 should be ) rather vague , until we develop more detailed theory of time evolution in sec . [ sec - scenario ] . to introduce a renormalization group transformation , we first define a _ scaling transformation _ @xmath31 which depends on real parameters @xmath32 , and @xmath3 : @xmath33 ( as will be clear @xmath3 here is related with , but not equal to , @xmath0 . ) now we make the fundamental * assumption s ( invariance under scaling ) * the system of pdes ( [ eq : evolution ] ) are invariant under the scaling transformation ( [ eq : sctrdef ] ) in the following sense : if @xmath34 is a solution of ( [ eq : evolution ] ) , then for suitably chosen constants @xmath35 , and for an arbitrary scale parameter @xmath36 ( close to @xmath37 ) , the new function @xmath38 is also a solution of the system ( [ eq : evolution ] ) . in the following we denote by @xmath35 the constants which make the above assumption hold for the system of pdes ( [ eq : evolution ] ) , and denote @xmath38 simply by @xmath39 unless otherwise stated . we consider only those systems for which the above assumption s holds in sec . [ sec - rg ] through sec . [ sec-discr.ss ] . in sec . [ sec-univ.class ] , we consider systems for which assumption s does not hold , and address the problem of universality in its wider sense . we now define a _ renormalization group transformation ( rgt ) _ acting on the phase space @xmath30 as @xmath40 where @xmath35 are constants of assumption s , @xmath41 , and @xmath34 is a solution of the pde system . that is , @xmath42 is given by developing the initial data @xmath43 from @xmath44 to @xmath45 by the pde , and rescaling the spatial coordinate @xmath21 and the unknown functions @xmath46 . we have written the argument of @xmath47 as @xmath48 rather than @xmath21 , in order to draw attention to the fact that the physical ( or geometrical ) length is given by @xmath49 rather than by @xmath48 itself . because of our assumption of the invariance of the pdes under scaling , @xmath50 forms a ( semi ) group : @xmath51 from this it follows immediately that the @xmath36-derivative of @xmath50 at @xmath52 is in fact an infinitesimal generator of @xmath50 : @xmath53 a _ fixed point _ of the renormalization group @xmath54 is a point @xmath55 in @xmath30 satisfying @xmath56 for all @xmath41 , and can also be characterized by @xmath57 . a function @xmath34 is called _ self - similar _ with parameters @xmath58 if it satisfies @xmath59 . each self - similar solution @xmath60 of the pdes with parameters @xmath58 of assumption s is related to a fixed point @xmath55 of @xmath50 by @xmath61 . the _ tangent map _ ( or _ frchet derivative _ ) @xmath62 of @xmath63 at @xmath47 is defined by @xmath64 note that @xmath62 is a _ linear _ operator for fixed @xmath36 and @xmath47 , in contrast with @xmath65 . the above definition suggests the following formal relation : @xmath66 where @xmath67 denotes a term whose norm is of order @xmath68 ( with some suitable norm @xmath69 ) . we define the @xmath36-derivative of @xmath62 at @xmath47 as @xmath70 which is formally related with @xmath71 by @xmath72 at a fixed point @xmath73 of @xmath50 , ( [ eq : ind.tan ] ) simplifies to @xmath74 and we denote the tangent map at @xmath55 simply by @xmath75 in the following . @xmath75 ( at @xmath55 ) forms a ( semi)group like @xmath65 , and can be represented in terms of its infinitesimal generator @xmath76 as @xmath77 an _ eigenmode _ @xmath78 of @xmath79 with eigenvalue @xmath1 is a ( complex - valued ) function satisfying ( @xmath80 ) @xmath81 note that we here consider eigenmode ( and spectral ) problem on complexification of @xmath30 , i.e. we allow complex eigenvalues and eigenvectors . this is necessary for our later analysis . although complex eigenvectors are not elements of @xmath30 , we can express vectors in @xmath30 by taking the real part of a linear combination of complex eigenvectors . these modes determine the flow of the rgt near the fixed point @xmath55 , we have to know not only the point ( discrete ) spectrum but also the continuous spectrum of @xmath24 . see assumption l1 of sec . [ sub - rgflow ] for details . ] . a mode with @xmath82 , a _ relevant mode _ , is tangent to a flow diverging from @xmath73 , and one with @xmath83 , an _ irrelevant mode _ , to a flow converging to it . a @xmath84 mode is called _ marginal_. * remark . * many equations of gravitational systems , including those of the perfect fluid , are not exactly of the form of ( [ eq : evolution ] ) . more precisely , we have two kinds of unknowns @xmath85 and @xmath86 ) , which satisfy the following pdes of two different classes : ( 1)the evolution equations : @xmath87 and ( 2 ) the constraint equations : @xmath88 in this case , the well posed cauchy problem is to specify the initial values of @xmath89 for @xmath90 ( and appropriate boundary conditions for @xmath91 s ) . in this sense , the phase space variables are @xmath46 s , while @xmath91 s should be considered as constraint variables . solving ( [ eq : constr ] ) for constraint variables @xmath91 s in terms of integrals of @xmath46 s and substituting them into ( [ eq : evolution2 ] ) yields evolution equations for @xmath46 s of the type of ( [ eq : evolution ] ) , with the difference that @xmath34 in the argument of @xmath92 now represents some complicated functionals ( integrals ) of @xmath46 s . in this form , we can apply our formalism of the renormalization group transformation . * in the above , we have presented the formalism in @xmath93 coordinate with unknowns @xmath46 , in order to directly reflect our physical interpretation . however , mathematical treatment can in principle be made much simpler by introducing new variables @xmath94 in terms of these , the scaling transformation @xmath95 becomes a simple translation in @xmath36 : @xmath96 and the parameters @xmath35 disappear . one can thus identify the scale parameter @xmath97 with the new `` time '' coordinate @xmath36 . the assumed scale invariance of the system implies that the eom do not include @xmath36 explicitly . now @xmath71 and @xmath76 can be expressed by usual @xmath36-derivatives on the functions @xmath98 . in @xcite , we formulated our renormalization group approach directly in terms of these variables . in this subsection we present our scenario of critical phenomena in gravitational collapse based on the renormalization group ideas . to describe our claims clearly , we first classify numerically observed ( u ) _ universality _ and ( s ) _ scaling _ into two : ( 1 ) those of the critical space - time and ( 2 ) those of the critical exponent . concerning the critical space - time , it has been observed that ( s1 ) : : near - critical ( but not exactly critical ) solutions always once approach , but eventually deviate from , a self - similar space - time , ( u1 ) : : which is unique . and concerning the critical exponent , it has been observed that ( s2 ) : : for slightly supercritical solutions , the mass of the black hole formed is expressed by a scaling law @xmath99 , and ( u2 ) : : the value of the critical exponent @xmath0 is the same for all generic one - parameter families in @xmath30 . we could add another kind of universality , universality over different systems . it has been found @xcite that the system for some potentials such as @xmath100 , etc . , also have the above properties with the same @xmath3 as the minimally coupled massless scalar field . this suggests that the following universality holds : ( u3a ) : : there are some models which show the same critical behaviour ( same self - similar solution , same critical exponent ) . in particular , in scalar field collapse , the critical phenomena is the same for any potential @xmath101 , or at least , for some potentials @xmath101 . note that the term `` universality '' is used in two ways . in ( u1 ) and ( u2 ) , it refers to the independence on the initial data ( or a family of initial data ) , while in ( u3a ) , it refers to the independence on the system s eom ( i.e. details in the definition of the model considered ) . in statistical mechanics , where the term was used for the first time in a similar context , `` universality '' refers to the latter ( although there is no `` initial data '' in equilibrium statistical mechanics and thus it is impossible to have a counterpart for the former ) . it is one of our main purpose in this paper to give sufficient conditions under which the two kinds of universality hold . historically , because of closeness of numerical values of the critical exponents of a scalar field @xcite , a radiation fluid @xcite , and gravitational waves @xcite , the following universality in still wider sense , though stated less precisely , was speculated . ( u3b ) : : the value of @xmath0 is the same over many systems , including a scalar field , a radiation fluid , and gravitational waves . the precise value 0.35580192 ... in ref . @xcite and the values for scalar field collapse ( 0.37 ) and gravitational wave collapse ( 0.38 ) shows that this is not true . we now present our scenario through two claims , and their derivations . we do this on two different levels . first , in sec . [ sec-sen.critical ] ( resp . [ sec-sen.exponent ] ) we present claims concerning qualitative ( resp . quantitative ) pictures on the critical behaviour , and present their ( nonrigorous ) ` derivation ' based on naive linear perturbation ideas . presentation in this sec . [ sec - scenario ] is intended to provide basic ideas , so the following claims and their ` proofs ' are _ not _ intended to be precise mathematical statements . in particular , naive perturbation is not sufficient to deal with long time asymptotics rigorously , and this is where the `` renormalization group '' idea comes in . we thus present in sec . [ sub - rgflow ] more detailed , rigorous derivation of theorem [ prop - final ] , a rigorous version of claims 1 and 2 , based on renormalization group philosophy . a mathematical remark about the phase space is in order . in order to formulate the time evolution , we consider our phase space @xmath30 to be a _ real _ banach space of functions as a space of gauge equivalence classes of functions . or we could fix the gauge and consider @xmath30 to be a usual function space . in the latter case , we should be careful in gauge - fixing so as not to count `` gauge modes '' as physically meaningful eigenmodes in linear perturbation analysis . ] @xmath102 s which satisfy certain decay and smoothness properties , with a norm @xmath103 under which the phase space is complete . concrete choice of the decay and smoothness properties should be determined on physical grounds for each models , together with the requirement that @xmath30 is invariant under the time evolution . the term `` approach '' and `` deviate '' in the statement ( s1 ) should be interpreted in terms of the norm of the phase space , although other notion of convergence , e.g. pointwise convergence , could equally work . precise final definition should be given based on rigorous mathematical analysis , which is beyond the scope of the present paper . in the concrete cases discussed later , the fixed point @xmath73 does not represent an asymptotically flat space - time . though one is usually interested in asymptotically flat space - time in gravitational collapse , we define our phase space @xmath30 to be sufficiently large to include @xmath73 . asymptotically flat data are included in the phase space ( if one chooses the functions @xmath104 appropriately ) and form a subspace of @xmath30 . we first consider the qualitative behaviour of critical solutions . [ claim-1 ] suppose there is a fixed point @xmath73 , with no marginal modes . if ( s1 ) and ( u1 ) hold with this @xmath73 , then there should be a _ unique _ relevant mode with a _ eigenvalue for this @xmath73 . conversely , if @xmath73 has a unique relevant mode , then ( s1 ) and ( u1 ) hold with this @xmath73 , at least for all the initial data close to @xmath73 . ( except , of course , for the exactly critical initial data . ) * remark . * because we are considering real pdes , the odes ] for eigenmodes are also real , and thus all eigenmodes with complex eigenvalues ( if any ) appear in complex conjugate pairs . this means that if such a system has a unique relevant mode , the eigenvalue must be real . * remark . * ( reason for employing the real formulation . ) one could proceed without employing the real formulation . however , in this case , dimension counting becomes more complicated ; because the desired critical behaviour is observed only when the relevant subspace is ( real ) one - dimensional , one has to see whether a given eigenvector spans a subspace of ( real ) dimension 1 or ( real ) dimension 2 . by reducing the system of pdes in its real form , we can avoid such subtleties . we emphasize again that this does not restrict the applicability of our claims in any way . for example , a system of pdes ( for complex unknowns ) which has exactly one relevant mode with a complex eigenvalue is easily seen to have a relevant invariant subspace which is ( real ) two - dimensional , thus leading to the same conclusion ( i.e. this does not show usual critical behaviour ) as is obtained from the real formulation . based on this observation , we in this paper drop `` re '' in the exponent relation @xmath105 . _ ` proof . ' _ suppose @xmath73 has a unique relevant mode . then the renormalization group flow around the fixed point @xmath73 is contracting except for the direction of the relevant mode ( fig . [ fig - flow0 ] ) , at least in some neighbourhood @xmath106 around @xmath73 . thus there will be a _ critical surface _ or a _ stable manifold _ @xmath107 of the fixed point @xmath73 , of codimension one , whose points will all be driven towards @xmath73 . and there will be an _ unstable manifold _ @xmath108 of dimension one , whose points are all driven away from @xmath73 . a one - parameter family of initial data @xmath109 will in general intersect with the critical surface , and the intersecting point @xmath110 will be driven to @xmath73 under the rgt : @xmath111 thus , @xmath110 is the initial data with critical parameter @xmath112 . an initial data @xmath113 in the one - parameter family @xmath114 , which is close to @xmath110 will first be driven towards @xmath73 along the stable manifold , but eventually be driven away along the unstable manifold . therefore , claims ( s1 ) and ( u1 ) hold . to show the converse , first suppose @xmath73 had no relevant mode . then any initial data ( sufficiently close to @xmath73 ) would all be driven towards @xmath73 until finally be absorbed into @xmath73 , which contradicts ( s1 ) . next suppose @xmath73 had more than one relevant modes . then the stable manifold of @xmath73 would have codimension more than one , and thus a generic one - parameter family of initial data would not intersect with the stable manifold of @xmath73 . this means that a generic critical or near - critical solution would not approach the self - similar solution @xmath73 ; instead , they would be driven towards a different self - similar solution with a unique relevant mode . thus claims ( s1 ) and ( u1 ) do not hold . @xmath115 ( [ eq : crit ] ) states that the fixed point @xmath73 is on the closure of the subspace of @xmath30 which consists of asymptotically flat data . this explains , in rather precise terms , that the self - similar solution plays such an important role that its perturbation gives an exact critical exponent even when one is interested only in asymptotically flat collapses . asymptotically flat data can converge to the self - similar fixed point with infinite mass in the above sense , even though the adm mass conserves . it should be emphasized that there is no contradiction involved here . that is , the mass measured in the unit of the _ scaled _ coordinate becomes infinite as the space - time approaches the self - similar solution , while the physical mass measured in the unit of the _ original _ coordinate remains finite . * some reader might doubt the validity of our claim ( [ eq : crit ] ) , because often the critical space - time satisfies @xmath116 . however , it is not difficult to find norms @xmath103 , for which ( [ eq : crit ] ) holds ; e.g. take @xmath117 the remaining , highly nontrivial , question is whether the assumptions in sec . [ sub - rgflow ] are satisfied in this norm , because the spectrum of perturbations can depend on the choice of the norm ( and thus on the function space ) , may not be so strong in the concrete case of perfect fluid , because the sonic point places an effective boundary condition ] . in other words , the norm and the function space @xmath30 should be so chosen that the assumptions hold ( and that @xmath30 includes @xmath55 ) : if we can find such a norm and @xmath30 , the critical behaviour should be observed ; this is what is proved in sec . [ sub - rgflow ] . the question of the right choice of the norm must be answered for each models under proper mathematical consideration . we emphasize that our goal is to present a _ general _ framework ( physically reasonable sufficient conditions ) under which the critical behaviour is bound to happen , which is independent of the details of specific models . * remark . * fixed points with more than one relevant modes are responsible for the so - called _ multicritical behaviour _ in statistical mechanics . we give examples of self - similar solutions with more than one relevant modes ( and thus are not responsible for generic critical behaviour ) in section [ subsec-pp.other-ss ] . another example is provided by the self - similar solutions of complex scalar fields @xcite . in this model the continuous self - similar solution has three relevant modes , and thus is irrelevant for a generic critical behaviour . in other words , we have to adjust three parameters ( instead of one ) in the initial condition , to observe a critical behaviour governed by this self - similar solution . we now turn our attention to the quantitative aspect of the critical behaviour , i.e. the critical exponent @xmath0 . [ claim-2 ] if the relevant mode is unique with a real eigenvalue @xmath1 , the black hole mass @xmath118 satisfies @xmath119 with @xmath120 for slightly supercritical solutions , where @xmath3 is the scaling exponent of the scaling transformation ( [ eq : sctrdef ] ) . here @xmath121 means there exist @xmath7-independent positive constants @xmath122 such that @xmath123 holds for @xmath7 sufficiently close to @xmath112 . _ ` proof . ' _ the claim , in particular the exactness of the relation ( [ eq : eq - exp0 ] ) , is one of the main conclusions of this paper . the derivation of this claim is rather lengthy , although it is a variation on rather standard treatment of similar problems in ( rigorous ) renormalization group analysis of critical behaviour in statistical mechanics . we first explain the rough idea , and then explain each step in detail . we first consider the fate of an initial data @xmath113 in the one - parameter family , which is close to @xmath110 ( @xmath124 ) : @xmath125 the data will first be driven towards @xmath73 along the critical surface , but eventually be driven away along the unstable manifold , until it finally blows up and forms a black hole , or leads to a flat space - time . to trace the time evolution , we divide the whole evolution into two stages : i ( @xmath126 ) and ii ( @xmath127 ) . see fig . [ fig - flow2 ] : but is close to @xmath128 , can be treated in a similar way as long as the flow along @xmath128 is regular . this is because the time spent before it reaches stage i is finite . ] the stage i corresponds to @xmath129 , while stage ii to @xmath130(black hole ) . the stage i is the portion of the journey where @xmath131 is close to @xmath73 and we can use the perturbation . stage ii is the region where perturbation breaks down , and @xmath132 is so chosen to make this separation possible . concretely , we choose @xmath132 as the smallest @xmath36 such that @xmath133 becomes of order 1 ( after @xmath131 begins to deviate from @xmath55 ) . as will be explained in detail , near - critical solution @xmath131 spends most of its time in stage i , and the time of stage ii is relatively short . based on the analysis of the flow , we then derive the exponent by interpreting the result in our original coordinate . as will be explained , what matters in determining the exponent is the behaviour of the flow in stage i. _ step 1 . flows in stage i : _ the analysis of this stage is essential , because the flow spends most of its time at this stage , and gives the value of the exponent . for stage i , we can rely on linear perturbations to get @xmath134 where @xmath135 is roughly the component of the relevant mode in @xmath136 the derivation of ( [ eq : lin1 ] ) requires some calculations , together with suitable assumptions on invariant manifolds of @xmath76 . for the ease of reading , we postpone this derivation to section [ subsub-rg.quan2 ] , and proceed assuming ( [ eq : lin1 ] ) is correct . due to ( [ eq : crit ] ) , we have ( for large @xmath132 ) @xmath137 our choice of @xmath132 implies @xmath138 ( here @xmath139 means a positive small which is independent of @xmath140 . ) note that @xmath132 diverges logarithmically in @xmath140 as @xmath141 . _ step 2 . flows in stage ii : _ we further evolve @xmath142 until it finally diverges or leads to a flat space - time . at this point , we _ assume _ that the phase space @xmath30 has a rather simple structure , i.e. all points ( at least in a neighbourhood of @xmath55 ) on one side of @xmath128 blow up and lead to a black hole , while those on the other side disperse and lead to a flat space - time . once one admits this nontrivial assumption on the _ global _ structure of @xmath30 , stage ii becomes relatively trivial , because of the large second term of ( [ eq : rg.result ] ) . that is , the data @xmath142 differs from @xmath73 so much that one can tell the fate of this data depending on the sign of @xmath140 . in particular , if a black hole is formed , the solution will blow up in a finite , rather short , time ( uniform in @xmath140 ) . this means the radius of its apparent horizon , and thus its mass , will be @xmath143 measured in @xmath48 . _ step 3 . going back to the original scale : _ finally , we translate the above result back into our original coordinate @xmath93 . the relation @xmath144 implies that the radius of the apparent horizon , which is @xmath143 measured in @xmath48 , is in fact @xmath145 measured in @xmath21 . so we have from ( [ eq : s-0value ] ) @xmath146 therefore the critical exponent is given _ exactly _ by @xmath147 this concludes our ` derivation . ' @xmath115 * remark . * as can be seen from the above proof, an intuitive understanding of the exactness of @xmath0 is as follows . in the limit @xmath148 , the time spent in a finite region around @xmath73 diverges whereas the time spent outside is usually finite before a black hole is formed , if there are no other fixed points , etc near the flow . so the change of the profiles of the fields are dominated by the linear stage so that the exact critical exponent is given in terms of the eigenvalue of linear perturbation around @xmath73 . the rigorous argument is presented in sec . [ sub - rgflow ] . it follows immediately from the above claim that conditions ( s2 ) and ( u2 ) hold and the critical exponent @xmath0 is exactly given by ( [ eq : beta.exp ] ) . the above ` proofs ' show that our claims are true not only in asymptotically flat cases but also in cases with any definition of mass proportional to the typical scale of the functions @xmath102 . if one admits claim 2 , it would in fact be a big surprise if the condition ( u3b ) did hold , since the critical solution of other systems , such as that of a scalar field system and that of a gravitational wave , are so different that one can not expect the eigenvalues of the perturbations around them to be the same . an explicit example of nonuniversality is given by the critical exponent @xmath0 for various parameter values of @xmath4 for the perfect fluid : see section [ sec - results ] . simple linear perturbations are not sufficient for rigorous study of the long time asymptotics of the solution . in this section , we describe the `` renormalization group '' type analysis in detail , which goes beyond the linear perturbation and provides firm grounds on our claims . our goal is to derive theorem [ prop - final ] , which is a rigorous version of claims 1 and 2 , from a few `` assumptions . '' the `` assumptions '' from which we start are grouped into two : ( 1 ) assumptions on the _ linear _ time evolution of the system , defined by @xmath76 ( or equivalently @xmath149 ) , and ( 2 ) certain regularity ( smoothness ) properties of @xmath50 and @xmath150 which physically look innocent . in the following @xmath151 is a positive constant suitably chosen for both of the following assumptions to hold . the first assumption specifies the properties of the _ linear _ part of the time evolution , generated by @xmath152 : * assumption l1 ( invariant subspaces of @xmath153 ) * for a suitably chosen @xmath151 , the whole tangent space @xmath154 is a direct sum of invariant subspaces of @xmath152 : @xmath155 here @xmath156 and @xmath157 are invariant subspaces of @xmath152 , of dimension @xmath158 and of codimension @xmath158 , respectively . there exists @xmath159 such that the restrictions of @xmath152 on these invariant subspaces satisfy : @xmath160 the estimates ( [ eq : ass.1.ones ] ) in assumption l1 could be further deduced from suitable assumptions on the spectrum of @xmath161 , because @xmath162 . one possibility of such an assumption would be : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ the spectrum of @xmath163 is a union @xmath164 . here ( i ) @xmath165 is a finite set with total multiplicity @xmath158 , ( ii ) @xmath166 and @xmath167 for some @xmath168 , and ( iii ) @xmath76 is a sectorial operator . _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ for a related estimate , see e.g. henry ( * ? ? ? * theorem 1.5.3 ) . before stating the second assumption , we set up the following convention on the choice of the norm @xmath103 . * convention on the choice of the norm @xmath103 . * for a vector @xmath78 , we denote by @xmath169 ( resp . @xmath170 ) the @xmath157 ( resp . @xmath156 ) component of @xmath78 : @xmath171 with @xmath172 , @xmath173 . assumption l1 guarantees the possibility of the decomposition . we define the `` box norm '' @xmath174 note that this box norm is equivalent to the original @xmath103 , because @xmath175 holds with some ( @xmath78-independent ) @xmath176 . the first inequality of ( [ eq : ass.1.decp2 ] ) is due to completeness , while the second is nothing but the triangle inequality . because this box norm simplifies our presentation ( e.g. conditions on various constants in propositions ) , we extensively use this box norm in the following , writing it simply as @xmath103 . sometimes , we write @xmath177 , @xmath178 . * notation * we extensively use following abbreviations throughout sec . [ sub - rgflow ] to sec . [ sec-univ.class ] . * we write @xmath179 . * @xmath180 denotes a scalar or vector depending on the context , whose magnitude ( measured in absolute value @xmath181 for a scalar , in the box norm @xmath103 for a vector ) is bounded by @xmath182 . * we also write @xmath183 to simultaneously represent @xmath184 and @xmath185 . that is , @xmath183 is equivalent to @xmath186 when @xmath187 . the second assumption refers to regularity properties of the flow , and essentially requires that the map @xmath188 has a lipschitz continuous frchet derivative . although it is sufficient to require only hlder continuity ( of the derivative ) with a positive exponent in the sequel , we here require lipschitz continuity , to make the presentation as simple as possible . * assumption l2 ( smoothness of @xmath189 ) * there exist positive constants @xmath190 such that + ( i ) the formal relation ( [ eq : rtrel ] ) holds for @xmath191 : @xmath192 ( ii ) @xmath193 is uniformly bounded for @xmath194 : @xmath195 we take @xmath196 sufficiently small so that @xmath197 . note that assumption l2 implies the following lipschitz - type continuity for @xmath188 and @xmath193 ( for @xmath198 ( @xmath199 ) ) : @xmath200 [ to prove ( [ eq : a.pert ] ) from assumption l2 , first note it suffices to consider the case of @xmath201 , because @xmath193 is a linear operator . then evaluate @xmath202 for @xmath201 in two ways using ( [ eq : rtrel2 ] ) : ( 1 ) express it in terms of @xmath203 , ( 2 ) expand each @xmath204 at @xmath205 and show that @xmath206 is @xmath207 . ] although assumption l2 physically looks rather innocent , its verification is a nontrivial mathematical problem depending on individual cases . we emphasize that @xmath208 term in assumption l2 is not necessarily bounded uniformly in @xmath151 , and can diverge for large @xmath151 . for example , they will certainly diverge ( as @xmath209 ) for non - critical initial data which will lead to a black hole . what is claimed in the assumption is that one can find finite nonzero @xmath151 and @xmath196 for which the above claims hold . we follow the time evolution by considering a discrete dynamical system generated by @xmath210 ( @xmath211 ) . the reason for considering the discrete system is twofold . first , it is naturally generalized to the case with discrete self - similarity , as discussed in sec . [ sec-discr.ss ] . second , the operator @xmath188 could be more regular than @xmath71 , thanks to a smoothing effect of the time evolution over a nonzero period @xmath151 . ( i.e. the linearization @xmath76 may not be bounded , while @xmath212 can be . ) based on the above assumptions , we now start our renormalization group study of flows . we here consider the qualitative aspects of the flow in a vicinity of the fixed point @xmath55 . we remark that there are considerable number of mathematical literature concerning the local structure of invariant manifolds . proposition [ prop-3 ] is a celebrated `` invariant manifold theorem '' ; see e.g. shub ( * ? ? ? * theorem 5.2 , theorem ii.4 ) or ( * ? ? ? * theorem iii.8 ) , for precise statements and a proof by sophisticated methods under slightly different assumptions . although the cited theorems are sufficient for our needs in this section , they do not cover the case considered in sec . [ sec-univ.class ] , where we consider the universality over different models . we therefore present our version of an elementary proof , in order to make the presentation more self - contained , and also to pave a natural way to our later analysis in sec . [ sec-univ.class ] . we consider the time evolution of an initial data @xmath47 ( close to @xmath55 ) under @xmath213 , by expressing it as ( @xmath214 ) @xmath215 where the second equality defines the difference @xmath216 between @xmath217 and @xmath55 . see fig . [ fig - dcmp](a ) . assumption l1 guarantees that we can always decompose this way . we begin with the following lemma , which claims that an initial data with @xmath218 is driven away from @xmath55 ( at least until @xmath219 becomes somewhat large ) . [ lem - ftriangle ] define @xmath220 by @xmath221 and consider the time evolution of an initial data @xmath222 with @xmath223 . then as long as @xmath224 , we have @xmath225 _ proof . _ we proceed by induction in @xmath20 . we start by obtaining recursion relations for @xmath226 and @xmath227 . assumption l2 reads @xmath228 which implies ( note : @xmath229 ) @xmath230 where in the second step we used the linearity of @xmath152 and the fact that @xmath169 and @xmath170 belong to invariant subspaces of @xmath152 . decomposing as in ( [ eq : rg.qual.dcmp ] ) , we have recursion relations for @xmath231 ( @xmath232 ) : @xmath233 our inductive assumption @xmath234 implies @xmath235 , by the definition of the box norm ( [ eq : def - boxnorm ] ) . so taking the norm of ( [ eq : rg.qual.4 ] ) and utilizing assumption l1 and assumption l2 to bound @xmath236 , we have recursions : @xmath237 now the recursion ( [ eq : qual.frec.1 ] ) immediately implies ( for @xmath238 ) @xmath239 on the other hand , ( [ eq : qual.frec.2 ] ) implies ( for @xmath240 ; use also the definition of @xmath241 ) @xmath242 these two prove ( [ eq : ftriangle2 ] ) . [ note : @xmath243 for positive @xmath244 . ] finally , ( [ eq : qual.frec.3 ] ) implies @xmath245 this , combined with @xmath246 just proven , implies ( [ eq : ftriangle1 ] ) , because @xmath247 . @xmath115 [ prop-3 ] under assumptions 1 and 2 , we have + ( i ) [ existence of critical solutions . ] for any @xmath172 with @xmath248 , there exists @xmath249 such that @xmath250 moreover , @xmath251 we define the `` local stable manifold '' as @xmath252 ( ii ) [ motion on @xmath253 . ] there exists @xmath254 such that @xmath255 and @xmath256 for any vector @xmath257 which satisfies @xmath258 . we can in fact take @xmath259 by taking @xmath260 small . + ( iii ) [ invariant manifolds . ] above @xmath253 in fact forms a manifold of codimension @xmath158 . we only present a proof for ( i ) , ( ii ) of the proposition . detailed information presented in ( iii ) is not necessary for our later analysis , so we only make brief comments on their proof at the end of the proof of ( i ) and ( ii ) . our proof is modeled after the argument by bleher and sinai @xcite , which first appeared in a similar context in statistical mechanics . _ proof of proposition [ prop-3 ] , ( i ) and ( ii ) _ we mainly consider the case of @xmath261 , which is of our main interest . for the other cases , see the end of this proof . we denote the normalized relevant mode of @xmath152 by @xmath262 : @xmath263 , @xmath264 . the proof is done in several steps . _ step 1 . the goal . _ we are interested in the time evolution of vectors @xmath265 , for fixed @xmath266 and various @xmath267 . our goal is to show that there is at least one value of @xmath267 ( called @xmath268 , as a function of @xmath266 ) for which @xmath269 . _ step 2 . recursion relations for @xmath270 and @xmath226 . _ we obtained recursions for @xmath226 and @xmath271 in the proof of lemma [ lem - ftriangle ] : ( [ eq : rg.qual.4 ] ) . utilizing assumption l1 , and using @xmath272 which follows from the definition of the box norm ( [ eq : def - boxnorm ] ) , we can convert ( [ eq : rg.qual.4 ] ) into recursion relations for @xmath270 and @xmath273 . the result is ( recall : @xmath179 ) @xmath274 roughly speaking , the recursion shows that the map for @xmath244 is expanding , while that for @xmath275 is contracting , with `` shifts '' of @xmath276 . if we neglect the second term , @xmath270 would blow up . thus the stable set @xmath253 is located by carefully balancing the inhomogeneous @xmath277 term with the first expanding term , so that @xmath270 will remain bounded ( in fact goes to zero ) as we iterate the recursion . that we can in fact find at least one such @xmath278 for given @xmath279 can be shown by using the argument due to bleher and sinai @xcite , as explained in steps 35 . _ step 3 . continuity of @xmath270 and @xmath280 as functions of @xmath267 . _ from ( [ eq : ass.lipshitz ] ) @xmath188 is a continuous function of its arguments . this implies in particular that @xmath281 and @xmath282 are continuous functions of @xmath270 and @xmath226 . iterating this for @xmath20 times , we see that @xmath270 and @xmath226 are continuous functions of the initial data , @xmath267 , for fixed @xmath266 . this implies @xmath273 is also continuous in @xmath267 . _ step 4 . solving the recursions . _ we now solve the recursion , following the technique of bleher and sinai . we will show that there are positive sequences @xmath283 such that ( a ) : : @xmath284 as @xmath285 ( b ) : : @xmath286 holds , as long as @xmath287 and @xmath288 . ( c ) : : as long as @xmath288 , the set of all @xmath281 when @xmath270 sweeps the interval @xmath289 contains the interval @xmath290 inside . the proof of ( a)(c ) is postponed to `` step 5 '' for the ease of reading . now given ( a)(c ) , we can show the existence of the critical @xmath268 as follows . let @xmath291 be the interval @xmath292 . because @xmath270 is a continuous function of @xmath267 , we can consider the inverse image @xmath293 of the interval @xmath291 under the @xmath20-iterations of the recursion . but ( a)(c ) show that @xmath293 are nonincreasing ( in @xmath20 ) sets , with nonempty intersections . so , taking an initial condition @xmath294 , we can guarantee for all @xmath214 @xmath295 because @xmath296 as @xmath297 , ( [ eq : lem3.ahbd ] ) shows @xmath298 for this special @xmath268 , i.e. @xmath299 , with @xmath300 : existence of a critical solution for given @xmath169 . _ step 5 . proof of ( a)(c ) . _ now we give an example of sequences @xmath283 which satisfy ( a)(c ) of step 4 . we define @xmath283 according to ( for @xmath301 ) @xmath302 then ( a ) is easily seen to be satisfied , because @xmath303 by the choice of @xmath241 . to prove ( b ) and ( c ) , first note the following sufficient condition : @xmath304 the choice of @xmath305 guarantees @xmath306 , and thus @xmath307 . this , together with @xmath308 , immediately shows that ( [ eq : bscond21 ] ) and ( [ eq : bscond22 ] ) are satisfied . eq.([eq : rntn - rel ] ) for @xmath309 proves ( [ eq : fu - fs ] ) . _ step 6 . motion of the critical solution . _ now that we have proven the existence of a critical solution , we concentrate on its time evolution . we first note @xmath310 holds for the critical solution for all @xmath214 , because otherwise the solution is shown to be driven away from @xmath55 by lemma [ lem - ftriangle ] . but then , the recursion ( [ eq : rg.qual.6 ] ) is reduced to @xmath311 h_{n } \label{eq : hnrecred2}\ ] ] which , in view of @xmath312 , implies @xmath313 with @xmath314 @xmath115 _ cases of @xmath315 and a comment on proposition [ prop-3 ] , ( iii ) _ all the above proves the existence of @xmath253 and the estimate ( [ eq : ws - behave ] ) for the case of @xmath261 . for @xmath316 we can proceed similarly and find the critical surface ( this time , we have to adjust @xmath158 parameters ) . for @xmath317 , similar argument shows that the map @xmath188 is a contraction in a neighbourhood of @xmath55 , thus proving that all the initial data close to @xmath55 is attracted to @xmath55 . proof of proposition [ prop-3 ] , ( iii ) can be performed , e.g. , as in ( * ? ? ? * theorem ii.4 ) . @xmath115 in the previous section , we established the existence of the critical surface ( stable manifold ) . we also showed that initial data which is sufficiently away from the critical surface ( characterized by @xmath318 ) is driven away from @xmath55 . in this and the following section , we concentrate on the fate of near ( but not exactly ) critical solutions . we in this section define an `` unstable manifold . '' then in sec . [ subsub-rg.quan2 ] , we show that all the near - critical initial data ( starting from an neighbourhood of @xmath55 ) are at first driven towards @xmath55 along the critical surface , but are eventually driven away along an unstable direction . we also give quantitative estimates on time intervals until the deviation of the data from @xmath55 becomes large . we concentrate on the case where there is a unique relevant mode ( @xmath261 in assumption l1 ) . more precisely we assume , in addition to assumptions l1 and l2 of sec.[subsub - rg.qual ] , * assumption l1a ( uniqueness of the relevant mode ) * the relevant mode of @xmath319 is unique , i.e. @xmath261 in assumption l1 . we denote the relevant mode by @xmath262 and the relevant _ real _ eigenvalue by @xmath1 ( @xmath320 ) : @xmath321 throughout this section , we use following notation . * @xmath322 , @xmath323 are positive constants chosen so that @xmath324 @xmath325 * given @xmath326 , define @xmath327 to be the largest integer such that @xmath328 holds for all @xmath329 . i.e. @xmath330 to trace the time evolution , we have to specify the unstable manifold . if @xmath188 is invertible , we can define the unstable manifold as @xmath331 interpreting @xmath332 for negative @xmath20 . however , we here do not assume the invertibility of @xmath188 . so the definition of @xmath333 requires some additional work . although our definition is somewhat artificial , it suffices our need . we define several sets , including the local unstable manifold @xmath334 . although all of these are defined with respect to the fixed point @xmath55 , we do not explicitly write @xmath55 except for @xmath334 , to simplify notation . we first introduce a line segment @xmath335 for @xmath336 as @xmath337 and define @xmath338 as [ @xmath339 is defined in ( [ eq : def - n1-gen ] ) ] @xmath340 then we define the local unstable manifold at @xmath55 as @xmath341 . \label{eq : wulocdef}\ ] ] lemma [ lem-6 ] below guarantees that @xmath334 is well defined , and all flows close to @xmath55 are squeezed around this @xmath334 . [ lem-6 ] ( i ) [ structure of @xmath334 ] @xmath334 of ( [ eq : wulocdef ] ) is a well - defined set , and is `` one - dimensional '' in the sense that its intersection with a plane @xmath343 exists as a unique point for @xmath344 . moreover , any @xmath345 satisfy @xmath346 with @xmath347 + ( ii ) [ contraction around @xmath334 in @xmath348-plane ] consider the time evolution of a vector @xmath47 under @xmath213 , with the initial data satisfying @xmath349 . we can find @xmath350 ( depending on @xmath322 ) such that @xmath351 holds as long as @xmath352 . for a proof , see appendix [ proof - lem-6 ] . finally we define sets of vectors @xmath353 where for a given set @xmath354 and a vector @xmath326 , we denote their `` distance '' as @xmath355 the set @xmath356 is schematically shown in fig . [ fig - flow2 ] of sec . [ sec-sen.exponent ] . we now trace the time evolution of a slightly super - critical initial data @xmath47 . our final goal is to trace it until it finally blows up and forms a black hole , but in this section we trace it until its deviation from @xmath253 becomes of order one ( stage i of sec . [ sec-sen.exponent ] ) as in ( [ eq : rg.result ] ) , so that we can tell that it is not critical . as shown , this part plays the essential role in determining the critical exponents . our presentation is modeled after a standard approach in applications of rigorous renormalization group techniques in statistical mechanics , see e.g. @xcite . the rest of the flow ( stage ii ) is considered in sec . [ subsub-rg.global ] . time evolution of an initial data sensitively depends on its location _ relative _ to @xmath253 . so we decompose the _ initial data _ @xmath47 as @xmath357 where @xmath358 [ i.e. @xmath359 as @xmath360 , @xmath262 is the normalized relevant eigenmode defined in ( [ eq : f1def ] ) , and @xmath361 is its coefficient . see fig . [ fig - dcmp](b ) in sec . [ subsub-rg.qual ] . proposition [ prop-3 ] ( i ) guarantees that we can find @xmath110 and decompose this way for @xmath258 . generically @xmath362 , because @xmath363 measures the component of @xmath262 in @xmath364 . now , for @xmath365 , we define @xmath216 in terms of the decomposition ) also for @xmath365 , i.e. @xmath366 with @xmath367 . however , we here employ ( [ eq : udiff - dcmp - non0 ] ) , because it also allows us to deal with the case of sec . [ sec-univ.class ] in essentially the same way . ] @xmath368 and further decompose @xmath216 in the spirit of ( [ eq : def - dcmpfsfu ] ) : @xmath369 see fig . [ fig - dcmp](b ) in sec . [ subsub-rg.qual ] . we do emphasize that the decomposition ( [ eq : udiff - dcmp - non0 ] ) differs from that of ( [ eq : rg.qual.dcmp ] ) in sec . [ subsub-rg.qual ] , in that @xmath78 there measures the difference between @xmath47 and the fixed point @xmath55 , while @xmath78 here measures the difference between @xmath47 and the ( time - evolved ) critical solution @xmath370 . also note that @xmath216 ( @xmath365 ) is not necessarily parallel to @xmath262 . the initial condition implies @xmath371 . after these preparations , concerning the flow in stage i , we have : [ prop-4 ] consider the time evolution of an initial data @xmath372 , with @xmath373 , @xmath374 , @xmath375 , and at each @xmath376 we decompose as in ( [ eq : udiff - dcmp - non0 ] ) : @xmath377 . let @xmath378 . then under assumptions l1 , l2 , and l1a , we can find @xmath379 ( independent of @xmath380 and @xmath322 ) such that the following holds : + ( i ) @xmath381 is finite and satisfies ( @xmath382 ) @xmath383 ( ii ) for sufficiently small @xmath384 , we can find @xmath385 such that @xmath386 where plus or minus sign is chosen according to the sign of @xmath380 . in fact we can take @xmath387 as @xmath388 . the above proposition gives an estimate on the time interval spent by the flow in stage i. it also provides information on the exit of stage i , by showing that the exit is essentially squeezed into the tiny tube for @xmath389 , we can reduce the tube @xmath356 to a disk , given by @xmath390 . ] @xmath356 around @xmath334 ( see fig . [ fig - flow2 ] ) . _ proof of proposition [ prop-4 ] . _ ( i ) we first derive recursion equations for @xmath391 and @xmath226 . using ( [ eq : rtrel2 ] ) with @xmath392 , we have @xmath393 this implies , according to our decomposition ( [ eq : udiff - dcmp - non0 ] ) , @xmath394 where in the second step we used ( [ eq : a.pert ] ) . this corresponds to the first line of ( [ eq : detail.phys.ex2 ] ) ; the difference is we here have @xmath395 . arguing as we did in deriving ( [ eq : rg.qual.4])([eq : qual.frec.3 ] ) , i.e. decomposing @xmath78 into @xmath169 and @xmath170 as in ( [ eq : udiff - dcmp - f ] ) , we this time obtain @xmath396 \label{eq : ap.ev33 } \\ & & b_{n+1 } = e^{\kappa\sigma } b_{n } \nonumber \\ & & \qquad + k_{1 } ( \| b_{n } \| \vee f_{n } ) \left [ { { \bar{o } } } ( 6 g_{n } ) + { { \bar{o } } } ( \| b_{n } \| \vee f_{n } ) \right ] . \label{eq : ap.ev34}\end{aligned}\ ] ] where we have introduced @xmath397 , @xmath398 to simplify the notation . our remaining task is to solve the recursions . this is an elementary exercise in calculus , and is summarized as lemma [ lem-5 ] in appendix [ proof - lem-5 ] : the recursive inequalities ( [ eq : ap.ev33 ] ) and ( [ eq : ap.ev34 ] ) are of the form of ( [ eq : ap.ev42 ] ) and ( [ eq : ap.ev41 ] ) there , with @xmath399 , @xmath400 , and @xmath401 . and that @xmath402 satisfies the assumption of the lemma follows from proposition [ prop-3 ] , ( [ eq : norm - bound - bs ] ) . part ( i ) of proposition [ prop-4 ] thus follows immediately from lemma [ lem-5 ] . \(ii ) we apply lemma [ lem-6 ] to the time evolution of the vector @xmath403 . here @xmath404 should be interpreted as its integer part . we have , by lemma [ lem-5 ] and proposition [ prop-3 ] , @xmath405 for @xmath406 $ ] , the above is bounded uniformly by @xmath407 . this also shows that the initial data @xmath408 for this part of time evolution satisfies @xmath409 as @xmath410 . so taking @xmath407 sufficiently small , we can apply lemma [ lem-6 ] to conclude that @xmath411 goes to zero as @xmath381 goes to infinity ( i.e. @xmath380 goes to zero ) . does not guarantee @xmath412 , because @xmath413 measures @xmath414 , and in general @xmath415 . however , because @xmath416 goes to zero as @xmath381 goes to infinity , the difference can be neglected as @xmath380 goes to zero . ] @xmath115 we are now at the stage of presenting the rigorous version of claims 1 and 2 , together with necessary assumptions . after the analysis in the previous section , that is , the analysis of the flow at stage i , we in this section consider stage ii of the flow . we want to trace the flow further until it blows up , and then calculate the mass of the black hole . this stage involves nontrivial physical and mathematical processes whose detailed analysis is beyond the scope of this paper . in this sense , this section is essentially a repetition of sec . [ sec-sen.exponent ] , with a difference that we here explicitly list up all the assumptions ( which are physically reasonable in view of numerical simulations ) on the global behaviour of the flow , under which we can present the rigorous version of claims 1 and 2 , i.e. theorem [ prop - final ] . * assumption g1 ( flows in stage ii ) * for sufficiently small @xmath417 and for @xmath418 defined in ( [ eq : deapm ] ) we have : * @xmath419 tends to a flat space - time as @xmath420 , for all @xmath421 , * @xmath422 blows up at some finite time @xmath423 , @xmath424 ; @xmath425 and @xmath426 are uniformly bounded above and below for all @xmath427 . [ recall that @xmath48 is related to @xmath21 by @xmath428 as in ( [ eq - xi - def ] ) . ] * assumption g2 ( mass inside the apparent horizon ) * the mass @xmath429 inside the apparent horizon at its formation is related with its radius @xmath430 by @xmath431 , where @xmath432 is a finite positive constant independent of @xmath430 . * assumption g3 ( relation between masses ) * the final mass of the black hole @xmath118 , and the mass at the formation of the apparent horizon @xmath429 , are of the same order . more precisely , there exists a positive constant @xmath433 such that @xmath434 holds for all the near - critical solutions . remarks on the meaning of these assumptions are in order . * remark . * 1 . assumption g1 specifies the behaviour of the flow in stage ii . thanks to our analysis in stage i , we have only to know the fate of the flow which is emerging from a very small neighbourhood of @xmath333 ( recall fig . [ fig - flow2 ] of sec . [ sec-sen.exponent ] ) . this means that physically it is sufficient to trace the time evolution of two vectors which is located on @xmath333 . 2 . assumption g2 is valid for a spherically symmetric gravitational collapse ( in fact , @xmath435 ) , which is the main topic in this paper . however we termed it as an assumption , in order to make our presentation more general . 3 . assumption g3 involves a subtle physical process after the formation of an apparent horizon . strictly speaking , the radius of the apparent horizon is not exactly related with the final bondi sachs mass which gives the mass of the black hole . the final mass @xmath118 differs from ( greater than , when the dominant energy condition is satisfied , ) the mass at the apparent horizon @xmath429 by the matter absorbed into the black hole after the formation of the apparent horizon . however , in our case , it is expected from the form of the relevant mode and the blow - up profile in numerical simulations that this effect can only make the black hole slightly heavier , say by a factor of two or so . ( it should be noted that usual numerical simulation assumes assumption g3 , and identifies the radius of apparent horizon with the final mass of the black hole . ) under these preparations , we can now state our rigorous version of claim 1 and claim 2 , which is the main result of this section . [ prop - final ] under assumptions l1 , l2 , l1a and assumptions g1 , g2 , g3 , the universality ( u1 , u2 ) and scaling ( s1 , s2 ) of sec . [ sec - scenario ] hold . more precisely , + ( i ) all the initial data sufficiently close to @xmath55 ( but not exactly on @xmath253 ) once approach , but finally deviate from , @xmath55 . the final fate of these data are either a flat space - time or a black hole , depending on which side of @xmath253 lies the initial data . + ( ii ) consider the time evolution of the initial data @xmath436 with @xmath437 , which is on the side of black hole of @xmath253 . then as @xmath363 goes to zero , the mass of the black hole formed satisfies @xmath438 with @xmath363-independent positive constants @xmath439 . here @xmath440 is the initial ( say , adm or bondi - sachs ) mass . in the expression ( [ eq : prop.final ] ) to avoid a nongeneric situation in which @xmath440 behaves like power of @xmath380 and @xmath0 becomes different from @xmath441 . ] * remark . * by refining the above argument leading to theorem [ prop - final ] , especially refining the proof of lemma [ lem-5 ] , we can see the following : if there is a marginal mode , but we still have a stable manifold of @xmath55 of codimension one , it can happen that the scaling of mass has a logarithmic correction term , like @xmath442 with some exponents @xmath0 and @xmath443 . _ _ proposition [ prop-4 ] is a rigorous version of step 1 of the ` proof ' of claim 2 of sec . [ sec-sen.exponent ] . assumptions g1 , g2 and g3 enable us to perform step 2 , proving ( i ) of the theorem . this also shows @xmath444 where @xmath432 is a positive constant , and @xmath445 is the total time until the formation of the apparent horizon . since @xmath446 by the definition of scaling transformation ( [ eq : sctrdef ] ) , we get ( ii ) of the theorem . @xmath115 # 1#1 # 1#1 significance of discretely self - similar solutions can be well understood with minor modifications to the scenario presented in sec . [ sec - scenario ] . let us begin with some definitions . a _ limit cycle _ of a renormalization group @xmath447 with fixed @xmath448 and @xmath3 is a periodic orbit of the flow @xmath449 where @xmath450 with fundamental periodicity @xmath451 . with a slight abuse of notation , we use @xmath452 to denote the orbit itself , while @xmath453 denotes a point on it . let us call a function @xmath34 _ discretely self - similar _ with parameters @xmath58 if it satisfies @xmath454 for some @xmath451 . each discretely self - similar solution @xmath34 of the pdes with parameters @xmath58 of the assumption s in sec . [ sec - scenario ] is related to a limit cycle @xmath452 by @xmath455 . for a limit cycle @xmath452 , it is convenient to consider eigenmodes of @xmath456 , where @xmath457 is defined in ( [ eq : ind.tan ] ) . an _ eigenmode _ @xmath78 of @xmath456 is a function satisfying @xmath458 the exponent @xmath1 , which is equal to the lyapunov exponent , gives the mean growth rate of the perturbation @xmath78 . _ relevant , irrelevant _ , and _ marginal modes _ are defined according to the sign of @xmath459 as in sec . [ sec - rg ] . it is easily seen that if @xmath78 is an eigenmode of @xmath456 then @xmath460 is an eigenmode of @xmath461 with the _ same _ @xmath1 . note that for a limit cycle @xmath452 there is a trivial marginal mode @xmath462 of @xmath463 which is the tangent vector of @xmath464 in @xmath30 : @xmath465 to measure the ` distance ' between a vector @xmath47 and the limit cycle @xmath452 , we introduce @xmath466 and we simply say `` @xmath47 is close to @xmath452 '' when @xmath467 is small . we can now show the discrete versions of the claims in sec . [ sec - scenario ] : * claim 1@xmath468 * _ suppose there is a limit cycle @xmath452 of periodicity @xmath469 , with no nontrivial marginal modes . if ( s1 ) and ( u1 ) hold with this limit cycle @xmath452 , then there should be a _ unique _ relevant mode for this @xmath452 . conversely , if @xmath452 has a unique relevant mode , then ( s1 ) and ( u1 ) hold with this @xmath452 , at least for all the initial data sufficiently close to @xmath452 . ( except , of course , for the exactly critical initial data . ) _ * claim 2@xmath468 * _ if the relevant mode is unique , the black hole mass satisfies @xmath470 for slightly supercritical solutions , where @xmath1 is the eigenvalue of the unique relevant mode of @xmath456 . _ here `` self - similarity '' in ( s1 ) and ( s2 ) should be understood as `` discrete self - similarity . '' rough idea of the ` proof ' is as follows ( see the following section for rigorous results ) . we reduce the problem of a limit cycle to that of a fixed point by considering the poincar section @xmath471 a codimension - one submanifold transverse to the limit cycle , and the poincar map @xmath472 on it induced by the renormalization group flow @xmath473 . ( see fig . [ fig - poin ] . ) @xmath474 defines a discrete dynamical system , for which the limit cycle in @xmath30 corresponds to a fixed point in @xmath475 . the behaviour of the map @xmath472 on the poincar section @xmath475 is the same as that of the map @xmath189 on the whole phase space @xmath30 in the continuously self - similar case . thus it follows from claims 1 and 2 that claims 1@xmath476 and 2@xmath476 hold . in this and next sections , we give rigorous results under certain assumptions . in this section , we consider the local structure of the phase space and show how the poincar map reduces the problem of the limit cycle to that of the fixed point . let us first list assumptions used in the derivation , which are similar to those in the case of continuous self - similarity . to simplify the notation , we often write @xmath477 for @xmath478 in the following . * assumption l1@xmath468 * ( invariant subspaces of @xmath479 ) the tangent space @xmath480 of @xmath30 at @xmath464 is a direct sum of invariant subspaces of @xmath481 : @xmath482 where @xmath483 is a one - dimensional eigenspace spanned by @xmath484 satisfying @xmath485 and @xmath486 and @xmath487 are of dimension @xmath158 and of codimension @xmath488 , respectively , and there exists @xmath489 such that @xmath490 * assumption l2@xmath468 ( smoothness of @xmath473 . ) * the renormalization group transformation @xmath473 considered as a function @xmath491 is of class @xmath492 in a tubular neighborhood of the limit cycle @xmath452 . namely , for @xmath493 and @xmath494 , @xmath495 * remark . * assumption l1@xmath468 together with the existence of @xmath71 in assumption l2@xmath468 is a natural limit cycle version of assumption l1 of sec . [ sub - rgflow ] . assumption l2@xmath468 is slightly stronger than assumption l2 . however , we consider it to be also physically reasonable and adopt it in order to rule out subtle and unexpected behaviour of the flow near the limit cycle @xmath452 . assumption l2@xmath468 implies the following : @xmath496 let us define the poincar map and reduce the problem to that of a dynamical system on a submanifold of codimension one . let @xmath475 be a submanifold of @xmath30 defined by @xmath497 a point moving along a flow curve @xmath498 starting from point @xmath47 on @xmath475 will return to @xmath475 for some @xmath499 , and we can define the _ poincar map _ @xmath500 so that @xmath501 . more precisely , we have the following lemma . [ lem - poin ] under assumptions l1@xmath476 and l2@xmath476 , the following ( i ) and ( ii ) hold . + ( i ) there is a @xmath502-neighborhood of @xmath464 in @xmath475 and the poincar map @xmath472 which is a local @xmath492-diffeomorphism from @xmath475 to @xmath475 satisfying @xmath503 where @xmath504 is @xmath492 and @xmath505 . in particular , the tangent map @xmath506 of @xmath472 at @xmath464 is given by @xmath507 + ( ii ) furthermore , a tubular @xmath502-neighbourhood of @xmath452 is foliated by sections @xmath508 and on each section there exists the poincar map @xmath509 which satisfies all of ( i ) with @xmath510 , @xmath472 , @xmath475 and @xmath464 replaced by @xmath511 , @xmath509 , @xmath512 and @xmath453 . in particular , we have @xmath513 the poincar maps @xmath509 for various @xmath36 are related by @xmath514 where @xmath515 ( @xmath516 $ ] ) is a foliation preserving local diffeomorphism on @xmath30 which maps @xmath512 to @xmath517 . here @xmath518 is a @xmath492-function determined by @xmath97 such that @xmath519 , and satisfies @xmath520 . the tangent map @xmath521 of @xmath522 at @xmath453 satisfies @xmath523 moreover @xmath524 is @xmath492 , as a function on @xmath525 . * the choice of @xmath475 being a linear space , though in general it need not to be , makes the subsequent discussion much simpler . in particular , by reducing the flow on @xmath30 to maps on @xmath475 , we can use the results in the fixed point case . * in fact , ( i ) is included in ( ii ) . one could define @xmath522 first and then the poincar map by @xmath526 . _ ( i ) the proof here is based on the standard treatment of the poincar map ( see , e.g. ( * ? ? ? * section 8) ) . assumption l@xmath527 immediately implies that @xmath475 is _ transverse _ to @xmath452 at @xmath464 , i.e. , the direct sum of the tangent spaces is the whole tangent space , since @xmath528 and @xmath529 . consider a map @xmath530 , @xmath531 . we have @xmath532 . we show that we can implicitly define maps @xmath533 and @xmath534 by @xmath535 for some positive @xmath536 , where @xmath537 is an open ball of center @xmath464 and radius @xmath536 in @xmath30 . assumption l@xmath538 implies that @xmath539 is @xmath492 , and the explicit form of the derivative of @xmath539 at @xmath540 with respect to @xmath541 is @xmath542 where @xmath543 is understood as a frchet derivative . if @xmath544 this has a smooth inverse @xmath545 for @xmath546 , where @xmath547 is the projection of @xmath548 into @xmath475 along @xmath549 . at @xmath550 there exists a smooth inverse ( [ eq : inv f ] ) because of the transversality of @xmath452 and @xmath475 , i.e. , @xmath551 . then the implicit function theorem implies that there exist @xmath552 and a one - to - one map @xmath553 satisfying @xmath554 , and that @xmath555 is as smooth as @xmath539 , i.e. , of class @xmath492 . denoting its partial derivative by @xmath556 , we have @xmath557 the explicit form of the derivatives are given by @xmath558 which follows from @xmath559 where the inverse derivative of @xmath539 , evaluated at @xmath560 , exists and is given by ( [ eq : inv f ] ) . the poincar map on @xmath475 is the restriction of @xmath561 to @xmath562 , which we denote by @xmath472 . the transversality of @xmath563 and @xmath475 implies that this restriction is a one - to - one map . ( [ eq : tp.explicit ] ) implies that @xmath564 where the second equality follows from invariance of @xmath510 under @xmath565 . \(ii ) it follows from assumption 1@xmath476 and @xmath566 that @xmath567 s are invariant subspaces of @xmath568 so that @xmath512 and @xmath569 are transverse . then by the same argument as ( i ) , one can define @xmath522 and the lemma is proved . ( the existence of a tubular neighborhood on which @xmath522 is well defined follows from compactness of @xmath114 . ) @xmath115 though we will consider qualitative behaviour mainly on the poincar section , we show here the qualitative behaviour in the whole phase space , which is the limit cycle version of proposition [ prop-3](i ) . * proposition [ prop-3]@xmath468 * * ( invariant manifolds of @xmath473 at @xmath452 ) * _ under assumptions l1@xmath476 and l2@xmath476 , we have the following : in a tubular neighbourhood of @xmath452 , there exist a stable manifold @xmath570 of codimension @xmath158 , and an unstable manifold @xmath571 of dimension @xmath488 , whose tangent spaces at @xmath453 are @xmath572 and @xmath573 , respectively . _ in terms of the poincar sections , there exist a stable manifold @xmath574 and an unstable manifold @xmath575 of each @xmath576 and are given by @xmath577 the stable and unstable manifolds @xmath570 and @xmath571 of the whole flow are written as @xmath578 _ proof . _ we now show that discrete dynamical system @xmath474 satisfies assumptions l1 and l2 of sec . [ sub - rgflow ] , i.e. , we show that the assumptions hold under the replacement @xmath579 assumption l2 holds for @xmath474 because @xmath472 is of class @xmath492 ( see ( [ eq : p - q])([eq : tp.conti ] ) ) . assumption l1 holds for @xmath474 because of ( [ eq : tp on s ] ) and assumption l1@xmath476 . then proposition [ prop-3 ] of sec . [ sub - rgflow ] implies that in @xmath475 there exists a stable manifold of codimension @xmath158 in @xmath475 and an @xmath158-dimensional unstable manifold of @xmath464 . again , more sophisticated argument guarantees their smoothness ( * ? ? ? * theorem 5.2 , theorem 5.ii.4 ) ; they are of class @xmath492 . next , consider a one - parameter family of discrete dynamical systems @xmath576 . lemma [ lem - poin](ii ) states that @xmath580 is a diffeomorphism from @xmath475 to @xmath512 which preserves the action of the poincar maps on them . this implies that @xmath581 if and only if @xmath582 so that @xmath522 sends the stable manifold @xmath583 to @xmath574 . it also implies that @xmath584 is tangent to @xmath585 so that @xmath586 is tangent to @xmath587 . we have a similar result for the unstable manifold ( consider @xmath588 in place of @xmath472 ) . thus there exist the stable and unstable manifolds @xmath570 and @xmath571 of codimension @xmath158 and dimension @xmath488 in @xmath30 , which are given by ( [ eq : w vs w on s;s ] ) and ( [ eq : w vs w on s;u ] ) , respectively . @xmath115 we now consider quantitative behaviour of the renormalization group flow . we assume the following , which , because of proposition [ prop-3]@xmath476 , asserts that the stable manifold is of codimension 1 and the unstable manifold is of dimension 2 . * assumption l1@xmath476a ( uniqueness of the relevant mode ) * [ ass - l1a ] @xmath486 is one - dimensional , i.e. , @xmath261 . thus @xmath486 is an eigenspace spanned by @xmath262 satisfying @xmath589 where @xmath1 is the unique positive eigenvalue of @xmath477 . consider the time evolution @xmath590 of an initial data @xmath113 in @xmath591 . we are going to trace the evolution on the poincar section @xmath475 . to measure how @xmath113 is deviated from the critical surface , we introduce a decomposition of @xmath113 in the tubular neighbourhood of @xmath452 . lemma [ lem - poin](ii ) implies that for any @xmath113 there is a unique @xmath592 . we can decompose any @xmath113 as ( [ eq : udiff - dcmp - init ] ) on @xmath592 : @xmath593 as in the continuously self - similar case , we generically have @xmath594 . in the special case of @xmath595 , proposition [ prop-4 ] applies for the dynamical system @xmath474 and we immediately have the estimates in proposition [ prop-4 ] . in a general case where @xmath596 ( say , the point @xmath597 of fig . [ fig - poin ] ) , all we have to do is to estimate how @xmath598 and @xmath363 behave before the flow intersects @xmath475 for the first time ( at the point @xmath599 of fig . [ fig - poin ] ) ; after that , we can directly apply proposition [ prop-4 ] . the estimate is given by the following lemma . [ lem - discr - step0 ] under assumptions l1@xmath476 , l2@xmath476 and l1@xmath476a there exist positive constants @xmath502 and @xmath140 which satisfy the following : let @xmath600 , @xmath601 and @xmath602 then we have @xmath603 where @xmath432 is a positive constant . _ let @xmath604 and @xmath605 . we have , for fixed @xmath606 , @xmath607 where we have used ( [ eq : tpbar ] ) at the fourth equality . ( @xmath608 above may depend on @xmath97 or @xmath606 . ) since @xmath609 , we have @xmath610 the first equality gives @xmath611 where we have used the smoothness of @xmath524 in lemma [ lem - poin ] . eq . ( [ eq : lem-9.second ] ) gives @xmath612 since @xmath613 , the @xmath608 terms above are bounded by a constant independent of @xmath606 . @xmath115 now that the local flow structures have been analyzed , we can state a rigorous theorem which corresponds to theorem [ prop - final ] , under moderate assumptions on the global structures of the flow just as in the case of continuous self - similarity . one possibility is : [ prop - final - discr ] suppose assumptions l1@xmath476 , l2@xmath476 , l1@xmath476a , g2 , and g3 hold , and assumption g1 holds for the discrete dynamical system @xmath614 . then the following hold . + ( i ) all the initial data sufficiently close to @xmath452 ( but not exactly on @xmath615 ) once approach , but finally deviate from , @xmath452 . the final fate of these data are either a flat space - time or a black hole , depending on which side of @xmath570 lies the initial data . + ( ii ) consider the time evolution of the initial data ( [ eq : init - decomp ] ) with @xmath616 , @xmath617 , which is on the side of black hole of @xmath570 . then as @xmath363 goes to zero , the mass of the black hole formed satisfies @xmath618 with @xmath363-independent positive constants @xmath439 . here @xmath440 is the initial ( say , adm or bondi - sachs ) mass . _ lemma [ lem - discr - step0 ] reduces the problem into the dynamical system @xmath474 on the poincar section @xmath475 , where theorem [ prop - final ] can be applied . @xmath115 in this section , we focus on the problem of the universality in the original sense of the term in statistical mechanics , based on renormalization group philosophy . we have largely benefitted from the formulation of @xcite . this is the rigorous version of the heuristic argument of universality class given in @xcite . consider general cases in which the pdes do not necessarily satisfy the fundamental assumption of scale invariance , assumption s in sec . [ sec - scenario ] , or self - similar solutions with the scale invariance is not relevant for a critical behaviour ( e.g. having more than one relevant mode ) . methods of renormalization group gives us a clear understanding on which systems can exhibit the same critical behaviour . the main point is that the renormalization group transformation drives the equations of motion to a fixed point , where they gain the scale invariance , and the problem reduce to the cases treated in previous sections . as a result , we understand why and to what extent condition ( u3a ) holds . knowing the _ universality class _ ( the class of models which exhibit the same critical behaviour ) is very important from a physical point of view , because it guarantees that certain models exhibit the same critical behaviour with a model in the same class , whose analysis could be much simpler . ( imagine we want to deal with a realistic matter . it would be almost impossible to know or deal with the exact form of the microscopic interaction . universality could make the problem easier and accessible , by reducing it to that for a simpler model , e.g. perfect fluid . ) suppose unknowns @xmath34 satisfy the pdes of the following form , which is not necessarily scale invariant : @xmath619 let us define the scaling transformation @xmath620 as in ( [ eq : sctrdef ] ) . then @xmath39 satisfies equations @xmath621 where @xmath622 and @xmath623 ( where @xmath624 is the number of component of @xmath92 ) is a set of constants which may depend on @xmath448 and @xmath3 and are chosen so that @xmath625 remains finite and nonzero when @xmath626 . it should be noted that if @xmath627 holds with some @xmath4 the system @xmath628 is invariant under scaling transformation in the sense of sec . [ sec - rg ] . a renormalization group transformation is now defined to be a _ pair _ of a transformation on the eom : @xmath629 , and the following transformation on the @xmath30 : @xmath630 where we added subscript @xmath92 to show explicitly the rgt s dependence on @xmath92 . see fig . [ fig - withv ] . we emphasize that the renormalization group transformation is now to be considered as a pair of two transformations on different spaces : the space of eom , and the space of initial data ( phase space @xmath30 ) . one can easily show the ( semi)group property @xmath631 or @xmath632 this states that long time evolution problem of the original equation defined by @xmath92 is the composition of finite time evolution problems of equations with ( in general ) different pdes @xmath633 . we now turn to the problem of determining the universality class . concretely , we consider the following situation . suppose , with a certain choice of @xmath448 , @xmath3 , and @xmath4 , @xmath625 approaches a fixed point ( i.e. a fixed function of its arguments ) @xmath634 asymptotically : @xmath635 it follows from definition ( [ eq : uc - l - scaled ] ) of transformation of @xmath92 that the fixed point @xmath636 satisfies the scale invariance in the sense of sec . [ sec - rg ] . suppose , in addition , there exists a fixed point ( or a limit cycle ) of @xmath637 which has a unique relevant mode . and the question is : under what circumstances do @xmath92 and @xmath638 exhibit the same critical behaviour ( i.e. in the same universality class ) ? at first glance the question looks rather trivial ; a naive guess would be that they are always in the same universality class , because @xmath625 approaches @xmath638 . however , there is a rather subtle problem involved here : values of @xmath625 and @xmath638 at the _ same argument _ will approach each other as @xmath420 ; but this does not necessarily mean that their difference at their respective ( generally different ) @xmath42 , which are defined by the time evolution by @xmath625 and @xmath638 , goes to zero . in fact , if one starts from the _ same initial condition _ this usually does not hold ( e.g. the critical value @xmath639 is shifted due to @xmath640 ) ; we have to prove structural stability for these systems . in the following , we provide a sufficient condition under which the above naive guess does hold . to simplify the notation , we define @xmath641 , and introduce @xmath642 which measures the `` strength '' of the difference @xmath643 . the strength @xmath644 should be so chosen that the following assumption 2c should hold . we write the time evolution operator in terms of @xmath92 by @xmath645 , and that by @xmath638 by @xmath473 ; and their tangent maps by the same superscript convention . we assume that the finite time evolution operators are regular with respect to this difference . concretely , we assume ( in addition to assumption l1 of and assumption l2 sec . [ sub - rgflow ] , which are about time evolutions _ without _ @xmath101 ) : * assumption l2v ( regularity in the presence of @xmath101 ) * there are positive @xmath646 and @xmath151 such that the following hold for @xmath647 and @xmath648 : @xmath649 these guarantee the structural stability of the flow in the neighbourhood of @xmath55 and @xmath650 . concretely , we have the following proposition , which corresponds to proposition [ prop-3 ] ( i ) of sec . [ sub - rgflow ] . we denote by @xmath157 and @xmath156 the invariant subspaces of @xmath152 , _ in the absence of @xmath101_. [ prop-3prime ] suppose there is a fixed point @xmath55 of @xmath637 with a unique relevant mode , i.e. assumptions l1 , l2 , l1a of sec . [ sub - rgflow ] hold , and @xmath651 for @xmath652 . suppose further that assumption l2v holds . then there exist small positive constants @xmath653 ( how small is specified in the proof ) such that if @xmath654 then for any @xmath172 with @xmath655 , there exists @xmath249 such that @xmath656 as @xmath285 . we define @xmath657 as we did in ( [ eq : defcrit.sol1 ] ) . note that now the definition depends on @xmath101 . _ sketch of the proof of proposition [ prop-3prime ] . _ the proof is done by bleher - sinai argument as was done for proposition [ prop-3 ] . we sketch the proof by mainly pointing out necessary modifications . _ step 1 . _ same as the proof of proposition [ prop-3 ] . we employ the same decomposition ( [ eq : rg.qual.dcmp ] ) , and for each @xmath169 , try to find @xmath268 such that @xmath658 _ step 2 . _ by ( [ eq : univ.reg ] ) , we have @xmath659 calculating @xmath660 just as in ( [ eq : detail.phys.ex2 ] ) , we this time obtain [ cf . ( [ eq : rg.qual.4 ] ) , ( [ eq : rg.qual.5 ] ) , ( [ eq : rg.qual.6 ] ) ] @xmath661 where @xmath662 is the presence of @xmath663 . _ step 3 . continuity . _ same as the proof of proposition [ prop-3 ] . assumption 2c guarantees the continuity in @xmath267 , even in the presence of @xmath101 . _ step 4 . solving the recursion . _ same as the proof of proposition [ prop-3 ] . our goal is to find sequences @xmath283 for which ( a)(c ) of step 4 of the proof of proposition [ prop-3 ] hold . _ step 5 . proof of ( a)(c ) . _ this part significantly differs from the proof of proposition [ prop-3 ] , due to the presence of the term @xmath663 in the recursion . this time , we define @xmath664 as follows . first define @xmath665 and choose @xmath666 and @xmath667 as @xmath668 then define @xmath669 recursively , starting from @xmath670 as @xmath671 or ( @xmath672 ) @xmath673 we now show that ( a)(c ) is satisfied by this choice of @xmath674 . first , dividing the sum over @xmath675 in ( [ eq : rn - def - v ] ) according to @xmath676 , ( [ eq : rn - def - v ] ) implies ( @xmath677 is non - increasing in @xmath675 by definition ) @xmath678 three terms on the rhs goes to zero ( as @xmath285 ) , either because @xmath679 or @xmath680 . this proves ( a ) . similar reasoning also shows the following uniform bound : @xmath681 now we turn to ( b ) and ( c ) . first note that it is sufficient to prove [ cf . ( [ eq : bscond21 ] ) , ( [ eq : bscond22 ] ) ] @xmath682 because we are taking @xmath669 and because @xmath683 , it is then sufficient to prove @xmath684 eq.([eq : rcond1.v ] ) is satisfied , by our definition of @xmath674 . and ( [ eq : rcond2.v ] ) is satisfied , if @xmath685 it is not difficult to see ( by induction ) that ( [ eq : rcond3.v ] ) is in fact satisfied under our choice of @xmath666 and @xmath667 . this completes step 5 . @xmath115 to complete the story , we have to make additional assumptions on the global behaviour of the flow in the presence of @xmath101 . concretely we assume : * assumption gv ( global behaviour in the presence of @xmath101 ) * consider the time evolution of initial data @xmath686 in the presence of @xmath101 . then there exist @xmath687 such that if @xmath688 then assumptions g1 , g2 , g3 hold for suitably chosen @xmath689 . in the above , @xmath690 was defined in ( [ eq : deapm ] ) , in terms of dynamics without @xmath101 . under these preparations , we can now state our theorem . [ theorem-11 ] suppose there exist @xmath691 and @xmath692 satisfying @xmath693 ) , where @xmath638 is a fixed point of the scaling transformation : @xmath694 . suppose there is a fixed point or limit cycle @xmath55 of @xmath637 with a unique relevant mode , i.e. assumptions l1 , l2 , l1a of sec . [ sub - rgflow ] hold . moreover , suppose assumption 2c and assumption gv hold . then , + ( i ) if there exists @xmath695 such that for @xmath194 we have ( 1 ) @xmath651 as @xmath420 and ( 2 ) @xmath696 for sufficiently small @xmath697 , then the system @xmath628 exhibits qualitatively the same critical behaviour as the system @xmath698 . ( i.e. there is a critical behaviour and near - critical solutions once approach and then deviate from @xmath55 ) . + ( ii ) if in addition @xmath699 sufficiently fast ( for example , @xmath700 integrable in @xmath36 on @xmath701 is sufficient ) , the system @xmath628 exhibits quantitatively the same critical behaviour as the system @xmath698 ( i.e. the same critical exponent ) . * remark . * in assumption gv , @xmath101 in fact means @xmath702 at the instant when the flow exits the perturbative region through the tube @xmath703 : the local analysis which corresponds to that of proposition [ prop-4 ] shows that the near ( but not exactly ) critical solutions once approach and then deviate from @xmath55 , and they exit from the perturbative region through the tubes @xmath356 . because @xmath704 as @xmath420 , the effect of @xmath101 goes to zero as @xmath705 , and could be neglected . however , we made the assumption gv in order to avoid pathological situation where the slight effect caused by this small @xmath706 leads to qualitatively different global behaviour . ( because the solution is expected to blow up , we can not expect nice regularity in @xmath101 in this region . ) theorem [ theorem-11 ] explains how and why claim ( u3a ) of sec . [ sec - scenario ] holds . the mere condition @xmath707 is not sufficient to guarantee the universality . the following proof in fact suggests the deviation in the critical behaviour : @xmath708 where @xmath709 \label{eq : univ - deviation2 } , \ ] ] with positive constants @xmath710 . concretely , @xmath711 will be a power of logarithm in @xmath712 for @xmath713 , and @xmath714 for @xmath715 ( @xmath716 ) . the deviation in ( [ eq : univ - deviation3 ] ) is greater than logarithmic , although it does not change the critical exponent ( if we define it as the limit of the logarithmic ratio of both hand sides ) . _ proof of theorem [ theorem-11 ] . _ proposition [ prop-3prime ] shows the existence of a critical solution . we now trace the time evolution of off - critical solutions . we employ the scheme of sec . [ subsub-rg.quan2 ] . that is , we decompose the initial data @xmath47 as [ cf . ( [ eq : udiff - dcmp - init ] ) ] @xmath717 and for @xmath718 decompose [ cf . ( [ eq : udiff - dcmp - non0 ] ) ] @xmath719 and further decompose @xmath216 as in ( [ eq : udiff - dcmp - f ] ) : @xmath720 note that @xmath110 here means the critical solution _ in the presence of @xmath101 _ ( and its time evolution ) . then arguing as we did in deriving ( [ eq : ap.ev1 ] ) , we now obtain , using ( [ eq : iniv.reg0 ] ) @xmath721 which implies @xmath722 now using ( [ eq : univ.regt ] ) and ( [ eq : a.pert ] ) , @xmath723 ( f ) + \bigl [ { \cal t}_{\sigma , u } - { \cal t}_{\sigma , u^ { * } } \bigr ] ( f ) \nonumber \\ & & { } + { \cal t}_{\sigma , u^ { * } } ( f ) \nonumber \\ & = & { \cal t}_{\sigma , u^ { * } } ( f ) + { { \bar{o } } } ( k_{3 } \|\|\| v ( u ) \|\|\| \cdot \\| f \\| ) \nonumber \\ & & { } + { { \bar{o } } } ( 6 k_{1 } \cdot \\| u - u^ { * } \\| \cdot \\| f \\| ) \end{aligned}\ ] ] so decomposing @xmath216 into @xmath226 and @xmath227 in ( [ eq : univ.fnplone ] ) , we have recursions for @xmath391 and @xmath398 , @xmath397 : @xmath724 \label{eq : rec - ineq - v1 } \\ & & f_{n+1 } \leq e^{-\bar{\kappa}\sigma } f_{n } \nonumber \\ & & \quad + ( \| b_{n } \| \vee f_{n } ) \bigl [ k_{1 } ( \| b_{n } \| \vee f_{n } ) + 6 k_{1 } g_{n } + k_{3 } v_{n } \bigr ] \label{eq : rec - ineq - v2}\end{aligned}\ ] ] as seen , the only difference here from ( [ eq : ap.ev42 ] ) and ( [ eq : ap.ev41 ] ) is the appearance of @xmath725 . the recursion is of the form considered in lemma [ lem-5 ] ( @xmath726 playing the role of @xmath402 there ) , and this proves the quantitative estimate which corresponds to proposition [ prop-4 ] . in particular , lemma [ lem-5 ] guarantees that the presence of @xmath727 does not affect the critical behaviour . the above concludes the local analysis of the flow in the neighbourhood of @xmath55 , which corresponds to that in sec . [ subsub-rg.qual ] and [ subsub-rg.quan2 ] . the analysis of global behaviour , corresponding to that of sec . [ subsub-rg.global ] , can be carried out in a similar manner under the assumption gv . @xmath115 * remark . * by refining the the proof of proposition [ prop-3prime ] , we can show in general is of definite sign , we can heuristically get a lower bound on the difference , and show @xmath728 this estimate gives an answer to the following question considered in @xcite : fix the value of @xmath7 to the critical value when @xmath650 ( i.e. @xmath729 ) , and vary the magnitude of @xmath101 . what kind of critical behaviour can we observe ? the above estimate , together with theorem [ theorem-11 ] shows that we will observe the _ critical exponent @xmath105 , because @xmath730 where we used theorem [ theorem-11 ] in the first step , and ( [ eq : pcvestimate2 ] ) in the second . ] ( @xmath731 denotes the critical value without @xmath101 ) @xmath732 we conclude this section by giving two examples to which the above analysis can be applied . note that the following arguments are not rigorous in contrast to what we have been doing , in the sense that no rigorous verification of assumptions are given . the first example is a perfect fluid , with equation of state ( @xmath539 is a given function ) @xmath733 the eom for this system is given by ( [ eq : eom - fll.rt ] ) of the next section , supplemented by the above equation of state . when @xmath734 , this is a usual perfect fluid , to be studied in detail beginning with the next section . it will be shown that the system with @xmath734 exhibits a critical behaviour governed by a self - similar solution @xmath55 , which is a fixed point of a renormalization group transformation induced from a scaling transformation @xmath735 whose effect on @xmath736 and @xmath7 is : @xmath737 where @xmath738 is arbitrary due to gauge invariance . the eom ( [ eq : eom - fll.rt ] ) is invariant under the scale transformation , and so is the equation of state with @xmath734 : @xmath739 . ( here we write @xmath740 because this is the sixth of eom . ) thus the system with @xmath734 is an example to which our scenario in sec . [ sec - scenario ] and sec . [ sub - rgflow ] applies . our question is to determine the models ( forms of @xmath539 ) which belong to the same universality class as the @xmath734 model . now under the scaling transformation the eom in question , @xmath741 , is transformed into @xmath742 where @xmath743 at the fixed point in question , @xmath744 and @xmath745 are finite smooth functions of their argument ( sec . [ sec - ss ] ) . so we can expect assumption 2c to hold in a neighbourhood of the fixed point , by simply taking @xmath746 . then theorem [ theorem-11 ] implies that the model belongs to the same universality class as that of @xmath734 , as long as ( for fixed @xmath747 ) @xmath748 this is the sufficient condition we are after . an example of @xmath539 which shows the same critical behaviour as @xmath734 is given by @xmath749 with @xmath750 . let us present another example , i.e. spherically symmetric collapse of a scalar field . we consider a scalar field @xmath751 with energy momentum tensor @xmath752 where @xmath753 is the metric tensor , @xmath101 is any , say , @xmath754- , function , and the abstract index notation is used . the line element of a general spherically symmetric space - time can be written as @xmath755 we fix the gauge by the condition @xmath756 . the equations of motion can be written in terms of retarded time @xmath34 , area radius @xmath21 and field variable @xmath757 as @xmath758=0 $ ] , where functional @xmath92 labeled by @xmath101 is given by @xcite @xmath759\equiv d_vh -\frac12 ( h-\av h)\;\frac { \partial{\bar g_v}[h]}{\partial r } + \frac12 r\ , g[h]\ , v'(\av h),\ ] ] where @xmath760 , @xmath761 , @xmath762 , and @xmath763 are functionals defined as @xmath764&\equiv & \exp\left({-4\pi r \aav{r^{-1}(f-\av f)^2}}\right ) , \nonumber\\ { \bar g_v}[f]&\equiv & \left\langle{g[f]\,(1 - 8\pi r^2v(\av f))}\right\rangle , \nonumber\\ d_v f&\equiv & \frac{\partial f}{\partial u } -\frac12{\bar g_v[f]}\frac{\partial f}{\partial r}. \label{eq : scalar : ein.,integro.}\end{aligned}\ ] ] in the equation of motion , @xmath758=0 $ ] above , metric variables @xmath761 and @xmath762 have been solved as functionals @xmath761 and @xmath765 , respectively , of @xmath275 , and the scalar field is given by @xmath766 . hereafter , for a functional @xmath78 , @xmath767 $ ] denotes a function determined by a function @xmath275 and @xmath767(a , b)$ ] denotes the value of function @xmath767 $ ] at @xmath768 . for functionals @xmath769 , we rewrite the values @xmath767(e^{-s}u , e^{-\beta s}r)$ ] in terms of @xmath770 using the relation @xmath771 : @xmath772(e^{-s}u , e^{-\beta s}r ) & = & \left ( g[h^{(s ) } ] \right ) ^{e^{-2\alpha s}}(u , r ) , \\ { \bar g_v}[h](e^{-s}u , e^{-\beta s}r ) & = & \bigl \langle \left ( g[h^{(s ) } ] \right ) ^{e^{-2\alpha s } } \nonumber\\ & & \hspace{-6em } { } \times \left ( 1 - 8\pi r^2 v^{(s)}(e^{-\alpha s}\av { h^{(s ) } } ) \right ) \bigr \rangle\,(u , r ) , \\ ( d_v h)(e^{-s}u , e^{-\beta s}r ) & = & \bigl ( e^{(-\alpha+1 ) s } \frac{\partial h^{(s)}}{\partial u } \nonumber\\ & & \hspace{-6em } -\frac12 e^{(-\alpha+\beta ) s } { \bar g_{v^{(s)}}[h^{(s ) } ] } \cdot \frac{\partial h^{(s)}}{\partial r } \bigr ) ( u , r ) . \label{eq : fnls - scaled - last}\end{aligned}\ ] ] where @xmath773 we find the renormalized functional @xmath774 with parameters @xmath448 , @xmath3 and @xmath4 by substituting ( [ eq : fnls - scaled - first])([eq : fnls - scaled - last ] ) into @xmath775(e^{-s}u , e^{-\beta s}r)$ ] and reinterpreting it as @xmath776(u , r)$ ] , the value of function @xmath776 $ ] at @xmath777 . let us find fixed points , i.e. , @xmath778 satisfying @xmath779 . it follows from ( [ eq : fnls - scaled - last])([eq : fnls - scaled - last ] ) that we must choose @xmath780 , @xmath781 . then ( [ eq : fnls - scaled - first])([eq : fnls - scaled - last ] ) and ( [ eq : v - scaled ] ) read @xmath782(e^{-s}u , e^{-s}r)&= & g[h^{(s)}](u , r ) , \\ { \bar g_v[h]}(e^{-s}u , e^{-s}r)&= & { \bar g_{v^{(s)}}}[h^{(s)}](u , r ) , \\ ( d_v h)(e^{-s}u , e^{-\beta s}r)&= & e^{-s}(d_{v^{(s ) } } h^{(s)})(u , r ) , \label{eq : av - scaled-2}\end{aligned}\ ] ] and @xmath783 transformation of @xmath92 is thus given by @xmath784(u , r ) & \equiv & e^{(\gamma-1 ) s}l[h](e^{-s}u , e^{-\beta s}r ) \nonumber\\ & = & e^{(\gamma-1 ) s}l_{v^{(s)}}[h^{(s)}](u , r ) . \label{eq : l - scaled}\end{aligned}\ ] ] choosing @xmath785 , we have @xmath786 , and from ( [ eq : v - scaled-2 ] ) we find that @xmath787 is the only fixed point , i.e. , @xmath788 is the only model having scale invariance , among the models considered here . choptuik s original result @xcite , which was further confirmed by a theoretical work by gundlach @xcite , was that the system @xmath788 shows critical behaviour with this choice of parameters , @xmath780 , @xmath781 , for the rgt . the system has a limit cycle which should have a unique relevant mode , according to our scenario in previous sections . it follows directly from ( [ eq : v - scaled-2 ] ) and ( [ eq : l - scaled ] ) that @xmath789 and the convergence is so fast that theorem [ theorem-11 ] is applicable . this shows that potential @xmath101 is an _ irrelevant _ term which vanishes asymptotically . therefore , all @xmath778 s are in the same universality class as @xmath787 , i.e. , all scalar field models with arbitrary smooth potential @xmath101 exhibits the same critical behaviour with the same critical exponent as the minimally coupled massless scalar field . we now begin our concrete analysis of gravitational collapse of perfect fluid , as a case study of renormalization group ideas . the line element of any spherically symmetric space - time is written as @xmath790 we assume the matter content is a perfect fluid having energy - momentum tensor @xmath791 where @xmath736 is the density , @xmath7 is the pressure , and @xmath792 is a unit timelike vector whose components are given by : @xmath793 here @xmath101 is the 3-velocity of fluid particles . introducing locally @xmath794 and @xmath795 , the equations of motion ( eom ) are given in this coordinate by : [ eq : eom - fll.rt ] @xmath796 @xmath797 @xmath798 @xmath799 @xmath800 in the above , first three are obtained by taking linear combinations of @xmath801 , @xmath802 , @xmath803 components of einstein equation . the last two are obtained by taking linear combinations of the `` bianchi identity '' @xmath804 . we have to supplement the above by the equation of state for the fluid . we consider the case @xmath805 where @xmath806 is a constant ( adiabatic index ) . in terms of variables @xmath807 , and introducing @xmath808 we can write the equations of the system in an autonomous form , which makes the scale invariance of the system transparent : [ eq : eom - fll ] @xmath809 @xmath810 @xmath811 @xmath812 @xmath813 only four out of the above five equations are independent . more precisely , the eqn . ( [ eq : eom - fll.5 ] ) is automatically satisfied by solutions of the set ( [ eq : eom - fll.3 ] ) , ( [ eq : eom - fll.4 ] ) , ( [ eq : eom - fll.1 ] ) and ( [ eq : eom - fll.2 ] ) , as long as they satisfy a boundary condition @xmath814 . in view of this , we pick up the above set of four equations as our basic equations of motion , and use ( [ eq : eom - fll.5 ] ) as an auxiliary equation at appropriate stages in the following . the only coordinate transformation which preserves the form of ( [ eq : le ] ) is @xmath815 which corresponds to @xmath816 where @xmath817 with @xmath818 . under this transformation , the variables @xmath275 transform to @xmath819 , where @xmath820 eqs.([eq : eom - fll ] ) are of course invariant under this transformation . fixing the value of @xmath158 at a point , for example , @xmath821 , of each constant @xmath36 line determines the coordinate system completely . we shall retain the degree of freedom for the time being . one note : the gauge degree of freedom present in our system enables us to take @xmath19 in our formulation of sec . [ sec - rg ] , _ without _ introducing new variables . to carry out the first step of our scenario , we first have to find out fixed points ( i.e. self - similar solutions ) of rgt . in this section , we find out ( almost ) all of the self - similar solutions of the eom . because this paper is intended to present the `` renormalization group '' approach , we here mildly try to exhaust all possible self - similar solutions . we first require that the space - time is self - similar , i.e. that @xmath158 and @xmath597 depend only on @xmath747 : @xmath822 . conversely , it can be shown ( see appendix ? ? ) that any spherically symmetric self - similar space - times can be expressed in that form if a freedom of coordinate transformation ( [ eq : ct ] ) is used . then it follows from ( [ eq : eom - fll ] ) that @xmath823 and @xmath824 are also functions of @xmath747 only : @xmath825 . in this sense , the space time we are interested in is a _ fixed point _ of the rgt . the equations for a self - similar solution , @xmath57 , are then given by omitting terms containing derivatives of @xmath36 in ( [ eq : eom - fll ] ) : [ eq : eom - ss ] @xmath826 @xmath827 @xmath828 @xmath829 @xmath830 from ( [ eq : eom - ss.3 ] ) and ( [ eq : eom - ss.5 ] ) , one obtains an algebraic identity @xmath831 one could eliminate one variable ( e.g. @xmath597 ) from ( [ eq : eom - ss ] ) using the above identity . however , to avoid making equations too complicated , we keep using four variables ( @xmath832 ) in the following , using the above ( [ eq : alg - ss ] ) as a check at appropriate stages of our numerical calculation . under the self - similar ansatz , the coordinate freedom ( [ eq : ct ] ) reduces to @xmath833 with constant @xmath675 , which corresponds to the translation of @xmath747 [ the other transformations alter the constant @xmath747 lines in @xmath93 space ] . this freedom of coordinate transformation allows one to adjust the value of @xmath158 at a given point arbitrarily . we make use of this freedom , and fix the coordinate system by requiring that the sonic point ( see below ) be at @xmath821 . the behaviour of self - similar solutions has been extensively discussed by bogoyavlenskii in @xcite , followed by other works @xcite . we here briefly outline our analysis , in our coordinate system , for completeness . we are interested in self - similar solutions which satisfy the following two properties : ( i ) : : the self - similar solution is analytic ( or at least smooth , in the sense of having an asymptotic expansion to all orders ) for all @xmath834 , ( ii ) : : the space - time and the matter are regular , @xmath835 and @xmath650 , at the center ( @xmath836 ) . the reason why we require these may be summarized as follows . in this paper , we are interested in self - similar solutions which could represent the space - time structure and radiation fluid profile at the edge of the formation of a black hole . such a critical space time is formed from a suitable initial condition through time evolution given by the eom , and we have introduced the self - similar coordinate @xmath837 in order to absorb any singularity in the process . thus , the solution should be smooth for all @xmath747 . this is the first condition . the second ( boundary ) condition is a physical , natural one ; @xmath835 at the center guarantees that the space time has no physical singularity at the center . @xmath650 at the center should hold for any spherically symmetric problem , except for the case where there is a source of fluid at the center . now , ( [ eq : eom - ss ] ) is a set of ode s for four variables ( @xmath832 ) , which satisfies the lipschitz condition except at the so - called _ sonic point _ ( explained in section [ sec-ss.sonic ] ) . so , there are three types of possible singularities for ( [ eq : eom - ss ] ) : @xmath836 , @xmath838 and the sonic point . we consider each of them in the following , but before that , we conclude this subsection with the following simple observation , which will be repeatedly used in subsequent sections . [ lem-3.1 ] any solution of ( [ eq : eom - ss ] ) is monotonically non - decreasing , as long as @xmath839 . a direct consequence of the lemma is that any solution , which is regular at the center ( i.e. @xmath840 as @xmath841 ) , must satisfy @xmath842 for all @xmath834 . _ proof of lemma [ lem-3.1 ] . _ follows from ( [ eq : eom - ss.3 ] ) ; the last term on the rhs is nonnegative , so we have @xmath843 @xmath115 eqs . ( [ eq : eom - ss.1 ] ) and ( [ eq : eom - ss.2 ] ) can be written in the form : @xmath844 where @xmath244 , @xmath363 , @xmath845 , @xmath846 , @xmath847 , @xmath539 are functions which do not depend on derivatives of @xmath848 and @xmath101 . solving the above in favor of @xmath849 and @xmath850 , the resulting equation violates the lipschitz condition at the _ sonic point _ , where the `` determinant '' of the coefficient matrix of @xmath849 and @xmath850 vanishes : @xmath851 physically , the sonic point is characterized by the fact that the velocity of fluid particles seen from the observer on the constant @xmath747 line is equal to the speed of sound @xmath852 . because generic solutions at the sonic point can be singular , the analytic condition ( i ) required at the sonic point places a severe constraint on possible form of solutions and in the end leaves only one free parameter , as we now explain . the sonic point is characterized by the vanishing of the determinant , ( [ eq : det0 ] ) . in order to have finite derivatives , @xmath849 and @xmath850 , at the sonic point , the rows of ( [ eq : eom - ss.12form ] ) must be proportional to each other : @xmath853 these two conditions , ( [ eq : det0 ] ) and ( [ eq : ss.rows ] ) , together with the identity ( [ eq : alg - ss ] ) between @xmath832 , enable us to express values of @xmath597 , @xmath158 , @xmath854 at the sonic point in terms of that of @xmath855 ( we write @xmath856 , etc . for values of these functions at the sonic point ) . to find out the explicit forms of these expressions , we first list up the following properties to be satisfied at the sonic point : 1 . @xmath857 , because of its definition in terms of lapse function . 2 . @xmath858 , from lemma [ lem-3.1 ] of section [ sec-ss.cond-ss ] . 3 . the derivatives @xmath849 and @xmath850 should exist at the sonic point . in other words , first order expansion coefficients , in particular those of @xmath854 and @xmath101 must exist as real numbers . note that ( [ eq : det0 ] ) , being quadratic in @xmath859 , in general allows two solutions for @xmath859 . however , one of them leads to @xmath860 , and is excluded from the second requirement above . the first and third requirements together restrict allowed values of @xmath861 to @xmath862 . as a result , the zeroth order expansion coefficients are given by : @xmath863 once these values at the sonic point are given , we can then determine the higher coefficients of power series expansion of @xmath864 , and @xmath101 with respect to @xmath747 using the above conditions and ( [ eq : eom - ss ] ) . to avoid making presentation too complicated , we do not give explicit expressions , but only remark that the equations for the first order expansion coefficients , given in terms of @xmath865 , have two solutions in general ; some of which are not allowed for some values of @xmath861 due to the third requirement above . the self - similar solution which is regular at the sonic point is thus characterized by a single parameter @xmath855 . in searching for desired self - similar solutions , it becomes desirable to understand the correct asymptotic behaviour of solutions as @xmath866 . behaviour of solutions of ode s ( [ eq : eom - ss ] ) , as @xmath868 , can be most easily understood by introducing a new variable @xmath869 and rewriting ( [ eq : eom - ss ] ) in terms of @xmath870 . it is easily seen that @xmath871 is a stationary ( fixed ) point of ( [ eq : eom - ss ] ) , i.e. all the @xmath747-derivatives vanish . moreover , linear analysis around this stationary point shows that there is one unstable direction for this stationary point , namely in the direction of @xmath872 . thus , a generic solution which satisfies the analyticity condition at the sonic point will diverge as @xmath841 , and only a small number of solutions with carefully chosen values of @xmath861 can satisfy the regularity condition at @xmath867 ; thus leaving only discrete acceptable solutions . by similar reasoning , it is easily seen that any self - similar solution _ which satisfies our boundary condition at the center _ ( i.e. has a smooth derivative ) , allows a ( possibly only asymptotic ) power series expansion in @xmath873 at @xmath874 , and it is easily seen that the asymptotic behaviour is @xmath875 @xmath876 with coefficients @xmath877 satisfying the relation : @xmath878 it is easily seen that @xmath879 is a stationary point ( parameterized by two free parameters @xmath880 ) of ( [ eq : eom - ss ] ) . linear analysis around the stationary point shows that this stationary point becomes an attracting node for @xmath881 as will be explained in section [ sec-ss.results ] , most of the self - similar solutions which are analytic for @xmath882 found numerically satisfy this condition , and thus are seen to be attracted to this stationary point . based on considerations presented in the above , we proceed as follows in order to find out desired self - similar solutions . we first fix the value @xmath861 at the sonic point @xmath821 , and use the power series expansion there ( section [ subsubsec-ss.sonicexp ] ) to derive the derivatives . then starting from the sonic point @xmath821 , we solve ode s ( [ eq : eom - ss ] ) towards @xmath867 , using runge kutta fourth order integrator , with reliable error estimates . if at some @xmath883 ( 1 ) either @xmath884 or ( 2 ) @xmath885 , we stop solving and conclude that this @xmath861 does not give rise to a desired self - similar solution . the first case ( @xmath886 ) is excluded based on the lemma [ lem-3.1 ] of section [ sec-ss.cond-ss ] the second case implies the existence of another sonic point between @xmath821 and @xmath867 . such a solution with more than one ( but with finite number of ) sonic points can always be found by searching for its part between its left - most sonic point and @xmath867 , and then solving ( [ eq : eom - ss ] ) from the left - most sonic point to the right ( towards @xmath838 ) . however , it is quite unlikely that such solutions with more than one sonic points exist , because the existence of only one fixed point already reduces the freedom of parameters to discrete ones , and other possible sonic point will further reduce it . our numerical results show that each allowed @xmath861 ends in either of these two possibilities . moreover it is seen , both numerically and mathematically rigorously , that a desired self - similar solution exists at those values of @xmath861 , where @xmath887 leads to the case ( 1 ) and @xmath888 leads to the case ( 2 ) . we have searched for self - similar solutions which satisfy the above conditions ( i ) and ( ii ) numerically , for all the allowed values of @xmath861 . as has been extensively studied in @xcite , there exists a sequence of self - similar solutions , each of which is characterized by the number of zeros of @xmath101 . the number of zeros is an integer starting from @xmath139 , and the one with exactly one zero is the solution cited in @xcite , which we call the evans coleman solution . due to a numerical difficulty , we could find only self - similar solutions with odd number of zeros of @xmath101 . in fig . [ fig4.1 ] , we show the profile of the evans coleman solution , which will play the central role in later sections . this is the same as fig . 1 of @xcite except for @xmath448 which is due to the coordinate condition for @xmath26 . we studied perturbations around self - similar solutions by two different methods . the first one is to directly find out eigenmodes by solving the system of ode s ( [ eq : eom - eigenmodes ] ) below , which characterizes the eigenmodes . the second one is to apply the so called lyapunov analysis . the former has the advantage that it reduces the problem to that of solving ode s , and enables us fairly accurate estimate of eigenvalues , but has the disadvantage that it is almost impossible to search for _ every possible _ eigenvalues . the latter has an advantage that it is quite suited to finding out eigenvalues in descending order of their real part , but has disadvantages that ( 1 ) eigenvalue estimates are not so accurate , and ( 2 ) we have to solve a pde . however , we here emphasize that the pde we have to solve for the lyapunov analysis is a very regular one , and can be handled by standard techniques for solving pde s . in this section , we explain the first method , that is , direct search for perturbations by shooting method . to simplify the notation , we express perturbations in terms of perturbations of @xmath889 . for example , the @xmath890-version of ( [ eq : eom - pp ] ) in fact means @xmath891 where @xmath892 perturbation equations are obtained by taking the first order variation in ( [ eq : eom - fll ] ) from the self - similar solution @xmath893 considered : @xmath894 where @xmath275 represents each of @xmath895 . in the following , to simplify the notation , we write @xmath101 for @xmath824 and write @xmath896 for perturbations . explicitly , we have : @xmath897 \left ( \begin{array}{c } \bar{a}_{\rm var } \\ \bar{n}_{\rm var } \\ \bar{\omega}_{\rm var } \\ v_{\rm var } \end{array } \right ) \nonumber \\ & & \hspace{10 mm } = \left ( \begin{array}{cccc } g_1 & 0 & g_3 & g_4 \\ h_1 & 0 & h_3 & 0 \\ e_1 & e_2 & e_3 & e_4 \\ f_1 & f_2 & f_3 & f_4 \end{array } \right ) \left ( \begin{array}{c } \bar{a}_{\rm var } \\ \bar{n}_{\rm var } \\ \bar{\omega}_{\rm var } \\ v_{\rm var } \end{array } \right ) \end{aligned}\ ] ] with coefficients given in terms of self - similar solutions as @xmath898 @xmath899 @xmath900 @xmath901 @xmath902 @xmath903 we also have @xmath904 we note that there is a gauge freedom due to the coordinate transformation of order @xmath140 , namely ( [ eq : ct:(s , x ) ] ) with @xmath905 . the transformation ( [ eq : h ] ) now becomes [ taking @xmath906 terms ; @xmath907 @xmath908 this in particular means that one can always require @xmath909 for a fixed @xmath910 . we consider eigenmodes of the form @xmath911 , with @xmath80 being a constant . substituting this form into ( [ eq : eom - var ] ) yields a set of linear , homogeneous first order ode s for @xmath912 . explicitly , we have @xmath913 \left ( \ ! \begin{array}{c } \bar{a}_{\rm p } \\ \bar{n}_{\rm p } \\ \bar{\omega}_{\rm p } \\ v_{\rm p } \end{array } \ ! \right ) \nonumber \\ & & { } \qquad = - \kappa \left ( \begin{array}{cccc } 0 & 0 & 0 & 0 \\ 0 & 0 & 0 & 0 \\ 0 & 0 & a_s & b_s \\ 0 & 0 & c_s & d_s \end{array } \right ) \left ( \begin{array}{c } \bar{a}_{\rm p } \\ \bar{n}_{\rm p } \\ \bar{\omega}_{\rm p } \\ v_{\rm p } \end{array } \right ) \end{aligned}\ ] ] and @xmath914 eq . ( [ eq : eom - eigenmodes.2 ] ) and the first row of ( [ eq : eom - eigenmodes ] ) provide an algebraic identity : @xmath915 just as in the case of self - similar solutions [ cf . ( [ eq : alg - ss ] ) ] . as in our treatment of self - similar solutions , we here treat @xmath916 as four unknown functions , and use ( [ eq : alg - pp ] ) as a check at appropriate steps of computation . we require @xmath917 to fix the coordinate freedom ( [ eq : ct.var ] ) . as in the case of self - similar solutions , we require ( i ) that the perturbations are analytic for all @xmath834 , and ( ii ) that the perturbed space - times are regular at the center ( @xmath918 at @xmath836 ) . it can be seen that apart from the overall multiplicative factor , the perturbation solutions which satisfy the analyticity condition ( i ) at the sonic point are specified by one free parameter @xmath1 , as we now explain . \(1 ) first we note that the sonic point is a _ regular singular point _ for the perturbations . that is , the system of ode s for perturbations ( [ eq : eom - eigenmodes ] ) is singular where the determinant of the matrix on the lhs of ( [ eq : eom - eigenmodes ] ) vanishes : @xmath919 ; this is identical with the sonic point condition for self - similar solutions , ( [ eq : det0 ] ) . \(2 ) next , in order to have a smooth solution at the sonic point , the third and fourth row of ( [ eq : eom - eigenmodes ] ) must be proportional to each other , like in the case of self - similar solution , ( [ eq : ss.rows ] ) . this yields an algebraic relation between @xmath920 , @xmath921 , @xmath922 and @xmath923 at the sonic point . \(3 ) third , we have an algebraic identity between @xmath924 , @xmath925 , @xmath926 and @xmath927 , given by ( [ eq : alg - pp ] ) . \(4 ) the above two algebraic relations , together with our choice of gauge @xmath928 enables us to express , e.g. @xmath922 and @xmath923 in terms of @xmath920 ; and then higher order expansion coefficients of perturbations are given in terms of @xmath920 . because the system is linear and homogeneous , overall normalization is irrelevant . we thus see that the solution which satisfies ( i ) is characterized by a single parameter @xmath1 . this , together with the regularity condition ( ii ) at the center , in general allows only discrete values for @xmath1 as now be explained . searching numerically for eigenmodes of perturbations is in some sense easier than searching for self - similar solutions , because the ode s are now _ linear and homogeneous _ in unknown functions . in particular , asymptotic behaviour of solutions as @xmath866 is easily obtained as follows . substituting the asymptotic behaviour , ( [ eq : ss - asymp1 ] ) , of self - similar solutions as @xmath841 , the system of ode s for perturbations ( [ eq : eom - eigenmodes ] ) take on a simple form , given by @xmath929 this homogeneous linear ode with constant coefficients has four independent solutions , given by @xmath930 the last one should be excluded in view of the identity ( [ eq : alg - pp ] ) , and thus asymptotic behaviour of arbitrary solutions of ode s is described by suitable linear combinations of the first three independent solutions . now , one of them ( the third one ) blows up as @xmath841 . therefore we see that any solution , which satisfies our boundary condition , should thus be obtained by choosing @xmath1 so as to eliminate the unwanted expanding mode . similarly , the asymptotic behaviour as @xmath931 is obtained . it is now seen that there are three independent solutions [ after taking into account the identity ( [ eq : alg - pp ] ) ] , whose @xmath747-dependence are @xmath932 as @xmath931 . it is thus seen that every solution stays bounded as long as @xmath933 , whereas for @xmath934 it blows up as @xmath931 . in searching for possible eigenmodes , one has to pay attention to a `` gauge mode '' which emerges from a coordinate transformation applied to the self - similar solution , as we now explain . suppose one has a self - similar solution @xmath935 . seen from another coordinate [ related by the coordinate transformation ( [ eq : ct:(s , x ) ] ) ] , its perturbation is given by setting @xmath936 in @xmath937 of ( [ eq : ct.var ] ) : @xmath938 for general @xmath939 , this does not behave like an eigenmode . however , with the choice of @xmath940 ( @xmath941 arbitrary ) , @xmath937 does behave as an eigenmode with eigenvalue @xmath941 : @xmath942 the mode emerges via the coordinate transformation of the self - similar solution @xmath943 , and thus can be considered as a result of pure gauge transformation , like a `` zero mode '' in translation invariant systems . because @xmath944 is _ arbitrary _ , this pure gauge mode forms a one - parameter family . it can also be seen that other ( non - gauge ) modes are transformed under the gauge transformation as follows : eq.([eq : ct.no - p ] ) for @xmath945 becomes @xmath946 as long as @xmath947 , by taking @xmath948 , this means transformed @xmath937 is also an eigenmode , with the same @xmath1 . thus we have a one - parameter family of eigenmodes ( with the same eigenvalue @xmath1 ) , which are mutually related by the gauge mode ( [ eq : gauge - mode ] ) . based on the above observations , we searched for desired eigenmodes as follows . we first fix a value of @xmath1 , and then starting from @xmath821 ( the sonic point ) , integrate ( [ eq : eom - eigenmodes ] ) to @xmath836 . when there appear nonzero components of the expanding modes ( for sufficiently large @xmath949 ) , we judge that this @xmath1 does not give a desired eigenmode and stopped . in this way , we searched for eigenmodes in the range of @xmath950 . we found several eigenmodes which are explained below . the profile of the eigenmode with the largest @xmath951 is shown in fig . [ fig-5.1 ] . it has the eigenvalue @xmath952 ; which corresponds to the exponent value @xmath6 , according to our scenario of section [ sec - scenario ] , and which we believe to be exact to the last digit . this is in good agreement with the value of @xcite . in our gauge , where @xmath953 , we observe the gauge mode with @xmath954 , as explained in section [ sec-pp.gauge-mode ] . at the sonic point , while maison @xcite and we in sec . [ sec - lyapunov ] use the gauge @xmath955 at the origin . the former gauge gives the gauge mode at @xmath956 , while the latter gives at @xmath957 ; the difference in the values of @xmath1 is well understood in view of ( [ eq : gauge - mode ] ) . ] to confirm our scenario , we performed a thorough search of other eigenmodes in the region @xmath958 , and found _ none _ , except for the above mentioned relevant mode and the gauge mode . we also performed a less complete search in the region @xmath959 . there is an eigenmode with @xmath960 , which is consistent with the results of the lyapunov analysis in sec . [ sec - lyapunov ] . it is not _ a priori _ obvious whether the eigenmodes form a complete set of basis functions . due to the complicated structure of the equations of motion , we have not found a beautiful argument which can restrict possible eigenvalues ( like that of @xcite and references therein ) . our search of eigenmodes has a serious drawback that it is theoretically impossible to cover the whole values of @xmath961 ( unless , of course , one employs more sophisticated mathematical techniques ) . to fill this gap , and to further confirm our scenario , we performed a lyapunov analysis , as explained in sec . [ sec - lyapunov ] . as has been stated in sec . [ sec-ss.results ] , there exists a series of self - similar solutions ( specified by the number of zeros of @xmath101 ) , in which the evans coleman solution can be considered as the first one ( exactly one zero for @xmath101 ) . we have searched relevant eigenmodes for the first several self - similar solutions and found _ more than one _ relevant modes ( see table [ tab : other fp ] ) . this implies that the other self - similar solutions are irrelevant for the generic critical behaviour . however , we here admit that our analysis is less complete for these higher self - similar solutions , because due to a difficulty in coding , ( 1 ) we have not done the analysis for ( possible ) self - similar solutions with even number of zeros of @xmath101 , ( 2 ) there can be more relevant modes than is reported in the table . ld self - similar solution & @xmath1 + 1 ( evans coleman ) & 2.8105525488 + 3 & 8.456 + & 3.464 + & 1.665 + & 0.497 + 5 & 15.80 + & 7.13 + & 3.22 + & 1.51 + & 0.500 + 7 & 15.97 + & 6.92 + & 3.20 + & 1.50 + to further confirm the uniqueness of the relevant mode we performed a lyapunov analysis around the critical solution . a lyapunov analysis and the shooting method for an ordinary differential equation adopted in @xcite and in the previous section are complementary to each other for the following reasons : ( 1 ) the former extracts eigenmodes in the descending order of its real part , whereas the latter affords information only of a finite region of complex @xmath1 plane . ( 2 ) the lyapunov analysis is useful even if the ( discrete ) eigenmodes do not form a complete set . ( 3 ) it is easier in the latter than in the former to numerically obtain accurate values of the eigenvalues @xmath1 , hence the critical exponent @xmath0 . we here present our method of analysis and its results . the lyapunov analysis is a method of extracting eigenvalues in descending order . it involves time integration the linearized eom ( [ eq : eom - var ] ) about the self - similar solution . it takes advantage of the fact that the eigenmodes with large @xmath962 dominates at late times and that the volume , defined by an ( arbitrarily chosen ) inner product on @xmath30 , of the parallelepiped spanned by the eigenmodes corresponding @xmath963 , @xmath964 , ... , @xmath965 approaches @xmath966 . during the integration , vectors in @xmath30 are orthonormalized by the gram schmidt procedure . this is essential for avoiding numerical overflows and underflows . for details see appendix [ app - lyapunov ] . we can also have information on the imaginary part of eigenvalues by the period of oscillation in time @xmath36 . we have performed the above - described calculation by numerically solving the pde s for the perturbation , ( [ eq : eom - var ] ) . we emphasize that solving the pde s for perturbations , ( [ eq : eom - var ] ) , is much easier than solving the full eom close to the critical point , because perturbations are well behaved in contrast with near - critical solutions whose derivatives are diverging in @xmath93 coordinate . we used the first order lax scheme , and the second order lax wendroff scheme for time evolution . we show the results from the lax wendroff scheme . to simplify coding , we employed a new gauge , @xmath967 . in this gauge we observe the gauge mode at @xmath957 . we have made appropriate coordinate transformations to stabilize the integration schemes . we imposed the free boundary condition at the sonic point @xmath968 . this can be done because no information can come in from outside the sonic point ( as is easily seen by studying the characteristic curves ) , for spherically symmetric metric perturbations contains no gravitational waves and are determined by the matter degrees of freedom . [ fig - kp - vs - gm ] shows the dependence of @xmath459 on @xmath4 . the largest eigenvalue is real and agrees well with the first eigenvalue ( which is real ) found by solving two - point boundary problem of an ode in sec . [ sec - pert ] . the second largest eigenvalue is that of the gauge mode , which should be exactly equal to one . for @xmath969 the third and the fourth largest eigenvalues are real , and the fifth is imaginary . for @xmath970 the third is real and the fourth and the fifth is imaginary , which are complex conjugate . those for @xmath971 are complex . we can consider that there is a cross - over of lyapunov exponents corresponding to the real eigenvalue and the complex pair between @xmath972 and @xmath973 . for @xmath974 the third and the fourth are a complex conjugate pair , so that there is another cross - over of lyapunov exponents corresponding to the real eigenvalue and the complex pair between @xmath975 and @xmath976 . for large @xmath4 it seems that many eigenmodes with close values of @xmath459 are present . our preliminary computation of ten modes for @xmath977 shows that there at least nine modes in the range @xmath978 , though it seems that not all of them are degenerate in @xmath459 . the figure establishes the uniqueness of the relevant mode in collapse of perfect fluids with @xmath979 , @xmath980 . from this we can conclude that the critical behaviour must be observed for all of these models with @xmath981 . we have performed the analysis presented in secs . [ sec - ss][sec - lyapunov ] for perfect fluid , with @xmath982 . as has been noted in @xcite , a self - similar solution of evans coleman type ( i.e. with one zero of @xmath101 ) ceases to exist for @xmath983 . for evans coleman type self - similar solutions , the lyapunov analysis of sec . [ sec - lyapunov ] establishes that there is a unique relevant mode , and thus we can observe the critical behaviour ( fig . [ fig - kp - vs - gm ] ) . then , the precise values of the relevant eigenvalue @xmath1 are obtained by the shooting method of sec . [ sec - pert ] , and the critical exponent is given by @xmath984 . the result is shown in table [ tab : kp vs gm ] . this in particular shows that the relevant eigenmodes found by maison @xcite are actually unique , and thus are responsible for the critical behaviour . the value of the critical exponent @xmath0 depends strongly on @xmath4 . moreover , the limit @xmath985 seems to be discontinuous . ( for the dust , @xmath986 , we expect a trivial critical behaviour with @xmath987 . ) this may be because the domain of attraction of the evans coleman type self - similar solution vanishes as @xmath988 . ldd @xmath4 & @xmath1 & @xmath0 + 1.00001 & 9.4629170 & 0.10567566 + 1.0001 & 9.45592488 & 0.10575380 + 1.001 & 9.38660322 & 0.10653481 + 1.01 & 8.74868715 & 0.11430286 + 1.03 & 7.61774326 & 0.13127247 + 1.04 & 7.16334221 & 0.13959964 + 1.05 & 6.76491004 & 0.14782163 + 1.06 & 6.41269915 & 0.15594058 + 1.08 & 5.81789124 & 0.17188358 + 1.1 & 5.33435815 & 0.18746398 + 1.12 & 4.93282886 & 0.20272343 + 1.15 & 4.44235059 & 0.22510605 + 1.18 & 4.0484584 & 0.2470076 + 1.2 & 3.82545008 & 0.26140715 + 1.22 & 3.62729455 & 0.27568756 + 1.25 & 3.36750228 & 0.29695600 + 1.28 & 3.14337431 & 0.31812947 + 1.3 & 3.00990875 & 0.33223599 + 1.32 & 2.88714829 & 0.34636253 + 4/3 & 2.81055255 & 0.35580192 + 1.36 & 2.66838221 & 0.37475891 + 1.38 & 2.57025726 & 0.38906611 + 1.4 & 2.47850858 & 0.40346844 + 1.42 & 2.39245265 & 0.41798110 + 1.44 & 2.31150728 & 0.43261815 + 1.46 & 2.23517329 & 0.44739260 + 1.48 & 2.16301995 & 0.46231659 + 1.5 & 2.09467339 & 0.47740140 + 1.52 & 2.02980720 & 0.49265763 + 1.55 & 1.93841621 & 0.51588508 + 1.58 & 1.85338883 & 0.53955219 + 1.6 & 1.79989076 & 0.5555893 + 1.62 & 1.74873002 & 0.5718436 + 1.64 & 1.69974510 & 0.5883235 + 1.66 & 1.65278973 & 0.6050376 + 1.68 & 1.60773076 & 0.6219947 + 1.7 & 1.56444628 & 0.6392038 + 1.72 & 1.52282404 & 0.6566747 + 1.74 & 1.48276003 & 0.6744180 + 1.76 & 1.44415717 & 0.6924454 + 1.78 & 1.40692422 & 0.7107703 + 1.8 & 1.37097467 & 0.7294081 + 1.88 & 1.2383842 & 0.8075039 + 1.888 & 1.2259859 & 0.8156700 + 1.889 & 1.2244458 & 0.8166960 + the aim of this paper was twofold : first , in sec . [ sec - rgscenario ] we presented a mathematically rigorous framework , under which most of the critical behaviour observed in gravitational collapse are expected to be analyzed . we have in particular shown rigorous sufficient conditions under which ( 1 ) critical behaviour with a continuous self - similar solution is observed and the critical exponent is given _ exactly _ in terms of the unique relevant lyapunov exponent @xmath1 by @xmath105 ( sec . [ sub - rgflow ] ) , ( 2 ) critical behaviour with a discrete self - similar solution is observed and the critical exponent is given by the same formula ( sec . [ sec-discr.ss ] ) , and ( 3 ) different models exhibit the same critical behaviour , i.e. _ universality _ in the sense of statistical mechanics ( sec . [ sec-univ.class ] ) . the sufficient conditions for ( 1 ) and ( 2 ) are essentially the _ uniqueness _ of the relevant mode , together with some regularity on the dynamics . the condition for ( 3 ) is essentially that the difference of eom is relatively small in the critical region . the key idea behind the analysis is the renormalization group . it naturally casts a system described by partial differential equations into a dynamical system on an infinite- dimensional phase space , where continuously and discretely self - similar solutions of the original equations are described by fixed points and limit cycles . this opened the way to the abstract analysis on dynamical systems described in sec . [ sec - rgscenario ] . with the help of techniques developed in statistical mechanics to study critical behaviour , we have been able to obtain detailed estimates on the time evolution of the deviation from the critical solutions ( or the stable manifold ) . we thus obtained general and precise results , including rigorous relations between the critical exponent and lyapunov exponent . another advantage of the renormalization group idea is that it enables us to treat systems with different eom by naturally introducing a space of dynamical systems , as described in sec . [ sec-univ.class ] . each system is in general driven along a renormalization group flow in this space , and this enabled us to clarify terms of eom which are `` relevant '' for the phenomena we are interested in . these point of view provide us a natural understanding on why critical behaviour are often observed in gravitational collapse . we then proceeded to confirm the picture for a concrete model of perfect fluid in secs . [ sec - eom][sec - results ] . we employed essentially two methods to show that the main assumption , uniqueness of the relevant mode , is satisfied : ( 1 ) usual shooting method , and ( 2 ) lyapunov analysis . their results agree well , and strongly suggest that the relevant mode is unique . we obtained values of the critical exponent @xmath0 in sec . [ sec - results ] . it should be emphasized that the general framework presented in this paper is useful for both mathematical and numerical analysis on concrete models of critical behaviour in gravitational collapse , because it tells that in which case one can rely on the intuitive picture of the phase space shown in the schematic diagrams figs . [ fig - flow0 ] , [ fig - poin ] , [ fig - withv ] and the argument presented in @xcite which states that the linear perturbation is enough to understand all aspects of the critical behaviour , apart from pathological behaviour mainly due to the infinite dimensionality of the phase space . for mathematical proof of the existence of the critical behaviour and estimation of the critical exponent , the essential step is to find the norm @xmath69 , which defines the functional space or the phase space @xmath30 , that satisfies the assumptions in the sec . [ sub - rgflow ] . this should be highly nontrivial and depends on each model considered . for numerical analysis the nonlinear simulation is needed only to confirm assumptions g13 in sec . [ sub - rgflow ] , especially assumption g1 . this is qualitative one rather than quantitative ( because only the finiteness of the time before blowup is needed ) , and requires the time evolution of very small region of initial data , essentially a set of initial data which is the fixed point plus the small relevant mode . the calculation of this e.g. in @xmath837 coordinate should be much easier than adaptive mesh refinement calculation in @xmath93 coordinates for wide region of the phase space and should allow a closer control of the numerical errors . this together with close analysis of the spectrum of the linear operator @xmath989 around the fixed point may lead to a computer assisted proof of theorem [ prop - final ] . there remain several open questions to be answered . first , it should be emphasized that our analysis in secs . [ sec - eom][sec - results ] confirms only the _ local _ behaviour of the flow , around the specific self - similar solutions . although self - similar solutions with more than one zeros of @xmath101 seem to be irrelevant for generic critical behaviour ( because they seem to have more than one relevant modes ) , it would be interesting to know the _ global _ behaviour of the flow , in particular , how these self - similar solutions are related . second , now that various kinds of critical behaviour for different models have been observed , it would be interesting to ask what happens in a mixed system ( e.g. perfect fluid @xmath990 scalar fields ) . we performed a preliminary analysis in this direction . that is , we considered a `` mixed '' system of radiation fluid with a real scalar field , whose energy - momentum tensor is given by @xmath991 , and studied the linear stability of the evans coleman self - similar solution in this mixed system . ( it is easily seen that eigenmodes decouple . ) lyapunov analysis in sec . [ sec - lyapunov ] , applied to this mixed system , shows that the evans coleman self - similar solution has the unique relevant mode ( given by the relevant mode of the radiation fluid ) , and all scalar eigenmodes are irrelevant . this means that the evans coleman self - similar solution is stable under the scalar perturbation , and we will observe the same radiation fluid critical behaviour for the mixed system ( at least when the scalar field is sufficiently weak ) . it would be interesting to investigate the behaviour of the system without the restriction that the scalar field is weak ; in particular , where does the cross - over between two types ( fluid , scalar ) of critical behaviour occur ? as is suggested from our general framework in sec . [ sec - rgscenario ] , a `` critical behaviour '' such as observed in gravitational collapse is not in fact limited to self - gravitating systems . similar , but slightly simpler , phenomena can be observed in much simpler systems , such as nonlinear heat equation , where we can rigorously carry out the analysis of sec . [ sec - rgscenario ] . however if a solution of eom blows up in such simple systems , it usually means that the equation is not physically applicable in the blow - up region , and the blow - up is an artifact of bad approximation . general relativity provides rare examples where the blow - up of solutions does have a physical meaning formation of a black hole . our analysis in secs . [ sec - eom][sec - results ] is not mathematically rigorous . in view of the general framework given in sec . [ sec - rgscenario ] , and in view of the above remark that the critical behaviour and blow - up of solutions are in fact physically meaningful , it is extremely desirable to develop a mathematically rigorous analysis of critical behaviour for a physically interesting model such as radiation fluid or scalar fields . we are planning to come back to this problem in the near future . our intuition on gravitational collapse still seems to be heavily based on few exact solutions , especially the limiting case of pressureless matter . the critical behaviour may provide a different limiting case that the final mass is small compared to the initial mass for more realistic and wider range of matter contents . it will be of great help to settle the problems in gravitational collapse such as cosmic censorship conjecture . we thank akio hosoya , hal tasaki and atsushi inoue for stimulating discussions . this work is partially supported by the japan society for the promotion of science ( t.k . ) and the ministry of education , science and culture ( t.h . and t.k . ) . all the computations were performed on several work stations of tokyo institute of technology . the proof consists of several steps , and is somewhat complicated . because the lemma itself is of auxiliary importance , and because the proof uses techniques which have been explained in detail in the proofs of other lemmas and propositions already proved in sec . [ sub - rgflow ] , we here sketch only the essence of each step . _ step 1 . @xmath338 is a continuous curve , connecting @xmath992 and @xmath993 , with @xmath994 , @xmath995 . _ we write @xmath996 , and @xmath997 . because @xmath992 satisfies the assumptions of lemma [ lem - ftriangle ] , * @xmath998 keeps growing until it finally exceeds @xmath241 . this also shows @xmath999 . * @xmath1000 for @xmath1001 . now @xmath335 is a straight line connecting @xmath992 and @xmath1002 ; and @xmath338 is a union of images of @xmath335 under @xmath213 for @xmath1001 : @xmath1003 because @xmath188 is continuous and because @xmath339 is finite , @xmath213 is also continuous ( for @xmath1004 ) . thus @xmath1005 is a continuous set . because endpoints of @xmath1005 are overlapping , this proves the continuity of @xmath338 . bounds on the norm follow immediately from the definition of @xmath339 . _ step 2 . @xmath338 is `` one - dimensional '' , and any @xmath1006 satisfy @xmath1007 . _ we first prove the claim for @xmath1008 by induction in @xmath20 , and then extend the result for general @xmath1006 . first , the claim is satisfied for @xmath309 , because @xmath1009 is a straight line connecting @xmath992 and @xmath1002 . now suppose the claim holds for @xmath1008 and try to prove it for @xmath1010 . [ any @xmath1011 should have at least one inverse image in @xmath1005 , so this is sufficient . ] we compute as we did in deriving ( [ eq : ap.ev1])([eq : ap.ev11 ] ) using ( [ eq : rtrel2 ] ) , ( write @xmath1012 ) @xmath1013 now , by the inductive assumption , @xmath1014 . so we have @xmath1015 \ , \\| f \\|_{\rm u } \\ & & \\| { \cal r}_{\sigma } ( u ' ) - { \cal r}_{\sigma } ( u ) \\|_{\rm u } \nonumber \\ & & \qquad \geq \bigl [ e^{\kappa \sigma } - 6 k_{1 } \\| u - u^ { * } \\| - k_{1 } \\| f \\|_{\rm u } \bigr ] \ , \\| f \\|_{\rm u}\end{aligned}\ ] ] taking @xmath1016 and @xmath1017 into account , this leads to @xmath1018 it is now easily seen that this recursion preserves @xmath1019 and @xmath1020 follows from our choice of @xmath1021 , i.e. ( [ eq : ep2del2def1 ] ) and ( [ eq : ep2del2def2 ] ) . all this proves the claim for @xmath1022 for some @xmath20 . because @xmath1023 is strictly increasing , we can combine this bound for each @xmath1024 , to conclude @xmath1007 for @xmath1006 which are not necessarily in some @xmath1024 . _ step 3 . contraction in @xmath348-direction . _ we now consider the distance between a flow @xmath1025 and @xmath338 , and show that their distance in @xmath1026-direction is decreasing . to be more precise , we introduce a decomposition ( see fig . [ fig - dcmp ] ( c ) ) : @xmath1027 where this time @xmath1028 and @xmath172 . because we have already shown that @xmath1029 is one - dimensional and extends from @xmath21 to @xmath322 in @xmath1030-direction , we can always decompose this way , as long as @xmath1031 . to derive a recursion for @xmath1032 , we start as usual @xmath1033 where @xmath1034 we have to define @xmath1035 so that @xmath469 is decomposed as @xmath1036 so that the entire expression can be written as @xmath1037 with @xmath1038 now ( [ eq : escon.deldcp1 ] ) implies @xmath1039 and @xmath1040 however , we have shown in step 2 that @xmath338 is in some sense `` parallel '' to @xmath156 , i.e. @xmath1041 combining ( [ eq : escon.deldcp3])([eq : escon.deldcp5 ] ) implies @xmath1042 and we finally have @xmath1043 . \nonumber \\ & & \label{eq : lem6.grec2}\end{aligned}\ ] ] by taking @xmath1044 and @xmath1045 sufficiently small as in ( [ eq : ep2del2def1 ] ) and ( [ eq : ep2del2def2 ] ) , we can bound the right hand side by @xmath1046 with some @xmath1047 . this in particular shows that intersections of @xmath338 and the plane @xmath343 for various @xmath21 s are cauchy sequences , and thus the limit @xmath1048 exists : existence of @xmath334 . finally , repeating the above argument replacing @xmath338 by @xmath334 , we get the contraction around @xmath334 as well . @xmath115 we here provide a lemma , which effectively solves recursive inequalities encountered in sec . [ sec - rgscenario ] , such as ( [ eq : ap.ev33 ] ) , ( [ eq : ap.ev34 ] ) , ( [ eq : rec - ineq - v1 ] ) , ( [ eq : rec - ineq - v2 ] ) . [ lem-5 ] let @xmath1049 , @xmath1050 , @xmath432 , @xmath502 and @xmath140 be positive constants satisfying @xmath1051 , @xmath1052 , and @xmath1053 let @xmath1054 be nonnegative sequences satisfying @xmath1055 \label{eq : ap.ev42 } \\ & & f_{n+1 } \leq \bar{\lambda } f_{n } + c \left [ g_{n } ( f_{n } \vee b_{n } ) + ( f_{n } \vee b_{n})^{2 } \right ] , \label{eq : ap.ev41 } \end{aligned}\ ] ] where @xmath402 ( @xmath214 ) satisfies @xmath1056 define @xmath1057 . + ( i ) @xmath381 is finite and the following hold for @xmath382 : @xmath1058 ( ii ) suppose in addition that @xmath1059 satisfies @xmath1060 then the following hold for @xmath382 : @xmath1061 where @xmath433 is some positive constant . the proof is an elementary exercise in calculus , which is done in several steps in a bootstrapping manner . we only consider @xmath382 , which in particular means @xmath1062 . \(i ) _ step 1 . rough estimates and @xmath1063 . _ we first prove by induction @xmath1063 ( i.e. ( [ eq : f - b - bound ] ) ) , and @xmath1064 these are satisfied for @xmath309 , because @xmath1065 , and because of our choice of @xmath140 . now suppose they hold for @xmath20 . our inductive assumption in particular includes @xmath1063 , which reduces the recursion ( [ eq : ap.ev42 ] ) to @xmath1066 \end{aligned}\ ] ] where we used uniform bounds on @xmath402 and @xmath391 . eq.([eq : roughbd3 ] ) immediately follows from our choice of @xmath502 and @xmath140 . on the other hand , @xmath1063 reduces the recursion ( [ eq : ap.ev41 ] ) to @xmath1067 \nonumber \\ & \leq & \left [ \bar{\lambda } + c ( \delta + \epsilon ) \right ] \frac{2}{\lambda+1 } b_{n+1 } \end{aligned}\ ] ] where in the last step we used ( [ eq : roughbd3 ] ) just proved . the factor before @xmath1068 is easily seen to be less than one , and ( [ eq : f - b - bound ] ) is proved . rough estimate on @xmath391 . _ first note that ( [ eq : roughbd3 ] ) for @xmath1069 and the definition of @xmath381 implies @xmath1070 i.e. ( [ eq : detlem.bn2rgh ] ) . also solving ( [ eq : roughbd3 ] ) backwards from @xmath381 to @xmath20 , we have @xmath1071 which , in view of @xmath1072 , is nothing but ( [ eq : detlem.best ] ) . refined estimate on @xmath391 . _ for later use , we note the following refined estimate , which is valid without the additional assumption ( [ eq : del2cond2 ] ) . first note that @xmath1073 , @xmath1074 , and @xmath1063 guarantee @xmath1075 . so the recursion ( [ eq : bn - rec - red ] ) implies ( @xmath382 ) @xmath1076 \nonumber \\ & \ ; { { \ , \raise-.24em\hbox{$>$ } \hspace{-0.8em } \raise.31em\hbox{$<$}\ , } } \ ; & \lambda^{n } \prod_{k=0}^{n-1 } \left [ 1 \pm \frac{c ( g_{k}+ b_{k})}{\lambda } \right ] \nonumber \\ & \ ; { { \ , \raise-.24em\hbox{$>$ } \hspace{-0.8em } \raise.31em\hbox{$<$}\ , } } \ ; & \lambda^{n } \ , \exp \left [ \pm \frac{(2 \ln 2 ) c}{\lambda } \sum_{k=0}^{n-1 } ( g_{k}+ b_{k } ) \right ] . \label{eq : refinedb.2}\end{aligned}\ ] ] in the second step above we used @xmath1077 , which is valid for @xmath1078 . [ note that @xmath1079 . ] \(ii ) given the above estimate ( [ eq : refinedb.2 ] ) , the rest is easy . making use of the assumption ( [ eq : del2cond2 ] ) on @xmath402 , and the estimate ( [ eq : detlem.best ] ) on @xmath1080 proven so far , it can be easily seen that the sum in the exponent of ( [ eq : refinedb.2 ] ) is bounded by a constant depending only on @xmath1081 and @xmath1082 , uniformly in @xmath20 and @xmath381 as long as @xmath382 . taking ( [ eq : detlem.bn2rgh ] ) into account , ( [ eq : refinedb.2 ] ) with @xmath1069 gives ( [ eq : detailedlemma ] ) . @xmath115 defriese and carter ( see @xcite ) showed that the line element of any self - similar , spherically symmetric space - time is written as @xmath1083 let @xmath1084 and @xmath1085 , where @xmath1086 one has @xmath1087 where @xmath448 and @xmath244 are functions of @xmath1088 . since @xmath1089 is a function of @xmath1088 only , @xmath448 and @xmath244 are functions of @xmath1089 . the lyapunov method has been extensively used in nonlinear analysis @xcite . to make our presentation self - contained and transparent , we present a concise mathematical description of the method , restricted to the situation we are interested in . let us consider the linearization of a flow on a real hilbert space ( or a real hilbert manifold ) around an orbit . it is determined by an equation @xmath1090 where each @xmath1091 is an element of real hilbert space @xmath101 , a complete vector space with inner product , and @xmath1092 is a real function determined by the ( background ) orbit which we are considering . let @xmath1093 denote the inner product and @xmath1094 denote the norm defined by the inner product . we have @xmath1095 in ( [ eq : lyp - pde ] ) when the orbit is a fixed point , and @xmath1096 when it is an periodic orbit with periodicity @xmath469 . we concentrate on these cases below . let us define time evolution operator @xmath1097 by @xmath1098 where @xmath539 is a solution of ( [ eq : lyp - pde ] ) . we wish to find eigenvalues @xmath1 and eigenvectors @xmath1099 of @xmath1100 satisfying @xmath1101 in particular those with large @xmath962 . here @xmath1 is a complex number and @xmath1099 is a complex vector with @xmath1102 . in ( [ eq : lyp - eigen ] ) and below , @xmath36 ( and @xmath151 ) can be any number in the case of a fixed point , whereas it is restricted to be integer multiple of @xmath469 in the case of a periodic orbit . we assume that the eigenvalue problem ( [ eq : lyp - eigen ] ) has a complete set of discrete eigenvectors , denoted by @xmath1103 , with the corresponding eigenvalues @xmath1104 , @xmath1105 . each @xmath1106 is called the @xmath1107th lyapunov exponent . ( here we implicitly assume that we can always find @xmath1108 with the largest real part after we have defined @xmath963 , ... , @xmath1109 . ) we observe that if @xmath1108 is not real there is an integer @xmath1110 with @xmath1111 and @xmath1112 , where asterisks ( except for those on @xmath1107 or @xmath1113 ) denote complex conjugate in this appendix . we define @xmath1114 by @xmath1115 if @xmath1108 is real , and @xmath1116 , @xmath1117 if @xmath1108 is not real , @xmath1118 and @xmath1119 . any real vector @xmath78 can be expanded by @xmath1120 or @xmath1121 as @xmath1122 where @xmath1123 . without loss of generality , we assume all eigenvectors are normalized so that @xmath1124 , although @xmath1125 s are not necessarily orthogonal with respect to the inner product @xmath1126 . let @xmath1127 denote an @xmath20-frame of ( linearly independent , real ) vectors and let @xmath1128 be the subspace of @xmath101 spanned by @xmath1129 . below we often omit @xmath1130 in @xmath1131 if no confusion occurs . we define several operations on frames . first given an operator @xmath1132 we define @xmath1133 , an operator on frames induced from @xmath1134 , by @xmath1135 . we in the following consider the cases where @xmath1134 is a linear operator ( such as @xmath1136 ) , or the normalization operator @xmath158 : @xmath1137 . second , we define the orthogonalization operator @xmath1138 , where @xmath1139 is defined by @xmath1140 we adopt the convention that @xmath1141 denote the @xmath1107th vector of @xmath20-frame @xmath1142 , when @xmath1133 is one of the operations on a frame defined above . distinct from @xmath1133 induced from @xmath1143 , @xmath1144 in general depends on @xmath1145 with @xmath1146 . note that @xmath1147 . * the first expression ( [ eq : rekappa0 ] ) of the proposition is mathematically straightforward , but is not suitable for numerical calculations , due to serious overflow and underflow problems in large time integration . to overcome this difficulty , we employ the second expression of the proposition , which is performed in the following manner in practice . 1 . prepare @xmath1129 and let @xmath1151 . 2 . evolve @xmath1129 in time by @xmath1136 , and define @xmath1152 . 3 . find @xmath1153 and @xmath1154 by the gram schmidt procedure : @xmath1155 4 . add @xmath1156 to @xmath1157 . define new @xmath1158 and go back to ( 1 ) . 1 . if @xmath1133 is induced from a _ linear _ operator @xmath1162 , then as operators on frames , @xmath1163 . in particular , @xmath1164 2 . if @xmath1161 are induced from operators @xmath1165 on @xmath101 , we have @xmath1166 . in particular , @xmath1167 3 . for @xmath1168 ( @xmath1149 ) we have @xmath1169 and @xmath1170 _ proof . _ ( i ) by the definition of @xmath1171 , any frame @xmath1129 is written as @xmath1172 with suitable scalars @xmath1173 s . since @xmath1134 is linear we have @xmath1174 . however , for scalars @xmath1173 s , @xmath1175 which can be shown by induction in @xmath1107 . so we have @xmath1176 . we assume @xmath1129 to be _ generic _ in the sense that @xmath1179 the set of all nongeneric @xmath1129 s is of measure zero in the space of @xmath20-frames , because each nongeneric @xmath1129 contains at least one such @xmath1107 that @xmath1180 lies in a subspace of @xmath101 of codimension one , @xmath1181 . from ( [ eq : generic - frame ] ) we can define @xmath1182 such that ( for @xmath1149 ) @xmath1183 note that we have @xmath1184 let us expand @xmath1185 in terms of @xmath1120 as @xmath1186 where @xmath1187 s are defined similarly as @xmath1188 s in ( [ eq : expand ] ) . condition ( [ eq : generic - frame ] ) implies that @xmath1189 for all @xmath1190 , and that no @xmath1191 s appear for @xmath1192 . time evolution of @xmath1193 is given by @xmath1194 where @xmath1195 and @xmath502 is some positive constant . since @xmath1189 , @xmath1196 is bounded and nonzero uniformly in @xmath36 . operating @xmath1171 on both sides of ( [ eq : eetilde ] ) , we thus have by ( [ eq : relation : oo ] ) , @xmath1197 for sufficiently large @xmath36 . the left hand side can be replaced by @xmath1198 leading to @xmath1199 because we have from ( [ eq : relation : o ] ) and ( [ eq : off ] ) that @xmath1200 for any @xmath36 . thus we have @xmath1201 for sufficiently large @xmath36 , where @xmath1202 , @xmath1203 are positive constants . ( [ eq : lambda : simple ] ) and ( [ eq:\|otff\| ] ) prove ( [ eq : rekappa0 ] ) . now we turn to the proof of equivalence of ( [ eq : rekappa0 ] ) and ( [ eq : rekappa ] ) . using ( [ eq : relation : n ] ) and ( [ eq : relation : n0 ] ) we have @xmath1204 and thus @xmath1205 so by the definition of @xmath1206 , we have @xmath1207 where in the last step we used ( [ eq : relation : tn ] ) and ( [ eq : relation : oo ] ) . we thus have @xmath1208 and the proposition is proved .	we present a general framework for understanding and analyzing critical behaviour in gravitational collapse . we adopt the method of renormalization group , which has the following advantages . ( 1 ) it provides a natural explanation for various types of universality and scaling observed in numerical studies . in particular , universality in initial data space and universality for different models are understood in a unified way . ( 2 ) it enables us to perform a detailed analysis of time evolution beyond linear perturbation , by providing rigorous controls on nonlinear terms . under physically reasonable assumptions we prove : ( 1 ) _ uniqueness _ of the relevant mode around a fixed point implies universality in initial data space . ( 2 ) the critical exponent @xmath0 and the unique positive eigenvalue @xmath1 of the relevant mode is _ exactly _ related by @xmath2 , where @xmath3 is a scaling exponent used in calculating the eigenvalue . ( 3 ) the above ( 1 ) and ( 2 ) hold also for discretely self - similar case ( replacing `` fixed point '' with `` limit cycle '' ) . ( 4 ) universality for different models holds under a certain condition . [ these are summarized as theorems [ prop - final ] , [ prop - final - discr ] and [ theorem-11 ] of sec . [ sec - rgscenario ] . ] according to the framework , we carry out a rather complete ( though not mathematically rigorous ) analysis for perfect fluids with pressure proportional to density , in a wide range of the adiabatic index @xmath4 . the uniqueness of the relevant mode around a fixed point is established by lyapunov analyses . this shows that the critical phenomena occurs not only for the radiation fluid but also for perfect fluids with @xmath5 . the accurate values of critical exponents are calculated for the models . in particular , the exponent for the radiation fluid @xmath6 is also in agreement with that obtained in numerical simulation . boxedeps # 1#1
a plethora of astrophysical and cosmological observations have firmly established that non - baryonic dark matter ( dm ) must exist in our universe , and contribute to the overall cosmological energy density about five times more than ordinary matter . however , none of the particles of the standard model ( sm ) can account for the dm , which therefore constitutes a clear hint of new physics . colorless , electrically neutral and weakly interacting massive particles with mass in the gev - tev range are ubiquitous in new physics models , and appear to be well suited to reproduce quantitatively the measured dm energy density if their stability on cosmological time scales can be ensured . + from the model - building point of view , dm stability is most commonly enforced by assuming some suitable symmetry that forbids its decay into lighter sm particles . for example , in supersymmetric models this role is played by r - parity that stabilizes the lightest supersymmetric state , in universal extra dimensional models conservation of kaluza - klein parity ensures that the lightest kaluza - klein state remains stable @xcite , @xmath8-parity stabilizes the lightest @xmath8-odd particle in the littlest higgs model @xcite , suitable @xmath9 parities play the same role e.g. in the scotogenic model @xcite , in the inert doublet model @xcite , and in several other cases . often these stabilizing symmetries are just imposed by hand on the low energy lagrangian , and it is certainly more satisfactory when their origin can be traced back to some high energy completion of the model in question . a plausible way to generate unbroken discrete @xmath10 symmetries relies on assuming extra gauged @xmath11 abelian factors which are only broken by order parameters carrying @xmath12 units of the @xmath11 charge @xcite ( see also @xcite ) . such a mechanism renders grand unified theories ( guts ) based on gauge groups of rank larger than four particularly interesting , since they contain extra cartan generators besides the @xmath13 of the @xmath14 sm gauge group that , when broken by vacuum expectation values ( vevs ) of scalars in appropriate representations , yield discrete @xmath10 symmetries which inherit all the good properties of the parent local gauge symmetry . in particular , this type of symmetries remain protected from gravity induced symmetry breaking effects @xcite which , although suppressed by the planck scale , could jeopardize dm stability @xcite . one of the most interesting gut groups that allows to preserve at low energies an unbroken discrete gauge parity is the rank five group @xmath0 . as is well known , @xmath0 has many theoretically appealing properties : it unifies all sm fermions in a single @xmath15 dimensional irreducible representation including one right handed ( rh ) neutrino , it can explain the suppression of neutrino masses via the seesaw mechanism @xcite , it allows for gauge coupling unification at a sufficiently high scale to account for proton stability , and is automatically free from gauge anomalies . in this paper we focus on a @xmath0 gut model in which the breaking to the sm gauge group is driven by vevs of scalars in the @xmath16 representation . it has been recently shown that this model is compatible with unification @xcite , and that it can fit all charged fermion masses and mixings as well as the low energy neutrino data @xcite , while simultaneously explaining the cosmological baryon asymmetry via leptogenesis @xcite . it is therefore interesting to see if this framework can also accommodate automatically stable dm candidates in the fundamental @xmath1 dimensional representation of the group . @xmath0 is a rank five group and thus with respect to the sm model it contains one additional cartan generator that , upon breaking of the unified group , can give rise to a new @xmath11 gauge group factor . the @xmath11 charges of the component fields are conventionally normalized by setting the smallest charge equal to one . then , if @xmath11 is further broken by vevs of scalars carrying @xmath17 units of charge with @xmath18 as their greatest common divisor , a discrete center @xmath19 remains unbroken @xcite . with vevs in tensor representations can result in an unbroken @xmath9 parity was already pointed out in @xcite , in relation to the possible appearance of extended topological structures of cosmological relevance . we thank q. shafi for bringing this reference to our attention . ] in our setup @xmath0 is broken to @xmath20 by vevs of scalars in @xmath21 , which are all @xmath0 tensor representations . with respect to the non - sm @xmath11 factor singled out in the maximal subgroup @xmath22 , which is the one whose breaking gives rise to the gauge discrete symmetry , all @xmath0 tensor representations branch to @xmath22 fragments which have even values of the @xmath11 charge . the lowest charge value for the fragments acquiring vevs is 2 [ e.g. : @xmath23 and therefore a @xmath24 parity survives , which can guarantee the stability of the lightest particles belonging to appropriately chosen representations . restricting to dimensions @xmath25 we have that , depending if they are fermions or bosons , stable particles appear in the following representations @xcite : @xmath26 for example , fermions in the vector @xmath27 can not decay into sm fermions in the @xmath15 since this is a spinorial representation for which all fragments under @xmath22 carry odd @xmath11 charges and upon @xmath11 breaking then acquire odd @xmath24 parity . various proposals for @xmath0 dm candidates that are stabilized by the @xmath9 parity of gauge origin have been put forth in the recent literature : a dedicated analysis of scalar dm in the @xmath28 was carried out in @xcite , while the possibility of fermionic dm in the @xmath29 was addressed in @xcite ( see also @xcite where the @xmath29 is allowed to mix with a @xmath1 ) . other more general studies regarding possible embeddings of dm in @xmath0 can be found in @xcite . indeed , so far a special attention has been devoted to dm in the scalar @xmath28 and in the fermionic @xmath29 , and a possible reason for this might be the fact that both these representations contain sm singlets . needless to say , identifying dm candidates with sm singlets can naturally explain why all experimental direct detection ( dd ) searches have been eluded so far . in contrast , the @xmath1 dimensional vector representation of @xmath0 has not attracted much attention , although this could well be considered as the minimal choice . perhaps this is due to the fact that the @xmath1 does not contain sm singlets , and in particular all its states carry hypercharge , and thus couple to the @xmath30 boson , which might led to the conclusion that this possibility is excluded by dd limits . + in this letter we argue that fermionic dm in the @xmath1 of @xmath0 is instead a viable possibility . our main observation is that in a scenario in which fermions in a vectorlike @xmath31 acquire tree level masses via a yukawa coupling with the ( antisymmetric ) @xmath32 , loop diagrams involving an insertion of @xmath33-@xmath34 mixing produce a mass splitting between the two lightest mass eigenstates , which ( in our minimal realization ) are two neutral dirac fermions . we show that the neutral @xmath6 gauge bosons couple non - diagonally the light eigenstate to the heavier one and , as a result , at the leading order only inelastic neutral current scatterings of dm off target nuclei is allowed . if the mass splitting between the light and heavy mass eigenstates is larger than the typical dm kinetic energy @xmath35 , then the scattering is kinematically forbidden and dd bounds are automatically evaded . since the loop - induced splitting is suppressed by the rh gauge boson mass , the previous requirement can be translated into an upper - bound on @xmath36 which , combined with the lower bounds from flavour and cp violating processes in the @xmath37 and @xmath38 meson systems @xcite and from direct searches at the lhc @xcite , results in @xmath39 . the fact that the null result of dd dm searches constrains @xmath40 to lie at a relatively low scale , can reinforce the hope that a rich phenomenology could be within the reach of the lhc . we assume that @xmath0 is broken at the unification scale to the intermediate group @xmath41 by the vev of a @xmath32 . @xmath42 is then broken at an intermediate scale @xmath43 to the sm gauge group @xmath44 by the vev of @xmath45 , and finally @xmath46 is broken by electroweak doublet vevs in a ( complexified ) @xmath47 down to @xmath48 : so(10 ) & & 3_c 2_l 2_r 1_b - l + & & 3_c 2_l 1_y 2 + & & 3_c 1_q 2 . in eq . ( [ bc ] ) we have introduced for the gauge groups the short - hand notation e.g. @xmath49 , and together with the unbroken continuous gauge symmetries we have also written down the discrete @xmath9 factor which survives down to the last breaking step . although the gut symmetry breaking triggered by an adjoint @xmath32 together with a @xmath45 develops instabilities at the tree level , it has been recently shown that the inclusion of quantum corrections can solve this problem and make the model viable @xcite . the first symmetry breaking is achieved via the @xmath50 singlet contained in the @xmath51 . the second step is driven by a @xmath52 vev in the sm singlet direction which also provides majorana masses for the rh neutrinos , and the last step is driven by the vevs of electroweak doublets contained in the @xmath47 . note that while a real @xmath47 is sufficient to drive the @xmath53 symmetry breaking , a @xmath47 containing complex fields is needed to reproduce realistic fermion masses @xcite and in particular to accommodate the @xmath54 mass ratio . moreover , to reproduce the complete charged fermion mass spectrum accounting also for yukawa non - unification of the lepton and down - type quarks of the first two generations , a contribution from the vevs of the electroweak doublets appearing in the @xmath45 is also necessary @xcite . these vevs are unavoidably induced when @xmath55 is broken by the @xmath56 @xcite . all in all , the masses of the sm fermions are generated from the following yukawa terms : - l_= @xmath57_i ( h_ij @xmath58_h + g_ij @xmath58^_h + f_ij @xmath59_h ) @xmath57_j , where @xmath60 , @xmath61 and @xmath62 are @xmath63 symmetric matrices in flavour space and @xmath64 are family indices . the fermion couplings to @xmath65 can be forbidden by assigning to the fields a global @xmath11 peccei - quinn charge @xcite . in practice this sets @xmath66 , simplifying the yukawa structure of the model , and providing a dm candidate for non - supersymmetric @xmath0 models in the form of axions . in the spirit of avoiding the introduction of additional symmetries , and given that we are interested in a weakly interacting dm candidate , we will not follow this route , and we allow for @xmath67 . possible fcnc arising from coupling quarks of the same type to two different higgs doublets , as it would happen in this situation , can be kept under control in various ways e.g. by assuming a hierarchy @xmath68 . + let us now add to the @xmath0 model outlined above a pair of fundamentals @xmath31 containing new fermions . and @xmath69 chiralities with calligraphic subscripts ( @xmath70 ) , while normal subscripts ( @xmath71 ) label the @xmath72 gauge group factors . ] the tensor product of two vectors of @xmath0 is : @xmath58 @xmath58= ^s @xmath73^a ^s , where the superscripts denote @xmath74ymmetric and @xmath75ntisymmetric representations . although our model does not contain a @xmath76 of fundamental scalars , loop corrections can generate mass contributions that mimic the coupling to ( effective ) representations , as long as these couplings are allowed by the symmetries of the model . to keep as general as possible it is then convenient to write down all the allowed gauge invariant yukawa couplings , which are : - l _ & = & _ a= , @xmath58_a ( m_a+ _ a _ h ) @xmath58_a + & + & , where , in order not to over - clutter the expressions , we have left understood the usual spinor notations . it is instructive to analyze these couplings in terms of representations of the @xmath77 . the branching rule for the @xmath0 vector is @xmath78 so that in the first line the invariant mass term @xmath79 multiplies the @xmath80 singlet from the product of the same @xmath81 . the second term involves the symmetric @xmath82 and , besides containing a @xmath83 coupling involving the @xmath84 adjoint , it also includes couplings of the symmetric @xmath85 to a pair of fundamentals : @xmath86 . let us note at this point that if the colorless @xmath72 triplet contained in the @xmath87 ( @xmath85 ) acquires a small vev , these terms would generate a majorana mass for the neutral components of the fermion doublets in the @xmath88 ( @xmath89 ) . however , the same is not true for the analogous term in the second line since it contains only terms that couple two different @xmath1 s . finally , since the @xmath90 is antisymmetric , it must couple different representations , and thus it appears only in the second line . the model we will now study is specified by the following ingredients : ( i ) the @xmath91 is absent ; ( ii ) the adjoint vev @xmath51 which can be written as : = ( a , a , a , b , b ) ( 0&1 1&0 ) , acquires a dimopoulos - wilczek structure @xcite with @xmath92 and @xmath93 ( since @xmath94 breaks @xmath95 we require @xmath96 in order to respect the symmetry breaking pattern eq . ( [ bc ] ) ) ; ( iii ) we set @xmath97 . this can be viewed as technically natural since in this limit a global @xmath11 symmetry @xmath98 arises ( we will briefly comment below on the consequences of relaxing this assumption ) ; ( iv ) finally , we will also work with @xmath99 . this is just a simplification : a term proportional to @xmath100 preserves the global @xmath11 symmetry obtained for @xmath101 which eventually ensures the dirac nature of the dm states ( see next section ) , and as long as @xmath102 a non - vanishing @xmath100 would not change the analysis . in the following we label @xmath103 representations as @xmath104 where @xmath105 denote the dimensions of the multiplets under the respective symmetry factor , and @xmath106 gives the value of the @xmath107 charge . each one of the two @xmath1 s contains one @xmath108 bi - doublet . we denote the bi - doublet contained in @xmath109 as @xmath110 , with components : _ , = _ , ^+- & _ , ^++ + _ , ^ & _ , ^-+ , where the superscripts carried by the component fields denote the @xmath111 isospin eigenvalues of @xmath112 in units of @xmath113 . given that the @xmath1 has vanishing @xmath107quantum number , the electromagnetic charge for this representation is simply @xmath114 . thus , @xmath115 and @xmath116 are neutral fields , while @xmath117 have electric charge @xmath118 . the mass term for the neutral states arising from @xmath51 in eq . ( [ fvvev ] ) reads : -_m = m_bwith @xmath119 . a similar mass term can be written also for the charged components @xmath120 which at lowest order are degenerate in mass with the neutral states . however , electromagnetic corrections from loops involving sm gauge bosons lift this degeneracy inducing a charged - neutral mass difference @xmath121mev @xcite , thus ensuring that the lightest states are the neutral ones . has vanishing @xmath106 charge , and @xmath122 for both charged and neutral states , so that the corresponding loop contributions cancel in the mass difference . therefore the sm result @xcite for the charged - neutral mass difference for fermion doublets holds also in this case . ] ( [ eq : mb ] ) describes a pair of neutral dirac fermions @xmath123 and @xmath124 with degenerate masses equal to @xmath125 . however , these states get mixed via loop diagrams ( depicted in figure [ fig:2 ] ) involving two external @xmath126 vevs , which generate a mass term : -_= _ m. the tree level eq . ( [ eq : mb ] ) and the loop induced mass term eq . ( [ eq : dm ] ) are sketched in figure [ fig:1 ] . upon diagonalization of the mass matrix ( see the appendix ) we end up with one heavier ( @xmath127 ) and and one lighter ( @xmath128 ) dirac fermions , respectively with masses m_h , l = m_b _ m . the important point is that the couplings of @xmath129 to the neutral gauge bosons @xmath6 are off - diagonal : both @xmath6 couple ( with an opposite overall sign ) to the vectorlike neutral current j_^nc= _ _ l + h.c .. two ingredients are crucial for this result : ( i ) a complex @xmath47 : this is in any case needed to reproduce the sm fermion mass pattern . the tensor product of two @xmath1 s contains in its symmetric part the @xmath76 , which can be regarded here as an effective @xmath91 coupled to the fermions via the loop diagram . moreover , since @xmath47 develops vevs in the @xmath130-@xmath69 doublet components , the effective @xmath131 contains a non - vanishing @xmath132 breaking vev in @xmath133 , which provides an independent mass term to the fermion bi - doublets . ( ii ) a tree level mass coupling between @xmath134 and @xmath135 : a non - vanishing loop mass contribution can only appear if a chirality flip can be inserted inside the loop diagram . in our simplified setup with @xmath136 this can only be provided by @xmath125 ( a contribution from @xmath137 which also couples @xmath134 to @xmath135 would act in the same way with the replacement @xmath138 ) . note that the global @xmath11 symmetry mentioned at the end of section [ framework ] remains unbroken also at the loop level . would instead break this @xmath11 . as a result , we can expect that the two dirac states will split into four majorana fermions with a spectrum determined by the relative sizes of @xmath139 and @xmath140 . ] ) -([eq : dm ] ) . the numbers given in parenthesis refer to the @xmath141 values carried by the mass insertions . ] the mass splitting @xmath142 is an important quantity , since for @xmath143tev , the @xmath6 mediated inelastic dm scattering off target nuclei @xmath144 is kinematically forbidden only if @xmath145 @xcite , and only in this case @xmath128 could have escaped dd dm searches . let us then proceed to estimate its value . we denote the bi - doublet scalar contained in @xmath47 as @xmath146 , with components : = ^+- & ^++ + ^ & ^-+ , where the superscripts have the same meaning as for the fermions eq . ( [ bidoublet ] ) . assuming for simplicity that there is only one scalar bidoublet and that its vevs are real , we can write = v_u & + & v_d , where @xmath147 is the electroweak breaking vev . one diagram contributing to @xmath148 is depicted in figure [ fig:2 ] . the crossed diagrams should also be added , and a similar pair of diagrams can be drawn for external fermions with exchanged lr isospin labels @xmath149 . taking @xmath40 as the largest mass scale in the loop , the diagram can be estimated as @xmath150 where we have used @xmath151 with @xmath152 the @xmath42 breaking vev of @xmath45 , @xmath153 with @xmath154 the weinberg angle , and @xmath155 is to a very good approximation the dm mass . in the last expression we have introduced the @xmath5-@xmath156 mixing parameter @xmath157 which is experimentally bounded in various ways . electroweak precision data set the upper limit @xmath158 @xcite . on the other hand , @xmath159 is bounded from above also by the ratio of gauge boson masses squared : _ lr= , where in the first equality we have approximated @xmath160 . flavour and cp violating processes in the @xmath37 and @xmath38 meson systems provide an absolute lower bound on the @xmath161 gauge bosons mass @xmath162tev @xcite . similar bounds on @xmath40 have been also obtained from direct searches at the lhc @xcite . using these figures we obtain the conservative upper bound @xmath163 . the mass splitting @xmath164 between the light and heavy neutral states can then be bounded from above as : 2_m 15 ( ) ^2 ( ) . the constraint from dm non - observations in dd experiments @xcite @xmath145 can then be translated in the following _ upper limit _ on the @xmath161 gauge boson masses : m_w_r 25 ( ) ^1/2 . indeed this result suggests the possibility of a non trivial interplay between dm searches in dd experiments , and searches for new physics at the lhc or at a future @xmath165tev hadron collider . , see e.g. , @xcite . ] before concluding this section , let us note that non - vanishing dd cross sections will appear at the loop level , with leading contributions involving the exchange of pairs of @xmath4 gauge bosons @xmath166 or @xmath167 . the quantitative effects of the corresponding diagrams have been studied for example in refs . @xcite , and it was found that the resulting cross sections do not exceed @xmath168@xmath169 , which is far below the current experimental bounds @xcite . in the relevant mass range ( @xmath170tev ) this remains also below the reach of next generation dd experiments @xcite and close to the neutrino scattering background . for @xmath171 . inclusion of the @xmath172 crossed diagram is left understood . an analogous diagram contributes the same @xmath148 also for @xmath173 . ] the relatively low value of the intermediate symmetry breaking scale implied by eq . ( [ mwr ] ) could be of some concern for what regards the light neutrino masses . in general , the fact that the seesaw mechanism @xcite can be automatically embedded within @xmath0 provides an elegant way to explain why the neutrino masses are so suppressed . @xmath0 unification implies relations between the light neutrino , rh neutrinos , and up - type quark masses , which generically require rather heavy rh neutrinos ( @xmath174 ) . the natural range for @xmath175 is then loosely determined by the up and top quark masses as @xmath176 . on the other hand @xmath177 s acquire their masses from the same @xmath45 vev that breaks @xmath42 and concurs to determine the value of @xmath40 , and thus we would expect their masses to be of the order of @xmath40 , which remains bounded from above by eq . ( [ mwr ] ) . such a relatively low mass scale for the rh neutrinos does not provide enough suppression . one might then be tempted to appeal to a different suppression mechanism . for example the inverse seesaw @xcite is an elegant option that allows to suppress neutrino masses even when @xmath178 . however , implementing the inverse seesaw requires the addition of a @xmath0 singlet with a ( small ) majorana mass term , which couples to @xmath177 and to some scalar representation with non - vanishing vev . given that @xmath179 the only option for writing down a renormalizable coupling is a scalar multiplet @xmath180 . however , @xmath181 would break the @xmath9 parity thus allowing dm decays . a similar conclusion can be reached also in the case the new fermion is not a @xmath0 singlet but is assigned to some suitable @xmath0 representation . we must then conclude that in our framework the inverse seesaw does not provide a viable alternative to explain the neutrino mass suppression . a straightforward , although not so elegant , way out , is to appeal to cancellations in the neutrino dirac mass matrix . this relies on the fact that , when projected onto sm multiplets , the fermion mass matrices originating from eq . ( [ stn ] ) acquire non trivial clebsch - gordan coefficients that weight the various vev contributions and that are different for different fermion species . for the up - type quark and dirac neutrino masses we have @xcite : m_u & = & + ( h v_u + g v_d ) + f _ u , + m_^d & = & - ( h v_u + g v_d ) + 3 f _ u , where @xmath182 are the symmetric yukawa matrices introduced in eq . ( [ stn ] ) and @xmath183 is the up - type induced doublet vev from @xmath45 . if we take e.g. @xmath184 , no particular cancellation is needed for the mass entries related to the up quark mass , while for those related to the heavy third generation , a tuning in the cancellation of up to one part in @xmath185 is required . while this is certainly unpleasant , we should not forget that non - supersymmetric @xmath0 suffers a naturalness problem from the start , which already requires a tuning in the theory at a much higher level than @xmath186 . for @xmath187 ( @xmath188 ) and @xmath189 ( @xmath190 ) . we now turn to the calculation of the relic density of the dm candidate @xmath128 . right above the intermediate scale ( of order tev ) our @xmath0 model corresponds essentially to a low - scale l - r extension of the sm @xcite with the addition of two ( massless ) fermion bi - doublets . after @xmath161 breaking , two degenerate dirac @xmath4 doublets with mass @xmath125 appear . below the electroweak breaking scale the neutral members of these doublets combine into two neutral fermions @xmath127 and @xmath128 with masses @xmath191 . let us recall that @xmath192 while the mass difference with the heavier charged partners is of about @xmath193 ( see the discussion in section [ spectrum ] ) so that both mass splittings are much smaller than the typical dm freeze - out temperature , and have negligible effects on the determination of the dm relic abundance . this is because in the limit of unbroken lr symmetry the fermion states are degenerate members of gauge multiplets , and ` cohannihilation ' channels can be simply taken into account by including appropriate factors of gauge multiplicities both in the annihilation process and in the counting of particle degrees of freedom . however , in our numerical study we have kept track of the small mass splittings induced by symmetry breaking thus differentiating between annihilation and cohannihilation , but the effects of this more refined treatment remain irrelevant . of course , eventually the charged states and the heavier neutral state @xmath127 will all decay to @xmath128 , thus adding their contribution to the dm relic abundance . + before tackling the calculation of the dm relic density in our model , let us first recall the generic features of the analogous computation in the case of a dm sm doublet . the relic density of the neutral component of an @xmath4 doublet of mass @xmath194 can be cast as @xcite : _ dmh^2 0.1 0.1 ( ) ^2 , where @xmath195 is the thermally averaged annihilation cross - section times the relative velocity , and includes sm gauge annihilation and co - annihilation processes . the scaling of @xmath196 with the square of @xmath197 follows from the scaling of the annihilation cross section as @xmath198 , when the dm mass is larger than the sm gauge boson masses . clearly , in the presence of two quasi - degenerate doublets with mass @xmath125 , the same value of @xmath196 would be reproduced for @xmath199 . however , in our case in addition to the sm gauge interactions we have new annihilation channels mediated by @xmath200 and @xmath5 . for @xmath201 the cross section does not scale simply as @xmath202 , and in particular for @xmath203 and @xmath204 annihilation in the @xmath74-channel proceeds via gauge boson resonances which drastically lowers the otherwise large relic density implied by eq . ( [ relicmdm ] ) . therefore , in addition to the solution at @xmath205 , which approximately holds for @xmath206 we expect other phenomenologically viable values of @xmath125 which will depend on the values of the rh gauge boson masses . in order to explore quantitatively the mass parameter space for dm , including also the effects of resonant annihilation and co - annihilation via @xmath207 , we have carried out a numerical analysis by implementing our model in micromegas @xcite . the results are summarized in figure [ fig : relic_density ] where the dependence of the relic density on the dm mass @xmath125 is shown for two benchmark values of @xmath40 : @xmath208 and @xmath209 ( the corresponding @xmath200 mass values are respectively @xmath210 and @xmath211 ) . the first solution corresponds to the expected value @xmath212 . however , other values of @xmath125 become viable close to the resonance regions @xmath203 and @xmath213 . the width of these resonances is large enough to allow for a clear separation of the two crossings that are present for each resonance . all in all , we have five phenomenologically viable values of @xmath125 for each value of @xmath40 . eventually , for very large dm masses @xmath214 we recover the scaling of eq . ( [ relicmdm ] ) . however , in this region the relic densities are at least one order of magnitude larger than the observed @xmath196 . + as we have seen in section [ spectrum ] , dm does not couple diagonally to the neutral @xmath6 gauge bosons . then the leading annihilation channel for indirect detection searches is into @xmath215 and @xmath216 ( respectively via @xmath217-channel exchange of @xmath218 and @xmath127 ) with comparable branching ratios for the two diboson final states ( for the dm solutions corresponding to @xmath204 , see figure [ fig : relic_density ] , @xmath219 final states are also kinematically accessible ) . the velocity - averaged cross - section for @xmath220 can be estimated as @xmath221 with @xmath222 the @xmath72 fine structure constant . a more accurate estimate , including non - relativistic sommerfeld corrections , gives for the same mass range an enhancement up to one order of magnitude @xcite . in spite of this the signal remains well below the present limits @xmath223 for the mass range @xmath224 @xcite . as regards collider limits , at the lhc the most sensitive searches in our scenario , in which @xmath225 are quasi - degenerate , are monojet signatures from processes like @xmath226 , where the pair of @xmath227 s are produced via the @xmath74-channel exchange of a sm gauge boson however , this signal is accompanied by large backgrounds from @xmath228+jets , which render the experimental search particularly difficult . a dedicated analysis of signatures of quasi - degenerate higgsino like dm , which closely resembles our scenario , indicates a surprisingly low reach @xmath229gev even for lhc-13 @xcite . we have no reason to expect that this limit could be largely exceeded in our case , so that we can conclude that even the non - resonant ( lowest mass ) dm solution with @xmath230tev remains unconstrained by collider searches . + before concluding this section , one additional remark is in order . our dirac dm candidates carry hypercharge @xmath231 , which can be used to distinguish particles ( e.g. @xmath232 ) from antiparticles ( respectively @xmath233 ) . during their thermal history , the @xmath227 s will unavoidably enter in chemical equilibrium with the thermal bath , inheriting an asymmetry similar to that of all sm hypercharged states . for example , in the temperature range @xmath234 ( @xmath235 denote the temperatures at which @xmath236 get broken ) that is the relevant range in which dm annihilates efficiently when the dm mass is above a few tev , processes like @xmath237 which occur as long as @xmath238 , or @xmath239 scatterings mediated by @xmath240 effective operators like @xmath241 which occur when @xmath242 , will enforce chemical equilibrium between the @xmath227 system and the sm thermal bath , and thus an asymmetry will develop in the dm sector as well . the issue whether such an asymmetry could play any relevant role in determining the final dm relic density by quenching the annihilations when the dm density becomes of the order of the density asymmetry was recently studied in @xcite for the general case of stable relics belonging to scalar and fermion hypercharged multiplets of dimension @xmath243 . it was found that for fermion doublets , as for most of the other cases , the effects of the dm asymmetry on the surviving relic density are generally negligible . however , the results of the analysis in @xcite rely on certain assumptions , among which : ( i ) there are no new hypercharged particles besides the multiplet in question ; ( ii ) the same operator responsible for the transfer of the asymmetry is also responsible for the mass splitting between the neutral hypercharged states . in the present case however , these two conditions are not satisfied : ( i ) besides the new fermions in the bi - doublets , also the charged gauge bosons @xmath5 carry hypercharge @xmath231 ; ( ii ) the loop operator that induces the mass splitting ( figure [ fig:2 ] ) is generated only after @xmath4 breaking , while the asymmetry transfer is mediated by tree level interactions with real or virtual @xmath5 bosons , and is most efficient well above @xmath244 . moreover , resonantly enhanced annihilation and co - annihilation , that were not present in the scenario analyzed in @xcite , here play a very important role , and many of the solutions for the correct dm relic density are found in dm mass regions close to the gauge boson resonances . nevertheless , in spite of all these differences , given that co - annihilation via @xmath5 exchange , that is one of the dominant annihilation processes , does not suffer from any asymmetry - related quenching , it is reasonable to expect that in most of the parameter space the dm asymmetry will be largely uninfluential in determining the final value of @xmath196 . resonance could be an exception . ] breaking the @xmath0 gut group to the sm via the intermediate group @xmath245 by means of vacuum expectation values in @xmath246 preserves an exact discrete gauge symmetry @xmath9 , which ensures the stability of the lightest among new fermions belonging to @xmath1-dimensional vector representations . we have added to the @xmath0 model two fermionic @xmath1 s , and we have argued that the lightest stable states belonging to these representations correspond to the four _ neutral _ members of the @xmath247 bi - doublets contained in the two @xmath1 s . thus , a first requirement for viable dm candidates , namely electric charge neutrality , is satisfied . after @xmath161 breaking the four neutral states arrange into two dirac spinors describing two types of fermions that are degenerate in mass , and after @xmath4 breaking loop corrections involving the @xmath5 gauge bosons mix these states in such a way that : ( i ) the mass eigenstates get splitted by an amount @xmath248 ; ( ii ) both the @xmath6 neutral gauge bosons couple non - diagonally the lighter neutral fermion to the heavier one . requiring that @xmath248 is large enough to forbid neutral current inelastic scatterings of dm into the heavier fermions allows to evade all the limits from dd dm searches , in spite of the fact that dm is constituted by weakly interacting dirac fermions . we have shown that the quantitative requirement @xmath249 can be satisfied only if the scale of @xmath161 breaking is not too large , and we have derived an _ upper _ limit on the rh gauge bosons mass @xmath250tev . this upper bound strengthen the possibility that the lhc might detect signatures of a low scale lr symmetry , and this is particularly exciting in view of the anomalies recently reported by atlas @xcite and cms @xcite which can all be explained with a rh gauge boson with mass @xmath251 . + before concluding , we should also point out some issues that within our minimal scenario are left open , and that might deserve further studies . in section [ neutrinos ] we have pointed out that , as in all @xmath0 derived low scale lr models , there is not enough suppression for the light neutrino masses from the seesaw mechanism , and to accommodate the neutrino mass scale we had to invoke some amount of tuning in the yukawa sector . a related issue is the fact that the scale of the rh neutrino masses is too low to allow for an explanation of the cosmological baryon asymmetry via the standard leptogenesis scenario @xcite ( see @xcite for a review ) . this is because the rh neutrinos are too light to provide sufficiently large cp violating asymmetries @xcite . finally , we have verified that with the minimal particle content that we have assumed , gauge coupling unification does not occur ; however , we expect that it can be recovered by adding new particles in suitable incomplete @xmath0 representations ( see for example @xcite for different ways to recover gauge coupling unification in low scale lr models ) . we thank davide meloni for discussions in the early stage of this project . we acknowledge financial support from the research grant theoretical astroparticle physics " number 2012cppyp7 under the program prin 2012 funded by the italian `` ministero dellistruzione , universit e della ricerca '' ( miur ) and from the infn `` iniziativa specifica '' theoretical astroparticle physics ( tasp - lnf ) . smb acknowledges support of the spanish micinn s consolider - ingenio 2010 programme under grant multidark csd2009 - 00064 . appendix * the two @xmath108 bi - doublet contained in the @xmath1 of @xmath0 can written in components as : _ , = . @xmath110 are multiplets of weyl fermions coupled to @xmath132 gauge fields through the gauge - invariant kinetic lagrangian : _ k= ^_i_d^_+ ^ _ i_d^ _ , where @xmath252 and @xmath253 with @xmath254 the pauli matrices acting in lorentz space , and d^= ^ -i g_l ^_l -i g_r ^_r , where @xmath255 are pauli matrices acting in @xmath256 group space , i.e. on the first superscript labels of the multiplet components in eq . ( [ ap - bidoublet ] ) , while @xmath257 of @xmath258 act on the second labels , and the reversed vector sign reminds that the action is from the right : @xmath259 . let us now define _ = _ 2 ^ * _ , ^_= ^t__2 . from the free weyl equation for r - chirality spinors @xmath260 and using the relation @xmath261 it is easily seen that @xmath262 satisfies @xmath263 and thus it contains @xmath130-chirality spinors . in terms of @xmath262 the second term in eq . ( [ weyl ] ) can be rewritten as : ^ _ i_d^_&= & ^ _ i _ ^ _ + & + & ^_i _ _ where we have integrated by parts the derivative term ( neglecting a @xmath264-divergence ) and an overall change of sign is due to anticommutation of the fermion fields . ( [ zetal ] ) shows explicitly that @xmath265 transforms in the @xmath247 conjugate representation @xmath266 with generators @xmath267 , where the last relation expresses the pseudoreality of @xmath72 representations . it is then convenient to define new @xmath265 multiplets transforming similarly to @xmath268 in @xmath269 : _ & = & . the neutral current interactions for @xmath270 ( @xmath271 ) read : ^nc & = & j_^nc ( g_lw^_3l - g_r w^_3r ) , + j_^nc & = & - _ i=1,2 _ t_3 _ i , with @xmath272 . in order to write @xmath273 in terms of the mass eigenstates , let us study how the mass terms are rewritten from the original basis of lh and rh weyl spinors @xmath110 to our basis of lh spinors @xmath268 and @xmath265 . the fermion bilinears multiplying the tree level mass term for charged and neutral states eq . ( [ eq : mb ] ) are rewritten as : m_b tr(^__+ ^ _ _ ) & = & -m_btr ( ^t__2 _ + ^__2 ^ * _ ) . written explicitly : ( ^t__2 _ ) & = & - ( ^+-_)^t_2 ^ -+_- ( ^-+_)^t_2^+- _ + & & + ( ^_)^t_2^++_+ ( ^++_)^t_2^_. similarly , the @xmath247 breaking loop induced mass term for the neutral states ( [ eq : dm ] ) reads : _ m , so that the lagrangian for the masses of the neutral states is : -_m^0 = _ 2^t _ 2 _ 1 + h.c . with = . being @xmath274 symmetric it can be factorized as @xmath275 with @xmath276 unitary and @xmath100 diagonal with real and positive eigenvalues . the two eigenvalues are @xmath277 , and the heavy and light mass eigenstates @xmath278 and @xmath279 are given by @xmath280 with v = . by redefining now @xmath281 and @xmath282 , the mass term can be written as : -_m^0 = m_l _ l + m_h _ h , where we have introduced the four spinor @xmath283^t$ ] and a similar one for @xmath127 , and we have adopted the usual convention @xmath284 with @xmath285 in the chiral basis . thus , upon diagonalization of the mass matrix the new fermions organize into two dirac mass eigenstates splitted in mass by @xmath286 . coming back to the neutral current gauge interactions , after rotating the interaction eigenstates @xmath270 in eq . ( [ neutral ] ) onto mass eigenstates , the neutral current reads : j_^nc & = & - _ i=1,2 ^_i i _ ( vt_3 v^ ) _ i + & = & _ _ l + h.c . , with @xmath287 . we see that the neutral gauge bosons couple the light mass eigenstates to the heavy ones , a result that follows from the fact that @xmath288 is anti - diagonal ( @xmath289 denotes the second pauli matrix , but with no relation here with gauge group factors or spinors ) . obviously , since @xmath290 the purely kinetic term for the mass eigenstates remains diagonal .	@xmath0 grand unified theories can ensure the stability of new particles in terms of the gauge group structure itself , and in this respect are well suited to accommodate dark matter ( dm ) candidates in the form of new stable massive particles . we introduce new fermions in two vector @xmath1 representations . when @xmath0 is broken to the standard model by a minimal @xmath2 scalar sector with @xmath3 as intermediate symmetry group , the resulting lightest new states are two dirac fermions corresponding to combinations of the neutral members of the @xmath4 doublets in the @xmath1 s , which get splitted in mass by loop corrections involving @xmath5 . the resulting lighter mass eigenstate is stable , and has only non - diagonal @xmath6 neutral current couplings to the heavier neutral state . direct detection searches are evaded if the mass splitting is sufficiently large to suppress kinematically inelastic light - to - heavy scatterings . by requiring that this condition is satisfied , we obtain the upper limit @xmath7 tev . dark matter , grand unified theories
when high - energy cosmic rays bombard the earth , muons are produced as part of particle showers within the upper atmosphere . these highly - penetrating particles are observed at sea level with a flux of approximately one per square centimetre per minute and momenta of several gevc@xmath1 . as charged particles , they interact with matter primarily through ionising interactions with atomic electrons and via coulomb scattering from nuclei . both of these mechanisms have been exploited in recent years in the field of muon tomography ( mt ) to probe the internal composition of shielded structures which can not be probed using convential forms of imaging radiation _ e.g._x - rays . since e.p . george measured the thickness of the ice burden above the guthega - munyang tunnel in australia in the 1950s @xcite and l.w . alvarez conducted his search for hidden chambers in the second pyramid of chephren in egypt @xcite a decade later , there has been a wealth of wide - ranging applications using cosmic - ray muons for imaging purposes , such as in the field of volcanology @xcite and nuclear contraband detection for national security @xcite . the seminal work outlined by borozdin _ et al._in ref . @xcite revealed the potential to locate and characterise materials within shielded containers using the coulomb scattering of cosmic - ray muons . this approach relies on precision reconstruction of the initial and scattered muon trajectories to determine the scattering location within the container and the scattering density , denoted @xmath2 . this scattering density is known to exhibit an inherent dependence on the atomic number @xmath0 of the material @xcite _ i.e._larger scattering angles are typically observed for objects with larger @xmath0 values . this work presents the design and fabrication processes of a small prototype mt system for use in the identification and characterisation of legacy nuclear waste materials stored in highly - engineered industrial waste containment structures . for this purpose , a detector with high spatial resolution was required . it was essential that the design of the system , and the materials and fabrication processes used , be scaleable to allow the future construction of an industrially deployable detector system . the requirement of industrial deployability mandated that the detection medium be radiation - hard , robust and capable of performing to a high degree of stability over prolonged periods of time for operation in the typical high radiation , high stress industrial environment found within a nuclear waste processing plant . a modular design based on plastic scintillating fibres was chosen to satisfy these criteria . the individual components and modular construction are described in section [ sec : components ] with the assembly process outlined in section [ sec : assembly ] . the readout electronics , data acquisition ( daq ) and complex fibre multiplexing requirements are described in section [ sec : mapping ] . results showing the performance of the constructed , commissioned system are presented in section [ sec : commissioning ] and first results from the image reconstruction of a test configuration of low , medium and high-@xmath0 materials are presented in section [ sec : images ] . each detector module comprised orthogonal detection planes of plastic scintillating fibres which were supported and positioned on a low - density rigid foam and aluminium structure . the two planes of fibres were bonded within this structure and further supported at the extremities by plastic polymer distribution blocks . fibres were coupled to photon detectors ( here , multi - anode photomultiplier tubes ) housed within lightproof boxes . the active area of each module was encased within a lightproof vinyl film . the design of the individual detector modules and all components and materials used in the fabrication process are detailed in the following section . prior to the design and fabrication processes , dedicated geant4 @xcite simulation studies were performed to assess the fibre pitch required for this system with regards to the anticipated light output and reconstructed image resolution @xcite . a design based on 2mm - pitch plastic scintillating fibres was chosen . the fibres used in the production of this muon tracker were saint - gobain bcf-10 round fibres comprising a polystyrene - based core ( 97% by cross - sectional width ) with polymethylmethacrylate ( pmma ) optical cladding ( 3% ) . the light emission output of this formulation , which peaked at 432 nm , provided excellent overlap with the sensitivity of the chosen photon detector , the hamamatsu h8500 mapmt . feasibility studies ( outlined in section [ sec : mapping ] ) revealed the potential to reduce the required number of readout channels by 50% through the coupling of two scintillating fibres to a single pixel on the 64-pixel segmented anode of the pmt . a single detection layer comprising 128 fibres coupled to a single pmt was therefore chosen for the final prototype design . the design for the complete detector system comprised four modules , two situated above , and two below the volume under interrogation _ i.e._the assay volume . with this modular design and the identification of a space point in each module , the initial and scattered muon trajectories could be reconstructed allowing the determination of the scattering position and magnitude required for image reconstruction . within each module , two orthogonal detection planes , comprising a single layer of 128 fibres , yielded an active area of 256 mm x 256 mm . to allow for minimal transmission losses and manageable strain as a result of the fibre bending required for pmt coupling , fibres of 860 mm length were used . in total , for eight detection layers , 1024 fibres and 8 pmts were required . . ] for each fibre layer , accurate fibre positioning and support was provided by a low-@xmath0 , precision - machined structure of rohacell^^ sheeting , a closed - cell rigid polymethacrylimide foam . this provided sufficiently high tensile strength in comparatively small thicknesses to support the individual fibre layers whilst providing only a negligible coulomb scattering effect on transient muons . a layered configuration of rohacell^^ support sheets of varying thicknesses and dimensions was fabricated to support the orthogonal layers of fibres . the assembly of this structure is described in detail in section [ sub : module ] . shown in figure [ fig : rohacell ] and measuring 300 mm in length , the central square sheet that supported the fibres was machined with shallow , parallel _ v _ grooves to position each of the 128 fibres per layer . grooves were cut in the x ( y ) direction on the top ( bottom ) side of the support sheet . on this central sheet , narrow non - grooved regions were located at the sides parallel to the fibre direction ( shown in figure [ fig : rohacell ] ) to accommodate the holes which allowed the sheet to be fixed onto locating pins on the aluminium base plate ( see section [ sub : aluminiumbaseplate ] ) . each sheet of rohacell^^ had locating holes in each corner to fix the sheet to these locating pins to ensure accurate alignment and uniformity across all four modules . a 3mm - thick flat sheet of aluminium was used to provide a robust support base for the multiple layers of rohacell^^ and scintillating fibres which formed the active area of each module . this square sheet of aluminium , shown in figure [ fig : al - plate ] , measured 460 mm with a central square region of 270 mm removed to minimise the material contributing towards coulomb scattering within the active volume . four stainless - steel locating pins were fixed at each corner of the inner hole in the base plate . these were used to secure the sheets of rohacell^^ and to provide fixed reference points which ensured precision alignment was maintained both internally ( _ i.e._with the fibres within the corresponding module ) and externally ( _ i.e._with the other three modules ) . at the sides of the aluminium plate , two locating holes were bored . these holes ( the innermost holes on figure [ fig : al - plate ] ) allowed fibre distribution blocks to be located on each axis and screwed into place . each side of the base plate had two further holes bored at the outer corners . these ensured the aluminium sheet could be fixed to an external aluminium profile stand . to provide extra support to the fibres , custom - built distribution blocks were fabricated from polyoxymethylene ( pom ) plastic . these also maintained the uniform distribution of fibres at the extremes of the base plate and , more importantly , at the pixelated surface of the pmt . this was chosen over other potentially abrasive candidate materials _ e.g._aluminium , to prevent causing damage to the fibres . the detector benefitted from the low cost , opacity and precision machining capability of this polymer . the first set of distribution blocks were rectangular pom pieces shown in figure [ fig : fibredistblock ] . these were located at the extreme of the base plate and in the same plane as the fibre layers . they contained two rows of 64 holes and measured 300 mm in length . the rows were both offset slightly , one above and one below , from the level of the fibres to provide support and stability to the fibres and to prevent unwanted strain causing them to lift off from the rohacell^^ surface after bonding . this also ensured that a constant tension was maintained across all fibres . the fibre distribution blocks were fixed in place on the aluminium base by screws through the holes in the plate . the holes which supported the fibres were large enough to allow the application of black nylon tubing to the fibres for lightproofing . this tubing covered the length of each individual fibre outwith the area covered by the base plate and mitigated the risk of external sources of light incident on the fibre from producing signals within the pmt . this tubing , and additional lightproofing measures , are outlined in section [ sub : lighttight ] . ) . ] the second pom block shown in figure [ fig : pmtholder ] measured 110mmx82mmx42 mm . it served to hold the pmt in place in an inner square cavity ( of equal dimensions to the pmt ) at the rear side , whilst at the front , it provided precision contact between the pixelated surface of the pmt and each coupled fibre pair . this latter characteristic was achieved via an array of 128 holes , positioned such that two fibres were in optical contact with a single pixel . once inserted into the holes of the pmt holder , the fibres protruded a short distance due to a series of stainless steel collars which were bonded to the end of the fibres ( see section [ sub : fibreassembly ] ) . these prevented the fibres extending further into the pmt holder and thus ensured that the small lengths of fibre that protruded uniformly contacted the pmt surface . this process also ensured that all the fibres exerted the same pressure on the surface of the pmt . a thin silicone gel protective pad , which had negligible effect on the transmission of light , was placed between the fibres and pmt to mitigate potential damage to the surface of the pmt by the pressure of the fibres . four threaded holes were positioned on the rear surface to allow a printed circuit board ( pcb ) to be connected for signal readout purposes . surrounding the edge of the pmt holder , a raised border allowed the holder to sit firmly within a specially - constructed housing on the aluminium profile frame , shown in figure [ fig : frame ] . all components outlined in the previous subsections were housed in an aluminium profile frame ( with 20 mm square cross - sectional area ) shown schematically in figure [ fig : frame ] . aluminium profile was chosen because of its lightweight structure and strength . this was constructed to support the base plate ( onto which it was screwed ) , rohacell^^ and orthogonal layers of fibres , in addition to preventing unwanted strain on the fibres in the non - active regions which could result in fibre breakages and/or loss of contact with the pmt surface . to minimise these risks , aluminium sheets were attached to the frame structure to provide further support to the fibres in this region . at the end of these two arms , three small aluminium profile struts were assembled to fix the pmt holders firmly in position . these struts were designed to accommodate the raised borders of the holders . the design of the module frame benefitted from the capability of aluminium profile to attach to an external support frame at any position and orientation along its length . this allowed complete freedom for future experimental studies on alignment and module positioning . custom - made pcbs were printed to couple to the h8500 pmt connector pins and allow the connection of 16-channel ribbon cables to facilitate the readout of signals from the pixels . the readout system will be described in detail in section [ sec : mapping ] . these pcbs were tightly - screwed to the pmt holder via the holes on either side . this helped provide uniform contact between the fibres and the face of the pmt . a final , rectangular pom box measuring 140mmx113mmx55 mm was fabricated and bolted to the aluminium profile to fully encase and lightproof the pmt and the associated voltage supply cables , the pcb and readout signal cables . this pom box is shown in figure [ fig : pmtbox ] . this was bolted to the aluminium profile of the module via four oblique holes in the corners . a rectangular slit at one side of the box allowed the four ribbon signal cables per pmt to be connected . a small groove at each end of the box provided access for the high voltage cable and ensured it was not damaged during assembly . a circular hole was drilled out at the back of the pmt cover to allow the signal cable to be inserted . this section will detail the preparatory and assembly processes associated with the fabrication of the detector modules , and the lightproofing measures taken . prior to assembly , both ends of the scintillating fibres were polished to optimise light transmission and to provide good optical contact with the pixelated surface of the pmts . a three - stage polishing process was undertaken using a series of lapping films of decreasing coarseness . a vital component of the assembly process was to maintain a uniform contact and pressure distribution between the fibres and the pmt . this was achieved using a series of metal tubes ( or collars ) fixed to one end of the fibres at precise positions . at this end of each fibre , a 30mm - long stainless steel collar was bonded with optical epoxy ensuring that a small length of fibre protruded from the collar . once the epoxy and collar had been applied , the epoxy was left to cure . after curing , a further lapping stage removed this protruding length of fibre and , with it , any epoxy covering the polished end . a smaller 5mm - long stainless steel collar was bonded onto the larger collar using a uv - curing adhesive . removable caps were placed on the fibres during this process to ensure that the small collars were all attached at a pre - set position along the larger collar . this ensured that every collared fibre protruded the same distance through the pmt holder , providing a uniform pressure on the pmt surface . this process of attaching the smaller collars was performed after the fibres had been bonded to the rohacell^^ sheets and covered in black nylon tubing , which are detailed in the following sections . figure [ fig : modulelayers ] illustrates the final layered structure of scintillating fibres , rohacell^^ and aluminium which comprised each detector module . the scintillating fibres were first bonded to the central , grooved rohacell^^ sheet using the same optical epoxy used for the bonding of the large collars . the grooved sheet of rohacell^^ was first attached to the aluminium base plate via the four locating pins . this base plate was clamped to a breadboard table with flat blocks of pom abutting the end of the rohacell^^ sheet , as shown in figure [ fig : fibres ] . this ensured that , when bonded , the open end of the fibres rested level with the edge of their support sheet . the repellent properties of pom ensured that the epoxy did not bond to these blocks during this process . epoxy was then applied to the top side of the sheet before fibres ( these would become the bottom @xmath3 plane ) were carefully positioned into the grooves . a flat , 4mm - thick sheet of rohacell^^ of otherwise equal dimensions to the grooved central sheet , was then placed over the locating pins and bonded on top of the layer of fibres . this was weighted evenly to ensure the fibres cured in their desired positions within the grooves . with this layer of fibres set in place , the configuration was then overturned to allow the top x plane of fibres to be bonded . this was placed onto the base plate once again , over a 6mm - thick piece of rohacell^^ of equal dimensions as the aluminium plate . to minimise the material in the active volume , the central square region of this piece , measuring 270 mm , was also removed . for the x plane of fibres , the same gluing procedure was performed with another 4mm - thick sheet of rohacell^^ bonded and weighted on top during the curing process . a 6mm - thick flat piece of rohacell^^ , the same dimensions as the other central pieces , was secured on top of the support pins . this sheet was necessary to ensure a level height with the pom distribution blocks screwed to the edge of the base plate . the fibres were then fed through these blocks with care taken to avoid damage . a final sheet with the central region removed was pinned to the structure to fully enclose the fibres . this was secured tightly with washers over the four locating pins . once the layered configuration had been fabricated , a rohacell^^ skirting , shown in figure [ fig : modulelayers ] , was pinned around the open edges of the module to seal off the active area and to provide a smooth - edged support for the lightproof covering described in the following subsection . this whole assembly was then secured in the aluminium profile module frame using purpose - built bolts placed in the troughs of profile ( shown previously in figure [ fig : frame ] ) which were then bolted to the holes in the base plate . each module was then prepared for final lightproofing measures . prior to the optical connection of the fibres to the pmts , the exposed regions outwith the encased active area , and the areas covered by the porous rohacell^^ , required lightproofing to prevent light from entering the active area and producing noise signals during data collection . to achieve this , lengths of black nylon tubing were applied over the collared ends ( prior to the bonding of the smaller collar ) forming a tight seal against the large aluminium collar , while leaving sufficient space for the small collar to be bonded . at the opposite end , the tubing was positioned tightly inside the holes of the fibre distribution block providing a firm seal . when all the fibres were covered in the tubing , each fibre was labelled sequentially to allow the fibre multiplexing schemes outlined in section [ sec : mapping ] to be implemented . the rohacell^^ structure around the active area was covered in sheets of tedlar^^ , a radiation - durable polyvinyl fluoride film with excellent light absorption properties , which ensured the fibres were not exposed to any external light . two layers of this film were wrapped around the central area of the module with four holes pierced in the film to allow the locating pins to protrude through . once in place , black electrical tape was applied around all the edges to hold the tedlar^^ in position and a black foam seal was placed over the locating pins and taped in place to restrict any light penetrating into the active area . the fully lightproof , and pmt - connected , detector module is shown in figure [ fig : tedlar ] . with the construction of the four detector modules completed , and lightproof testing successfully performed , the modules were assembled in an aluminium profile vertical support stand . the modules were precisely aligned , to make certain that all four were parallel , and positioned at predetermined , optimum positions in the vertical @xmath4 direction . these positions were extensively studied using detailed geant4 simulations of the detector geometry to optimise the tracked muon flux and reconstructed image resolution . a maximum separation of 900 mm between the outermost modules was imposed to negate any destabilising effects introduced by extending to longer lengths of aluminium profile for the vertical stand . limiting this to 900 mm ensured the stability and alignment of the system . the separations between the top and bottom module pairs were fixed at 250 mm leaving a 400 mm spacing in the assay volume to accommodate test objects for future image reconstruction . the individual modules were attached to the support stand via a series of clamps around the central region . these ensured firm and stable attachment to the vertical stand , and allowed the modules to be connected at any position in the vertical direction if required . the clamps were designed with two holes on the outside and one on the raised , middle section . this middle hole provided the ability for the clamp to be bolted onto the vertical struts of the external detector stand , while the outside holes bolted onto each module frame . the four clamps were positioned at the active area of each module with two on the outside of the frame and two on the inner side . a model of the detector , fully assembled in the vertical support structure , is shown in figure [ fig : gmtdetector ] . the square base of the vertical support stand was firmly secured to a breadboard surface to provide additional stability . alignment studies , presented in section [ sub : alignment ] , revealed the high degree of precision resulting from the various support measures taken during the assembly of this detector system . the hamamatsu pmts offer one negative - voltage read out channel per pixel and a unified signal from the second - to - last dynode ( dynode-12 or dy12 ) in the amplification chain with positive polarity that is sensitive to a signal in any pixel . the dy12 signals were used in this system to trigger the read out and recording of the pmt signal . in order to reduce noise , a coincidence requirement was imposed ; data were only recorded when three separate pmts recorded signals in any pixel . the system made use of a nim electronics crate housing coincidence gates , dual gate generators , discriminators and fan - in / fan - out units . data collection was performed via a custom c++ data acquisition program ( daq ) running on a commodity laptop . this was connected via usb to a caen vme crate and the pixel signals were fed isomorphically into the signal inputs of caen charge to digital converter units ( qdcs ) . the vme crate also housed a scaler unit that was used to monitor signal peaks above a pre - set threshold voltage . the positive polarity dy12 signals were fed into the nim electronics crate , inverted , amplified and shaped before being passed to a set of coincidence units that produced the final physics trigger . when the trigger condition was satisfied and the qdcs had recorded the analog signal , the daq software transferred the data from the qdc registers over the vme bus . the data were zipped and written to disk before the next event was recorded . this potential delay had no significant impact on the detection efficiency due to the relatively slow event rate . meanwhile , the scalers monitored the number of signals received from each pmt ( each dy12 signal was fed to an individual scalar channel ) , the number of read out gates generated by the nim hardware , the number of trigger gates generated by the daq software and the number of events that were read out and recorded . for each event , the data and scaler signals were packaged together with a timestamp and status information on the qdc modules . the information from the scaler was used for both hardware and software debugging purposes ; any light leaks are reflected in the number of signals generated by the pmts while inefficiencies would be found if the number of physics triggers and the number of final triggers diverged . during normal data collection , events were recorded in runs of 500 . at the beginning of every data run , pedestal runs were recorded to monitor the amount of non - signal charge in the qdcs . as explained in section 5 , these pedestal files were used to identify any fluctuations in the electronics performance and were used during data quality monitoring to correct the corresponding data signals from the system prior to pmt gain correction . in order to minimise the number of read - out channels required for the experimental setup , the signals from all 128 scintillating fibres per detection layer were read out by a single hamamatsu h8500 mapmt . here the anode is segmented into a 64-pixel square array , with each pixel measuring 5.8 mm in width . two fibres were coupled to each pixel via dedicated pairing ( or multiplexing ) schemes , an example of which is shown in figure [ fig : mapping ] . simulation studies performed using a geant4 simulation of the detector geometry prior to the construction of the prototype system showed that these multiplexing schemes allowed the correct identification of the eight struck fibres per event , thus maintaining the high spatial resolution provided by the 2 mm pitch . this was achieved using a likelihood - based demultiplexing algorithm which relied on a small scattering angle assumption and the use of four unique multiplexing schemes , one in each of the four detection layers in the top and bottom pairs of modules . these had the effect of exaggerating the detected incident angles of the two muon trajectories for incorrectly - identified fibres such that the two partial tracks do not combine , or if so , are significantly less likely to have a scattering angle below a given threshold . from these simulation studies , the correct eight fibres struck per event were successfully identified in over 98% of events . further studies have shown this small misidentification of fibres which remains in the imaged data to have a negligible detrimental effect on the images reconstructed . once construction of the prototype detector system was completed and all daq systems had been tested , data were collected for commissioning and alignment purposes . for the studies performed throughout this work , excluding the pmt gain normalisation and signal transmission studies ( the experimental setups for these tests are described in sections [ sub : gain ] and [ sub : fibretest ] respectively ) , the three - fold daq trigger described previously was implemented . to recap , the top and bottommost layers were included in coincidence with one of the innermost layers to ensure confidence that the trigger originated from a single muon passing through the complete detector acceptance . for studies performed using the complete system ( studies in section [ sub : qdc ] onwards ) , dedicated pedestal - only and data runs were recorded at regular intervals during data collection . the identification of the struck fibre was made by first requiring events in which the five layers outwith the trigger conditions registered a signal , in at least one channel , with a qdc signal 3@xmath5 ( from a gaussian fit to the pedestal - only data ) above the pedestal mean . the pixel ( or pixel cluster ) coupled to the struck fibre was identified by selecting the highest pedestal - subtracted , gain - normalised qdc signal per pmt . this yielded two potential fibres per detection layer as a consequence of the fibre multiplexing schemes . the eight struck fibres per event , and therefore the four space points , were reconstructed via the likelihood - based demultiplexing algorithm described previously . prior to data collection on the full prototype system , each of the eight hamamatsu h8500 mapmts to be used required detailed characterisation to assess the anticipated effects from cross - talk , clustering and local gain variations across the pixelated surface . results from independent , detailed studies on the cross - talk and clustering characteristics of the h8500 pmt in ref . @xcite by montgomery _ et al._revealed these effects to be small . observations made during data collection consolidated these findings , revealing the influence of cross - talk on the surface of the pmt to be small in relation to the desired signal . typical cluster sizes of 2 to 3 pixels , and mean cluster multiplicities of around 1.5 per pmt were observed . although coupled on diagonally - opposite corners of the pixel and not the centre , the fibres ( with collars ) only occupied 23% of the active area of the pixel . this ensured that the majority of the signal was deposited within a single , distinct pixel . the effect of cross - talk resulting from light transmission to neighbouring fibres was also investigated and found to have a negligible influence . there was also no significant contribution arising from any dispersive effects introduced by the protective silicone pad applied to the face of each pmt . as a consequence of this small cross - talk effect , the pixel corresponding to the qdc channel with the highest gain - normalised , pedestal - subtracted signal per pmt was considered as being coupled to the struck fibre . despite this , in approximately 5% of events , there existed a secondary ( or in even fewer cases , a tertiary ) cluster with a larger integrated signal than the cluster containing the pixel with the highest individual signal . assignment of the corresponding fibres in such instances ( in the vast majority of cases this equated to a single layer ) manifested clear discrepancies in alignment data which were not observed when the pixel with the highest signal was selected . these clusters were subsequently attributed to noise within the system . it was concluded from these observations , that only the pixel with the highest signal per pmt should be considered for selection . this was the case for all studies presented in this work . negligence of any local gain variation effects across the pmt surface could , if their extent was large enough , result in the misidentification of the struck fibres through the selection of the pmt pixel with the highest signal . to minimise the risk of this occurring , low and high resolution scans of each pmt were performed to determine the relative variations in gain across their surfaces . these were performed using a laser tuned to the expected light output level of the scintillating fibres via a combination of filters . scans were undertaken at the operating voltages of the pmts , with the low resolution scans stepping from the centre of each pmt pixel to the next , and the high resolution studies employing a step size of 1 mm . the efficiency and uniformity of the dynode-12 signal were also studied during these laser scans to assess their potential use within the experimental trigger condition . results from the low resolution laser scans of one of the pmts used in the prototype system are shown in figure [ fig : gain ] . these show the relative pixel gains normalised to the highest gain per pmt . in rare , extreme cases , pixels are shown to drop to 40% of the maximum gain of the pmt , with uniformities ( defined here as the ratio of the maximum pixel gain to the mean gain ) ranging from 1.2 to 1.6 across the eight pmts . with only a single pmt per detection layer on which to identify the hit pixel , it was not necessary to extend this localised normalisation across all eight pmts . the light transmission through the fibres was studied using a single detector module prior to assembly in the vertical support stand . small scintillator paddles , covering an area of approximately 50mmx50 mm , were placed above and below the active area to provide an external muon trigger which removed any internal thresholds from biasing the efficiency measurement . the entire length of the active area was scanned with selected positions along the remaining length of fibres to assess the extent of the anticipated degradation . all measurements are presented normalised to the value ( this is set to 100 ) obtained at the face of the pmt . these studies confirmed the expected decrease in signal strength at increasing distances from the pmt due to the light transmission properties of the fibre , loss of light through the open end of the fibre , and the bending of the fibres at the pmt . figure [ fig : fibretest ] reveals the extent of this degradation in the observed signal in the region of this bending . the signal remains relatively constant across the active area of the module at a value of approximately 35% that of the signal observed at the pmt face . the corresponding detection efficiencies ( defined in section [ sub : eff ] ) also remain constant at values in the region of 75% except in the final efficiency measurement . it was subsequently discovered that the scintillator paddles did not fully cover the active area and as such , a larger uncertainty was attributed to account for this . all results were obtained using the same 3@xmath5 selection criteria for the signal size above pedestal , and confirmed that the degraded signals observed were still sufficiently large enough to avoid failing this requirement . values observed from pedestal - only data for all 512 qdc channels prior to ( red filled circles ) and after ( blue open circles ) optimisation of the detector and readout setup . in particular , channels in the qdc channel ranges [ 463 , 478 ] and [ 479 , 494 ] corresponded to two 16-channel ribbon cables which were later found to be severely kinked . this induced noise on the corresponding channels . in both sets of data , the characteristic parabolic distribution in @xmath5 values is observed across the cable . here , the bold ( dashed ) lines separate regions belonging to the same detector layer ( readout cable ) . ] throughout the commissioning process of the detector system , minor adjustments and improvements were made to the setup in preparation for image reconstruction studies . dedicated pedestal - only runs were recorded at regular intervals to allow stability assessment . the widths of the pedestal distribution in each qdc channel , characterised by the @xmath5 ( alternatively , @xmath6 in figure [ fig : sigmas ] ) from an applied gaussian fit , were extracted . across all , the mean @xmath5 was observed to be less than one qdc channel . however , in channels read out by connectors at the edges of the ribbon cables , characteristically - broader pedestals were found with @xmath5 values in the region of two channels . this is an understood effect relating to the placement of the grounding pins . the @xmath5 values for each of the 512 read out channels are shown in figure [ fig : sigmas ] for the optimised setup in comparison with initial pedestal data collected using what later transpired to be damaged ribbon cables which introduced noise into the system . in each of the five fibre layers that did not form part of the trigger scheme , the detection efficiencies were determined as the percentage of recorded events in which the layer registered at least one qdc signal more than 3@xmath5 ( from a gaussian fit to the pedestal - only data ) above the pedestal mean . the detected efficiencies varied between 70% and 86% across the five layers with a stable mean operating efficiency of 34% shown in figure [ fig : stability ] . this value is defined as the percentage of events within the 3-fold trigger which contained at least one qdc signal above pedestal in each of the other detection layers and indicated a mean detection efficiency of 80% per layer . the deficiencies observed relate to fibres which were damaged during the construction or assembly processes and/or dead or noisy channels . these channels included those qdc channels connected at the extremes of the ribbon cables whose pedestals exhibited a characteristically - broader distribution . here , the 3@xmath5 upper threshold imposed could restrict the detection of muons which deposited low signals within the fibre . this effect was observed in figure [ fig : sigmas ] which compares the @xmath5 values for each channel before and after optimisation of the setup . a further source of signal loss could have been a potentially weak signal originating from the muon interacting with a small volume of scintillating material close to the edge of the fibre . this 80% efficiency could therefore translate to the detection of a muon signal only when it had interacted with the central 80% ( by diameter _ i.e._1.55 mm ) of the active circular cross - section of the fibre . however , this assumption can not be verified with current experimental data . further studies were performed by changing the innermost detection layer which was included in the trigger scheme , thus allowing the efficiency of the layer regularly used for triggering to be determined . this allowed the efficiency of this layer to be determined , and it was found to be consistent with the range identified previously . position - sensitive studies , similar to those shown in figure [ fig : fibretest ] , were performed using small external scintillator paddles on the outermost detector layers which confirmed these layers consistency with the six internal layers . t , the time between successive triggers using the three - fold condition described in the text . the black line through the data points represents a second - order polynomial fit to the data . a stable rate of 0.15hz was observed over the entire data collection period , corresponding to a mean @xmath7 t of approximately 6.57s . ] x and @xmath8y axes arise from the small degree of misidentification of the hit fibre from the demultiplexing algorithm and/or from selection of a noise signal in one or more of the pmts.,title="fig : " ] x and @xmath8y axes arise from the small degree of misidentification of the hit fibre from the demultiplexing algorithm and/or from selection of a noise signal in one or more of the pmts.,title="fig : " ] throughout data collection , the prototype mt system exhibited high levels of stability which was observed in various facets of the data including the detection rates and efficiencies shown in figure [ fig : stability ] for a one month period . a steady 3-fold trigger rate of 0.15hz was recorded which equated to approximately 6000 candidate tracking events ( _ i.e._with a signal in every layer ) per day . the interaction rate of cosmic muons with matter follows a poisson - distribution . thus the time of arrival between two events @xmath7 t exhibits an exponential decay . from data collected , the semi - logarithmic distribution of @xmath7 t is shown in figure [ fig : rates ] with a mean @xmath7 t of 6.57s . nevertheless , the time between two events might exceed several tens of seconds . data were collected in 2012 with the assay volume empty of material to establish the extent of any misalignment between detector modules . despite the precision fabrication of the aluminium profile frame and the individual detector modules , minor misalignments were observed from the analysed data . these were determined via an iterative process of constructing and projecting the track from one module pair to the four other detector layers and minimising the residuals @xmath8x and @xmath8y , defined as the difference between the measured and projected positions in each plane . once the misalignments in these layers were accounted for , the process was repeated using the track formed by these modules . this process continued until the residuals in all eight layers were no greater than the fibre pitch . the largest misalignment observed was approximately 5 mm in both the x and y planes of one module . this is shown before and after alignment correction for this particular module in figure [ fig : alignment ] . the small misalignments identified revealed the internal alignment of the fibre planes within the module to be negligible in relation to the external alignment of the modules themselves . the small fraction of misidentified fibres arise from the demultiplexing algorithm and from noise hits which manifest as tails along the @xmath8x and @xmath8y axes in figure [ fig : alignment ] . with a small scattering angle assumption in place for data taken with material within the assay volume , placing a restriction on these tails further reduced the influence from fibre misidentification . -direction ) surrounding a 12 mm diameter stainless steel cylindrical rod and a 30 mm long , 20 mm diameter cylinder of machined uranium metal ( right _ i.e._the positive @xmath3-direction ) suspended below . the aluminium profile frame , ribbon cables , tedlar^^ and locating pin coverings described in the text are also observed . ] in preparation for image reconstruction studies , data collection commenced in late-2012 with a test configuration of objects placed within the assay volume . this setup is shown in figure [ fig : testsetup ] . this consisted of a stainless - steel cylindrical bar measuring 12 mm in diameter positioned through a 40 mm cube of lead . a machined cylinder of uranium metal , 20 mm in diameter and 30 mm in length , was suspended beneath the bar . this bar was fixed to the aluminium profile frame along the y - direction , centred on z=0 mm and x=@xmath95 mm in the coordinate frame used throughout this work . an image reconstruction algorithm based on the probabilistic maximum likelihood expectation maximisation ( mlem ) method introduced in ref . @xcite by schultz _ et al._was further developed for this application . the assay volume was chosen as a cube of dimension 300 mm in the central region of the detector assembly . prior to the imaging analysis , this volume was divided into small volume elements ( alternatively , voxels ) . voxel dimensions are predetermined by the analyser and are at the heart of the trade - off between image resolution and measurement time ; smaller voxels provide potentially greater definition up to the resolution of the detector system but require correspondingly greater measurement times . cubic voxels , 10 mm in dimension , were used throughout this work as simulation showed that this size provided the best compromise between resolution and data collection time . for each muon , the incoming and scattered vectors were first projected to their point of closest approach ( poca ) . the mlem method then requires the calculation of a normalised scattering probability in every voxel that the muon was determined to have passed through . after many muons , the most likely scattering density in each voxel was determined . this was used as the imaging metric in the final results presented in this work . first image reconstruction results from experimental data taken using this prototype detector system are presented in figure [ fig : image ] . shown are two 10 mm slices ( or tomograms ) , one through the xy - plane ( _ i.e._parallel with the detection planes ) which encompassed the stainless - steel bar , and the other in the yz - plane . these images have been reconstructed from several weeks of exposure to cosmic - ray muons . sensitivity to atomic number @xmath0 and discrimination between the @xmath2 values of the stainless - steel bar , the two high-@xmath0 material blocks , and the surrounding air is observed . the non - uniformity of the @xmath2 values of the reconstructed high-@xmath0 objects is attributed to a combination of possible factors ; a spread in @xmath2 values for high - density materials as a result of increased coulomb scattering ( here , the poca input to the mlem method reconstructed only the average scatter ) and non - uniform voxel coverage of the objects ( _ e.g._for the uranium object , the central voxel with a reconstructed @xmath2 in excess of 70mrad@xmath10cm@xmath1 was assumed to fully occupy the uranium , whereas the surrounding voxels occupy a combination of uranium and air which acted to dilute the reconstructed @xmath2 value ) . in the presented tomograms , the high image - resolution of this detector system in the xy - plane is observed . smearing and stretching of the image in the z - direction is also noted . this is an inherent effect associated with the reconstruction of the scattering location in the principle axis of two near - parallel tracks , and is artificially exaggerated in this work by the small angular acceptance of the prototype system . studies which address this issue , and dedicated simulation studies using this detector geometry , are the subject of ongoing work and as such will not be described here , other than to highlight the consistency with expectation . this effect was observed with images reconstructed from dedicated simulation studies of this test configuration of objects @xcite which also provided excellent agreement with the image shown in figure [ fig : image ] . the design , fabrication and assembly processes undertaken to develop a modular tracker system for use in the cosmic - ray muon tomography of legacy nuclear waste containers has been presented . the system comprised four modules with orthogonal layers of 2mm - pitch round scintillating fibres . these were supported within a low-@xmath0 structure fabricated using thin sheets of rohacell^^ and aluminium , and covered with a lightproof tedlar^^ film and black nylon tubing . the fibre signals were read out to qdc modules and hamamatsu h8500 mapmts with two fibres coupled to a single pixel , and one pmt ( _ i.e._128 fibres ) per detection layer . the modules were supported in an adjustable vertical support stand made from aluminium profile . performance studies revealed a high level of stability , both structurally and in relation to the data collected over prolonged periods of time . high muon detection efficiencies of up to 86% per layer were recorded with low levels ( less than 5% ) of misidentification and misalignment ( less than 5 mm in the worst case prior to correction in software ) . first images reconstructed from data collected with a test configuration of materials within the assay volume verified the high-@xmath0 detection capabilities of this system and revealed promising low , medium and high-@xmath0 discrimination which will be investigated further in future work . all these studies will directly influence the design and construction of a large - scale detector system which will assay an industrial waste container in preparation for the industrial deployment of this technology . the authors gratefully acknowledge sellafield ltd . , on behalf of the uk nuclear decommissioning authority , for their continued funding of this project . 00 e. p. george , commonwealth engineer , 455 ( 1955 ) l. w. alvarez , science * 167 * , 832 ( 1970 ) h. k. m . _ , nuclear instruments & methods in physics research a * 555 * , 164 ( 2005 ) g. ambrosi _ , nuclear instruments & methods in physics research a * 628 * , 120 ( 2011 ) k. borozdin _ et al . _ , nature * 422 * , 277 ( 2003 ) k. gnanvo _ et al . _ , ieee nuclear science symposium conference record , 1278 ( 2008 ) l. schultz _ _ , nuclear instruments & methods in physics research a * 519 * , 687 ( 2004 ) s. agostinelli _ et al . _ , nuclear instruments & methods in physics research a * 506 * , 250 ( 2003 ) a. clarkson _ et al . _ , arxiv:1309.3400 ( 2013 ) r. montgomery _ _ , nuclear instruments & methods in physics research a * 695 * , 326 ( 2012 ) l. schultz _ et al . _ , ieee transactions on image processing * 16 * , 1985 ( 2007 ) d. f. mahon _ et al . _ , nuclear instruments & methods in physics research a ( 2013 ) , http://dx.doi.org/10.1016/j.nima.2013.05.119	tomographic imaging techniques using the coulomb scattering of cosmic - ray muons are increasingly being exploited for the non - destructive assay of shielded containers in a wide range of applications . one such application is the characterisation of legacy nuclear waste materials stored within industrial containers . the design , assembly and performance of a prototype muon tomography system developed for this purpose are detailed in this work . this muon tracker comprises four detection modules , each containing orthogonal layers of saint - gobain bcf-10 2mm - pitch plastic scintillating fibres . identification of the two struck fibres per module allows the reconstruction of a space point , and subsequently , the incoming and coulomb - scattered muon trajectories . these allow the container content , with respect to the atomic number @xmath0 of the scattering material , to be determined through reconstruction of the scattering location and magnitude . on each detection layer , the light emitted by the fibre is detected by a single hamamatsu h8500 mapmt with two fibres coupled to each pixel via dedicated pairing schemes developed to ensure the identification of the struck fibre . the pmt signals are read out to qdcs and interpreted via custom data acquisition and analysis software . the design and assembly of the detector system are detailed and presented alongside results from performance studies with data collected after construction . these results reveal high stability during extended collection periods with detection efficiencies in the region of 80% per layer . minor misalignments of millimetre order have been identified and corrected in software . a first image reconstructed from a test configuration of materials has been obtained using software based on the maximum likelihood expectation maximisation algorithm . the results highlight the high spatial resolution provided by the detector system . clear discrimination between the low , medium and high-@xmath0 materials assayed is also observed . muon tomography , scintillator detectors , nuclear waste 96.50.s- , 29.40.mc , 89.20.bb
searching for possible candidates of a quantum spin liquid state has been one of the central issues in condensed matter physics . organic materials provide an important research area @xcite and a number of organic insulators , especially @xmath3-(bedt - ttf)@xmath2cu@xmath2(cn)@xmath0 ( ref . ) [ where bedt - ttf is bis(ethylenedithio)-tetrathiafulvalene ] and etme@xmath0sb[pd(dmit)@xmath1@xmath2 ( ref . ) ( where dmit is 1,3-dithiole-2-thione-4,5-dithiolate ) are strong candidates of the realizations of the spin liquids . although the nearest neighbor antiferromagnetic couplings are @xmath15@xmath9220 - 250 k for both compounds as speculated from the high - temperature magnetic susceptibility data , @xcite they do not exhibit long - range magnetic order at least down to the temperature @xmath16@xmath95 k. the mechanism of survival of the spin degree of freedom at low temperatures is yet to be clarified . strong electronic correlation and the resulting large quantum fluctuation are proposed to realize such a ground state , @xcite especially for low - dimensional systems , but it is not so simple to establish and identify the real materials as spin liquids . the low - energy structure of these organic compounds is often described in terms of a single - band model crossing the fermi level . this corresponds to the dimerization limit of the anti - bonding band of the highest occupied molecular orbital ( homo ) of organic molecules forming the dimer . in this limit , the system is described by a triangular - lattice structure and the geometrical frustration in the lattice is considered as a possible origin of the quantum fluctuation . on the other hand , however , an appreciable anisotropy of the transfer structure in the triangular lattice has been reported by several _ ab initio _ density functional calculations . @xcite in addition , a recent dielectric measurement @xcite suggests that the system exhibits a relaxer - type response at low temperatures , @xmath16@xmath960 k , thus indicating that the charge degree of freedom might bring about some kinds of disorder in the low - energy electronic state . this is clearly beyond single - band physics . recently , the ground - state properties of etme@xmath0sb[pd(dmit)@xmath1@xmath2 are of great interest , for which the extremely low - temperature measurements have been performed . @xmath17c nmr spectra under 7.65 t do not show significant broadening down to 19.4 mk , @xcite while the spin - lattice relaxation rate @xmath18 exhibits a sharp drop below 1 k with a temperature dependence @xmath19 . it might suggest a continuous phase transition around 1k due to some unknown symmetry breaking accompanied by the existence of gapless excitations in the low - temperature phase . on the other hand , the temperature dependence of heat - capacity@xcite and thermal - conductivity@xcite data give a @xmath16-linear term down to 0.1k , in contradiction with the @xmath20 behavior expected from @xmath18 . furthermore , the magnetic torque measurement indicates a pauli paramagnetic - like uniform magnetic susceptibility down to 30 mk , @xcite indicating a feature similar to the one - dimensional mott insulator represented by heisenberg or hubbard models at half filling at low temperatures . such seemingly contradicting experimental data require further careful theoretical analyses on the origin and mechanism of the spin liquid behavior . the low - energy electronic structure of etme@xmath0sb[pd(dmit)@xmath1@xmath2 is known to be rather unique . in this material , the level difference between homo and lowest unoccupied molecular orbital ( lumo ) is small compared to the dimerization gap corresponding to the bonding - antibonding splitting of each orbital @xcite and the resulting band structure shows unusual level inversion of a pair of the homo - lumo bands . @xcite then from the lower energy , the homo - bonding ( homo - b ) , lumo - bonding ( lumo - b ) , homo - antibonding ( homo - ab ) , and lumo - antibonding ( lumo - ab ) are stacked separated by a finite gap . the homo - ab band is half filled and is sandwiched by the lumo - ab and lumo - b bands residing above and below 0.5 ev of the fermi level , respectively [ see fig.[fig1](a ) ] . such a unique band structure may easily provide potential relevance of multiband physics , for instance , charge fluctuations within a dimer corresponding to the polarization between the crossed homo - lumo level . in addition , a valence - bond - solid phase exists in similar compounds , @xcite where superconductivity appears in the vicinity . on the other hand , the geometrical frustration in the half - filled homo - ab band , @xcite as well as interactions at a dimer site and/or between dimer sites , can also be relevant to the survival of the spin - liquid phase . to clarify roles of various factors in the low - energy physics , _ ab initio _ derivation of the low - energy model from first principles is highly desired . for _ ab initio _ derivations for the low - energy model , a scheme based on the constrained random - phase approximation ( crpa ) ( refs . and ) for constructed maximally localized wannier orbitals ( mlwo ) ( ref . ) has been applied to a wide range of materials . @xcite furthermore , by utilizing a quasi - low - dimensional character of materials as organic conductors , the methodology for obtaining _ ab initio _ effective models in reduced dimensions has been developed @xcite and applied to a two - dimensional ( 2d ) effective - model derivation for @xmath3-(bedt - ttf)@xmath2cu(ncs)@xmath2 . @xcite in ref . , a single - band model for the homo - ab band crossing the fermi level was derived and analyzed by a multi - variable variational monte - carlo method to demonstrate a quantitative reliability of the derived model parameters through the study for the metal - insulator transition . it was found that , while this material is experimentally located in the metallic region with the superconductivity at low temperatures close to the border of the metal - insulator transition , @xcite the derived _ ab initio _ model and its monte - carlo solution predicted an antiferromagnetic insulator . with the 20% reduction of the interaction parameters , the _ ab initio _ model just gave a metallic solution . as a possible origin of the discrepancy between the theory and experiment , the multiband nature missing in the model construction was discussed , especially in terms of dynamical effects between the target band and the closest high - energy band ( in this material , the homo - b band ) . in fact , in the study on the onsite hubbard model for the @xmath3-bedt - ttf system , @xcite with the fluctuation exchange approximation , the symmetry of the superconducting gap function seems to be sensitive depending on whether the homo - b band is included in the degree of the freedom of the effective model or not . in the present paper , we derive 2d multiband models of etme@xmath0sb[pd(dmit)@xmath1@xmath2 and @xmath3-(bedt - ttf)@xmath2cu(ncs)@xmath2 from first principles . the single - band models are also derived and compared with the multiband models to understand the essence of the multiband nature . we found that , for etme@xmath0sb[pd(dmit)@xmath1@xmath2 , onsite intraorbital @xmath7 and interorbital @xmath21 interactions are both @xmath90.7 ev and the hund s rule coupling @xmath15 is 0.1 - 0.2 ev . an offsite interaction @xmath8 is @xmath90.25 ev and a transfer @xmath10 is 40 - 50 mev . as a result , our estimated correlation strength ( @xmath7@xmath11@xmath8)/@xmath10 gives a substantially large value as @xmath910 . the other compound @xmath3-(bedt - ttf)@xmath2cu(ncs)@xmath2 also exhibits large correlation strength as @xmath910 . we also estimate the reliability of the downfolding whether the multiband model can be reduced to the single - band model with the _ ab initio _ crpa framework . the measure of the reliability is given by an order estimate of the vertex correction to the effective interaction . we found that , for both etme@xmath0sb[pd(dmit)@xmath1@xmath2 and @xmath3-(bedt - ttf)@xmath2cu(ncs)@xmath2 , the vertex correction is not small and may well affect the low - energy physics , thus suggesting that the multiband analysis beyond the single - band one is needed for describing the low - energy physics properly . the present paper is organized as follows : in sec . ii , we define effective models to be derived and describe a scheme for the derivation . computational details and results for etme@xmath0sb[pd(dmit)@xmath1@xmath2 and @xmath3-(bedt - ttf)@xmath2cu(ncs)@xmath2 are given in sec . we discuss characteristic aspects of the derived effective models including correlation strength and the estimate of the reliability of the downfolding in sec . summary is given in sec.v . here , we describe basic procedures for the derivation of effective low - energy models for the two compounds . the basis of the hamiltonian is the wannier function associated with anti - bonding / bonding states of homo or lumo of dmit or bedt - ttf molecules that form a dimer . the derived hamiltonian is explicitly given in a form of the 2d multiband extended hubbard model as @xmath22 with @xmath23 ( @xmath24 ) being a creation ( annihilation ) operator of an electron with spin @xmath25 in the @xmath26th wannier orbital localized at the @xmath27th dimer site . the @xmath28 and @xmath29 parameters are given by @xmath30 and @xmath31 respectively , with @xmath32=@xmath33 and @xmath34 being an effective one - body kohn - sham hamiltonian . in this model for etme@xmath0sb[pd(dmit)@xmath1@xmath2 , we take three wannier orbitals , i.e. , those of lumo - b , homo - ab , and lumo - ab , to represent the hamiltonian . in @xmath3-(bedt - ttf)@xmath2cu(ncs)@xmath2 , we take two wannier orbitals ; homo - ab and homo - b . note that the wannier orbitals are made for each band individually , since the bands separate with each other . with this construction of the wannier function , the transfers for different orbitals are zero ( i.e. , @xmath35=0 for @xmath26@xmath36@xmath37 ) . the @xmath38 and @xmath39 parameters in eq . ( [ h_hub3 ] ) are screened coulomb and exchange integrals in the wannier - orbital basis , respectively , expressed as @xmath40 and @xmath41 with @xmath42 being a 2d screened coulomb interaction in the low - frequency limit . here , @xmath43 is nothing but the onsite hubbard @xmath7 , while @xmath44 is set to zero by definition . if one keeps only the homo - ab band crossing the fermi level and treats other bands as screening bands , the multi - band model is reduced to a single - band model @xmath45 where we drop the orbital index @xmath26 . the calculation for @xmath42 in eqs . ( [ v_ij ] ) and ( [ j_ij ] ) follows ref . , where a new crpa framework was developed for the purpose to derive effective interactions of models defined in lower spatial dimensions . this new scheme is suitable for quasi - low - dimensional materials as the present system . the crpa method is originally formulated in the rpa framework with the constraint for the band degree of freedom to eliminate only the degrees of freedom far from the fermi level in energy . this is called the band downfolding . in the proposed scheme of the supplementary downfolding , however , the concept of the constraint is additionally relaxed to include the screening by the polarization in the other layers / chains even within the target bands . this utilizes the real - space representation of the polarization function and elimination of specific - polarization blocks associated with the spatial dimensions perpendicular to the target layer / chain , which leads to a low - dimensional effective interaction for the target layer / chain . we call it the dimensional downfolding . practically , the band+dimensional downfolding is performed in two steps : we first perform the band downfolding to derive a three - dimensional model for a small number of bands near the fermi level . @xcite this is followed by the dimensional downfolding in the second step . @xcite with this idea , we can naturally derive the low - energy model in any dimension . in the present case , we use it for the derivation of multiband and single - band 2d models for etme@xmath0sb[pd(dmit)@xmath1@xmath2 and @xmath3-(bedt - ttf)@xmath2cu(ncs)@xmath2 . _ ab initio _ density - functional calculations were performed by the program package xtapp , which is a massively parallelized version of _ tokyo ab initio program package _ ( tapp ) ( ref . ) and is able to perform the present large scale calculations efficiently . the xtapp program adopts plane - wave basis sets and the present calculations were performed with norm - conserving pseudopotentials @xcite and generalized gradient approximation ( gga ) ( ref . ) for the exchange - correlation potential . the experimental structure of etme@xmath0sb[pd(dmit)@xmath1@xmath2 at 4 k was taken from x - ray crystallography , @xcite while the atomic structure of @xmath3-(bedt - ttf)@xmath2cu(ncs)@xmath2 at 15 k was taken from neutron data . @xcite for both of the systems , the positions of the hydrogen atoms were relaxed . the cutoff energies in wavefunctions and charge densities were set to 36 ry and 256 ry , respectively . 5@xmath465@xmath463 and 5@xmath465@xmath465 @xmath47-point samplings were employed for etme@xmath0sb[pd(dmit)@xmath1@xmath2 and @xmath3-(bedt - ttf)@xmath2cu(ncs)@xmath2 , respectively . the construction of mlwo follows ref . . the polarization function was expanded in plane waves with an energy cutoff of 5 ry and the total number of bands considered in the polarization calculation was set to 2000 for etme@xmath0sb[pd(dmit)@xmath1@xmath2 and 750 for @xmath3-(bedt - ttf)@xmath2cu(ncs)@xmath2 . this condition corresponds to considering the excitation up to @xmath920 ev above the fermi level . the brillouin - zone ( bz ) integral on the wavevector in the polarization calculation was evaluated by the generalized tetrahedron method . @xcite in the dimensional downfolding procedure , the calculated 2d effective coulomb / exchange integrals were extrapolated to values with the infinite number of screening layers . @xcite figure [ fig1 ] ( a ) shows our calculated gga band structures ( solid lines ) of etme@xmath0sb[pd(dmit)@xmath1@xmath2 . the band in [ @xmath110.1 ev : 0.3 ev ] is the homo - ab band which is sandwiched by the lumo - b and lumo - ab bands at @[email protected] ev . the dotted lines represent the tight - binding band obtained with the transfer parameters in table [ param - dmit ] , where the definition of each transfer is given in the panel ( b ) . the transfer parameters were estimated from the mlwos . in the panels ( c ) and ( d ) , we display our calculated mlwos for the homo - ab and lumo - b bands , respectively . as the initial guess for the homo - ab wannier orbital , we used the @xmath49-type gaussian with the width 1.67 @xmath50 , centered at the dimer center . also , the initial guess for the lumo - b and lumo - ab mlwos are superposition of the @xmath49-type gaussians with width 1.18 @xmath50 put on the s atoms adjacent to pd in each dmit molecule . of etme@xmath0sb[pd(dmit)@xmath1@xmath2 . the crystal structure consists of alternating layers ( parallel to the @xmath51 plane ) of the pd(dmit)@xmath2 anion and polymeric etme@xmath52sb@xmath53 cation . band dispersions are plotted along the high - symmetry points in the @xmath51 plane , where @xmath54=(0 , 0 , 0 ) , x=(@xmath55/2 , 0 , 0 ) , m=(@xmath55/2 , @xmath56/2 , 0 ) , y=(0 , @xmath56/2 , 0 ) . the zero of energy is the fermi level . the homo - ab band is drawn by the dark blue dotted lines and the lumo - b and lumo - ab bands are described by light green dotted lines . ( b ) triangular lattice and the definition of transfers . the system contains two equivalent pd(dmit)@xmath2 layers and the dimer units stack along the @xmath57+@xmath58 direction in one layer and the @xmath57@xmath11@xmath58 direction in the other layer . in this figure , the latter - layer configuration is depicted . the @xmath5 direction is defined as the stacking direction and the @xmath4 direction is defined to be parallel to the @xmath58 axis . the @xmath59 direction is given as the direction of the remaining side of the triangular lattice . ( c ) our calculated wannier function of the homo - ab state in view along the @xmath58 axis , drawn by vesta . @xcite the dark blue surfaces indicate positive isosurface and the light yellow surfaces indicate negative isosurface . ( d ) the same plot for the lumo - b state.,scaledwidth=50.0% ] lr@ r@ r@ r & & & & + + [ -8pt ] lumo - b & @xmath11490 & 0 & 8.4 & @xmath1116.1 + homo - ab & 34.8 & 54.4 & 44.9 & 40.2 + lumo - ab & 495 & @xmath1125.7 & 24.8 & 15.1 + homo - ab exhckel & - & 28.2 & 27.0 & 25.0 + [ param - dmit ] the transfer parameters for the homo - ab band in the extended hckel method are given in the bottom of the table . @xcite note that , the notation for the transfers in ref . is different from the present paper ; in the correspondence between the former and latter , @xmath60=@xmath5 @xmath61=@xmath4 , and @xmath62=@xmath59 . in comparison with the _ ab initio _ values , the hckel values are smaller than the _ ab initio _ ones by the factor of 1/2 - 2/3 . on top of that , the _ ab initio _ parameters exhibit an appreciable one - dimensional anisotropy along the @xmath5 direction , i.e. , @xmath5@xmath63@xmath4@xmath9@xmath59 . in contrast , the hckel parameters is close to the equilateral triangular lattice as @xmath5@xmath9@xmath4@xmath9@xmath59 . we next show in fig . [ fig2 ] our calculated interaction parameters for the 2d single - band model . the panels ( a ) and ( b ) give the diagonal and off - diagonal parts of the interaction , respectively . in the presentation , we define a reference site as a site indicated by the arrow . the value in the ellipsoid represents the interaction value between the reference site and this ellipsoid one , in the unit of ev . for example , in the panel ( a ) , 0.61 ( ev ) in the reference site is the onsite hubbard @xmath7 and 0.22 ( ev ) next on the right is the offsite coulomb integral between the reference site and this site . in the figure , we show the interaction value up to the tenth neighbors and the further distant interactions are negligibly small . note that the system has a c@xmath2 symmetry . the offsite exchange integral in the panel ( b ) is given in the unit of mev . in principle , the interaction range is short as up to the nearest shell and the values are so small as nearly half of those of @xmath3-(bedt - ttf)@xmath2cu(ncs)@xmath2 . @xcite sb[pd(dmit)@xmath1@xmath2 in eq . ( [ h_hub ] ) . the arrow indicates a reference site . the view of the triangular lattice follows fig . [ fig1](b).,scaledwidth=50.0% ] the interaction values of the 2d three - band model are displayed in fig . [ fig3 ] . in the multi - band case , the interaction is given as the 3@xmath463 matrix as @xmath64 v_{01j1 } & v_{01j2 } & v_{01j3 } \\ v_{02j1 } & v_{02j2 } & v_{02j3 } \\ v_{03j1 } & v_{03j2 } & v_{03j3 } \\ \end{array } \right ) { \rm and } \left ( \begin{array}{ccc } \phantom { } & \phantom{vvvv } & \phantom { } \\[-4 mm ] j_{01j1 } & j_{01j2 } & j_{01j3 } \\ j_{02j1 } & j_{02j2 } & j_{02j3 } \\ j_{03j1 } & j_{03j2 } & j_{03j3 } \\ \end{array } \right ) , \end{aligned}\ ] ] where @xmath65 ( @xmath66 ) indicates the coulomb ( exchange ) integral between the @xmath26th orbital in the @xmath67th reference site and the @xmath37th orbital in the @xmath68th site . in the orbital labels , @xmath69 ( lumo - b , homo - ab , lumo - ab ) . the onsite coulomb interactions are nearly @xmath90.7 ev and hold @xmath7@xmath9@xmath21 . on the offsite interactions , the orbital dependence is negligible , thus being well described as the function of the distance between the centers of the wannier functions . on the basis of this aspect , for the interactions beyond the first shell , we show the averaged value over the orbitals . the hund s rule coupling at the reference site is rather large ( e.g. , the exchange integral between the lumo - b and lumo - ab orbitals as @xmath90.22 ev ) . the exchange integrals with the neighboring site are rather orbital dependent , and there exist rather large values more than 10 mev . sb[pd(dmit)@xmath1@xmath2 in eq . ( [ h_hub3 ] ) . the arrow indicates a reference site . the view of the trianglar lattice follows fig . [ fig1](b).,scaledwidth=50.0% ] we next move to @xmath3-(bedt - ttf)@xmath2cu(ncs)@xmath2 . figure [ fig4-bedt ] ( a ) displays our calculated gga band ( solid lines ) and the tight - binding band ( dotted lines ) with the transfer integrals in table [ param - bedt ] . the definition of each transfer follows the panel ( b ) and the wannier function of the homo - b and homo - ab are plotted in the panels ( c ) and ( d ) , respectively . the extended hckel transfers for the homo - ab band @xcite is given in the bottom of table [ param - bedt ] . in @xmath3-(bedt - ttf)@xmath2cu(ncs)@xmath2 , @xmath5@xmath9@xmath4 holds and the lattice is approximated as an isosceles triangle with @xmath10=@xmath70 and @xmath71=@xmath59 . for both the _ ab initio _ and extended hckel results , this trend well holds , but the former frustration strength @xmath72 as @xmath90.66 is somewhat smaller than the latter one as @xmath90.84 . we also note that , in the _ ab initio _ results , the @xmath73 transfer has a finite magnitude . of @xmath3-(bedt - ttf)@xmath2cu(ncs)@xmath2 . the crystal structure contains alternating layers ( parallel to the @xmath74 plane ) of bedt - ttf donor molecules and polymeric cu(ncs)@xmath75 anions . band dispersions are plotted along the high - symmetry points in the @xmath74 plane , where @xmath54=(0 , 0 , 0 ) , y=(0 , @xmath56/2 , 0 ) , z=(0 , 0 , @xmath76/2 ) , and m=(0 , @xmath56/2 , @xmath76/2 ) . note that the @xmath57 axis is interlayer axis . the zero of energy is the fermi level . the homo - ab band is drawn by the dark blue dotted lines and the homo - b band is described by light green dotted lines . ( b ) definition of the transfers . the directions of the @xmath58 and @xmath77 axes forming the plane are also drawn . ( c ) our calculated wannier function of the homo - b state in view along the @xmath57 axis , drawn by vesta . @xcite the dark blue surfaces indicate positive isosurface and the light yellow surfaces indicate negative isosurface . ( d ) the same plot for the homo - ab state.,scaledwidth=45.0% ] lr@ r@ r@ r@ r & & & & & + + [ -8pt ] homo - ab & 0.02 & @xmath1164.8 & @xmath1169.3 & 44.2 & @xmath1111.5 + homo - b & @xmath11447 & @xmath1140.3 & @xmath1125.9 & @xmath1145.8 & 12.1 + homo - ab exhckel & - & @xmath1169.3 & @xmath1165.8 & 56.6 & - + [ param - bedt ] the interaction parameter for the 2d single - band model is given in fig . [ fig5-bedt ] , where the panels ( a ) and ( b ) are diagonal and off - diagonal parts for the interaction , respectively . how to see the figure is the same as fig . [ fig2 ] . the interaction range for @xmath3-(bedt - ttf)@xmath2cu(ncs)@xmath2 is shorter than that of etme@xmath0sb[pd(dmit)@xmath1@xmath2 as far as we compare the distance with the values up to @xmath90.05 ev . this is due to the fact that the interdimer distance of @xmath3-(bedt - ttf)@xmath2cu(ncs)@xmath2 is larger than that of etme@xmath0sb[pd(dmit)@xmath1@xmath2 [ 7.68 @xmath50 for the @xmath3-bedt - ttf salt and 6.31 @xmath50 for the pd(dmit)@xmath2 salt ] . furthermore , in general , the in - plane interaction range is roughly determined by the interlayer distance , @xcite because , in the dimensional downfolding , the interlayer metallic screening is switched on from this distance [ 16.44 @xmath50 for the @xmath3-bedt - ttf salt and 18.52 @xmath50 for the pd(dmit)@xmath2 salt ] and , within this range , the in - plane interaction gives finite values . we also mention that the exchange value of @xmath3-(bedt - ttf)@xmath2cu(ncs)@xmath2 is nearly twice as large as that of etme@xmath0sb[pd(dmit)@xmath1@xmath2 , which may reflect the difference in the dimer orientation of @xmath3-(bedt - ttf)@xmath2cu(ncs)@xmath2 and etme@xmath0sb[pd(dmit)@xmath1@xmath2 [ compare fig . [ fig1](b ) and fig . [ fig4-bedt](b ) ] . -(bedt - ttf)@xmath2cu(ncs)@xmath2 in eq . ( [ h_hub ] ) . the arrow indicates a reference site . the view of the triangular lattice follows fig . [ fig4-bedt](b).,scaledwidth=45.0% ] we next show in fig . [ fig6-bedt ] our calculated 2d interaction parameters for the two - band model . the interactions are given by the matrix form as @xmath78 v_{01j1 } & v_{01j2 } \\ v_{02j1 } & v_{02j2 } \\ \end{array } \right ) { \rm and } \left ( \begin{array}{ccc } \phantom { } & \phantom{vv } \\[-4 mm ] j_{01j1 } & j_{01j2 } \\ j_{02j1 } & j_{02j2 } \\ \end{array } \right).\end{aligned}\ ] ] in the orbital labels , @xmath79 ( homo - ab , homo - b ) . the notations are the same as fig . [ fig3 ] , but note that the symmetry of the interaction matrix is different from that for etme@xmath0sb[pd(dmit)@xmath1@xmath2 . the onsite coulomb interaction is @xmath90.8 ev and the hund s rule coupling is @xmath90.36 ev , being larger than those of etme@xmath0sb[pd(dmit)@xmath1@xmath2 . the offsite coulomb integrals are not orbital dependent , while the offsite exchange integrals are largely orbital dependent and the largest value is as large as @xmath930 mev . -(bedt - ttf)@xmath2cu(ncs)@xmath2 in eq . ( [ h_hub3 ] ) . the arrow indicates a reference site . the view of the triangular lattice follows fig . [ fig4-bedt](b).,scaledwidth=50.0% ] here , we first remark that the antiferromagnetic exchange coupling @xmath15 is estimated in the strong - coupling perturbation as @xmath15 @xmath9 @xmath80/@xmath81@xmath11@xmath82 = 245 k for the pd(dmit)@xmath2 salt and @xmath15 @xmath9 360 k for @xmath3-bedt - ttf salt , when we use @xmath83 to compare with the experimental data deduced from the fitting to the isotropic triangular heisenberg model in the high - temperature expansion . @xcite when we consider the insufficiency of the single - band model and the experimental uncertainty coming from the lack of the data above 300 k , the results are favorably compared with 250 k estimated for the both compounds from such fittings of the magnetic susceptibility . now we discuss the screening effects due to the band downfolding from the multiband to single - band models and due to the dimensional downfolding as well . in fig . [ fig7 ] , we draw a schematic diagram showing how the onsite interaction value of the homo - ab band of etme@xmath0sb[pd(dmit)@xmath1@xmath2 is screened . the bare coulomb interaction is 3.55 ev and in the stage of the 3d three - band model with the conventional band downfolding , the interaction value becomes 0.89 ev . with considering the dimensional downfolding , the value is reduced further to 0.69 ev by 0.2 ev . when we consider the screening due to the lumo - b and lumo - ab bands in addition , to construct the single - band model , the value becomes 0.61 ev ( the reduction of 0.08 ev ) . so , the largest part of the screening comes from the high - energy - electron screening considered in the initial band - downfolding stage , while the screening effects due to the dimensional downfolding , as well as the reduction to a single - band model may not be neglected quantitatively . we note that the fully - screened - rpa value is 0.14 ev . sb[pd(dmit)@xmath1@xmath2 , generated by the band+dimensional downfolding.,scaledwidth=40.0% ] the same diagram for @xmath3-(bedt - ttf)@xmath2cu(ncs)@xmath2 is shown in fig . [ fig8 ] . in this material , while the reduction of the value due to the dimensional downfolding is nearly the same as the etme@xmath0sb[pd(dmit)@xmath1@xmath2 case , the reduction due to the band downfolding from the two - band to single - band model is appreciable ( the reduction of @xmath90.2 ev ) . this is probably due to the fact that the spatial distribution of the eliminated homo - b orbital is similar to that of the target homo - ab orbital and largely overlaps , resulting in a more efficient screening . the bare and fully - screened - rpa values are 3.61 and 0.19 ev , respectively . -(bedt - ttf)@xmath2cu(ncs)@xmath2 , generated by the band+dimension downfolding.,scaledwidth=40.0% ] we next discuss the degree of the correlation strength of the 2d effective models , which is measured by ( @xmath7@xmath11@xmath8)/@xmath84 , where @xmath7 and @xmath8 are the onsite and offsite coulomb interactions , respectively , and @xmath84 is the bandwidth . since we are interested in low energy , we focus on the homo - ab band . table [ param - woverd ] lists our calculated parameters and the correlation strength . for both etme@xmath0sb[pd(dmit)@xmath1@xmath2 and @xmath3-(bedt - ttf)@xmath2cu(ncs)@xmath2 , ( @xmath7@xmath11@xmath8)/@xmath84 is comparable to the unity and thus these systems are classified as the strongly correlated electron system . although the correlation strength of the single - band model is somewhat smaller than that of the multiband model , it is still large as close to 1 . lr@ r@ r@ r@ r@ r@ r & & & & & & & + + [ -8pt ] etme@xmath0sb[pd(dmit)@xmath1@xmath2 ( single - band ) & 0.61 & 0.20 & 0.44 & 0.93 & - & - & - + etme@xmath0sb[pd(dmit)@xmath1@xmath2 ( three - band ) & 0.69 & 0.23 & 0.44 & 1.04 & 0.15 & 0.44 & 0.33 + @xmath3-(bedt - ttf)@xmath2cu(ncs)@xmath2 ( single - band ) & 0.64 & 0.18 & 0.56 & 0.82 & - & - & - + @xmath3-(bedt - ttf)@xmath2cu(ncs)@xmath2 ( two - band ) & 0.83 & 0.19 & 0.56 & 1.14 & 0.16 & 0.33 & 0.49 + [ param - woverd ] we now detail the validity of reducing from the multiband to single - band model in the present case , by applying a general criterion . @xcite in crpa for deriving the effective interaction of the single - band model , the multiband is divided into the target homo - ab band and the rest . then , with excluding a polarization formed in the target band , the effective interaction is calculated . figure [ fig9 ] ( a ) exhibits a schematic diagram for deriving the single - band effective interaction illustrated symbolically in the second order one , where low - energy electrons in the homo - ab band ( external lines ) is screened by a polarization ( internal lines ) via high - energy electrons ( namely , electrons belonging to the bands away from the fermi level ) in the eliminated bands [ i.e. , the lumo - ab / lumo - b bands for the pd(dmit)@xmath2 salt and the homo - b band for the @xmath3-bedt - ttf salt ] . here , the capital characters l " and h " describe propagators of the low- and high - energy electrons , respectively . in the figure , an electron - hole pair formed in the low- and high - energy bands is displayed . the infinite series of such a diagram gives the effective interaction used in the single - band model . in this treatment , the vertex which modifies the interaction is totally dropped . in the panels ( b ) and ( c ) , we show the correction to the original interaction process in ( a ) ; the vertex in the shaded area corrects the polarization [ panel ( b ) ] or introduces an additional interaction process [ ( c ) ] . to the first approximation , the vertex is approximated by the single fully screened rpa interaction ; the correction to the polarization is written as an exchange process ( d ) and the additional interaction process for the propagators is written as a single ladder process ( e ) . let us consider the magnitude of these corrections to the original process . the polarization is given by @xmath85 with @xmath86 and @xmath87 being the propagators of the low- and high - energy electrons , respectively , and this magnitude roughly scales as @xmath88 , where @xmath14 is the minimum energy distance of the high - energy band from the fermi level . we note that , in this estimate , the effect of the transition matrix element is dropped for the simplicity . similarly , the correction term described in the panel ( d ) is given by @xmath87@xmath86@xmath87@xmath86@xmath89 with @xmath90 being the fully - screened interaction between the low- and high - energy electrons . @xcite thus , the order of the correction to the original polarization is scaled as @xmath90/@xmath14 . @xcite from the similar discussion , the order of the correction described in the panel ( e ) is also given as @xmath90/@xmath14 . this value of @xmath90/@xmath14 is a measure for a stable downfolding from a multiband to a single - band model . when this value is small , the vertex correction to the crpa interaction can be neglected . on the contrary , when the value is large , the crpa downfolding breaks down and one must solve the multiband model directly without reducing to the single - band model . we show in table [ param - woverd ] our estimated @xmath91 , where we have approximated @xmath90 by the full rpa form @xmath13 ; in etme@xmath0sb[pd(dmit)@xmath1@xmath2 , @xmath13 is calculated as the full - rpa coulomb interaction matrix element between the homo - ab and lumo - b wannier orbitals , and , in @xmath3-(bedt - ttf)@xmath2cu(ncs)@xmath2 , @xmath13 is estimated as the matrix element between the homo - ab and homo - b wannier orbitals . the resulting @xmath92 is as large as 0.3 - 0.5 . this value is rather large compared to the transition - metal oxides . in the transition - metal oxide , @xmath93 bands ( @xmath94 or @xmath95 bands ) form the low - energy band and the closest high - energy bands consist of oxygen-@xmath49 bands . in this case , the matrix element of the full - rpa interaction @xmath96 between @xmath94/@xmath95 and o@xmath97 orbitals is substantially smaller , because the centers of the wannier orbitals are spatially apart with each other and the full - rpa interaction does not work for such a length scale . on the other hand , in the case of the organic compounds , the wannier orbitals of the low- and high - energy bands have the same centers and therefore the matrix element for the full - rpa interaction does not disappear . this is a characteristic feature of the organic compound compared to the transition - metal oxide . as a result , in the case of the organic compound , the downfolding treatment based on crpa to reduce to the single - band model would be insufficient and the multiband analysis is strongly recommended . one of the compounds exhibiting quantum spin - liquid behavior is @xmath3-(bedt - ttf)@xmath2cu@xmath2(cn)@xmath0 . the _ ab initio _ parameters for the 3d single - band systems have been derived in ref .. since the homo bonding orbital appears to be entangled with other lower - energy bands , one may need to treat more than two bands for the multi - band description , in contrast to @xmath3-(bedt - ttf)@xmath2cu(ncs)@xmath2 , which is left for future studies . to summarize , in the present paper , we derived the 2d multiband models for etme@xmath0sb[pd(dmit)@xmath1@xmath2 and @xmath3-(bedt - ttf)@xmath2cu(ncs)@xmath2 , with the _ ab initio _ crpa plus mlwo framework suitable for strongly correlated electron systems . the correlation strength for the derived model , ( @xmath7@xmath11@xmath8)/@xmath84 , is estimated as a large value of @xmath91 , as shown in table [ param - woverd ] , indicating the strongly correlated nature of electrons in these systems . the transfer structure of etme@xmath0sb[pd(dmit)@xmath1@xmath2 is characterized by a lattice of quasi-2d scalene triangular lattice with 1d anisotropy , whereas that of @xmath3-(bedt - ttf)@xmath2cu(ncs)@xmath2 approximately has a 2d square - lattice structure with a weaker next - nearest neighbor transfer . to discuss whether the multiband model can be reduced to the single - band model with the usual crpa treatment , we estimated the order of the vertex correction to the crpa interaction . we found that the vertex process between the target homo - ab band and others [ i.e. , the lumo - ab / lumo - b bands for the pd(dmit)@xmath2 salt and the homo - b band for the @xmath3-bedt - ttf salt ] is not small . therefore , the multiband model should directly be analyzed , not by reducing to the single - band model , to clarify the realistic correspondence with strongly correlated phenomena including non - fermi liquid behavior , quantum spin liquid phase and/or unconventional superconducting mechanism found in the real organic compounds . this remains to be explored . we would like to thank reizo kato and takeo fukunaga for providing us with the structural data for etme@xmath0sb[pd(dmit)@xmath1@xmath2 . we also thank shiro sakai for fruitful discussions . calculations were done at supercomputer center at institute for solid state physics , university of tokyo . this work was supported by grants - in - aid for scientific research ( no . 22740215 , 22104010 , 23110708 , 23340095 , 22104006 , 19051016 ) from mext , japan . a part of this research has been funded by the strategic programs for innovative research ( spire ) , mext , and the computational materials science initiative ( cmsi ) , japan . k. nakamura , y. yoshimoto , t. kosugi , r. arita , and m. imada , j. phys . . jpn . * 78 * , 083710 ( 2009 ) . h. c. kandpal , i. opahle , y .- z . zhang , h. o. jeschke , and r. valenti , phys . lett . * 103 * , 067004 ( 2009 ) . e. p. scriven and b. j. powell , arxiv:1108.4456 . h. tajima , t. naito , m.tamura , a. kobayashi , h. kuroda , et al . , solid state commun . * 79 * , 337 ( 1991 ) e. canadell , n. j. chem . * 21 * , 1147 ( 1997 ) . t. miyazaki and t. ohno , phys . b * 59 * , r5269 ( 1999 ) . e. canadell , s. ravy , jp . pouget , l. brossard , solid state commun . * 75 * , 633 ( 1990 ) . m. tamura and r. kato , chem . . lett . * 387 * , 448 ( 2004 ) . a. nakao and r. kato , j. phys . jpn . * 74 * , 2754 ( 2005 ) . m. tamura , k. takenaka , h. takagi , s. sugai , a. tajima , r. kato , chem . * 411 * , 133 ( 2005 ) . y. shimizu , h. akimoto , h. tsujii , a. tajima , r. kato , phys . lett . * 99 * , 256403 ( 2007 ) . y. ishii , m. tamura , r. kato , j. phys . . jpn . * 76 * , 033704 ( 2007 ) . t. itou , a. oyamada , s. maegawa , k. kubo , h.m . yamamoto , r. kato , phys . b * 79 * , 174517 ( 2009 ) . f. aryasetiawan , m. imada , a. georges , g. kotliar , s. biermann , and a. i. lichtenstein , phys . b * 70 * , 195104 ( 2004 ) . i. v. solovyev and m. imada , phys . rev . b * 71 * , 045103 ( 2005 ) . n. marzari and d. vanderbilt , phys . b * 56 * , 12847 ( 1997 ) ; i. souza , n. marzari , and d. vanderbilt , phys . b * 65 * , 035109 ( 2001 ) . m. imada and t. miyake , j. phys . jpn . * 79 * , 112001 ( 2010 ) . k. nakamura , r. arita , and m. imada , j. phys . 77 * , 093711 ( 2008 ) . k. nakamura , y. yoshimoto , y. nohara , and m. imada , j. phys . . jpn . * 79 * , 123708 ( 2010 ) . h. shinaoka , t. misawa , k. nakamura , and m. imada , j. phys . . jpn . * 81 * , 034701 ( 2012 ) . h. urayama , h. yamochi , g. saito , k. nozawa , t. sugano , m. kinoshita , s. sato , k. oshima , a. kawamoto , and j. tanaka , chem . lett . , 55 ( 1988 ) ; s. gartner , e. gogu , i. heinen , h. j. keller , t. klutz , and d. schweitzer , solid state commun . * 65 * 1531 ( 1988 ) ; k. d. carlson , u. geiser a. m. kini , h. h. wang , l. k. montgomery , w. k. kwok , m. a. beno , j. m. williams , c. s. cariss , g. w. crabtree , m .- h . whangbo , and m. ezain , inorg . * 27 * , 965 ( 1988 ) . k. oshima , t. mori , and h. inokuchi , h. urayama , h. yamochi , and g. saito , phys . b * 38 * , 938(r ) ( 1988 ) . k. kuroki , t. kimura , r. arita , y. tanaka , and y. matsuda : phys . rev . b * 65 * , 100516(r ) ( 2002 ) .	we present _ ab initio _ two - dimensional extended hubbard - type multiband models for etme@xmath0sb[pd(dmit)@xmath1@xmath2 and @xmath3-(bedt - ttf)@xmath2cu(ncs)@xmath2 , after a downfolding scheme based on the constrained random phase approximation ( crpa ) and maximally - localized wannier orbitals , together with the dimensional downfolding . in the pd(dmit)@xmath2 salt , the antibonding state of the highest occupied molecular orbital ( homo ) and the bonding / antibonding states of the lowest unoccupied molecular orbital ( lumo ) are considered as the orbital degrees of freedom , while , in the @xmath3-bedt - ttf salt , the homo - antibonding / bonding states are considered . accordingly , a three - band model for the pd(dmit)@xmath2 salt and a two - band model for the @xmath3-(bedt - ttf ) salt are derived . we derive single band models for the homo - antibonding state for both of the compounds as well . the homo antibonding band of the pd(dmit)@xmath2 salt has a triangular structure of the transfers with a one - dimensional anisotropy in contrast to the nearly equilateral triangular structure predicted in the extended hckel results . the ratio of the larger interchain transfer @xmath4 to the intrachain transfer @xmath5 is around @xmath6 . our calculated screened onsite interaction @xmath7 and the largest offsite interaction @xmath8 are @xmath90.7 ev and @xmath90.23 ev , respectively , for etme@xmath0sb[pd(dmit)@xmath1@xmath2 and @xmath90.8 ev and @xmath90.2 ev for @xmath3-(bedt - ttf)@xmath2cu(ncs)@xmath2 . these values are large enough compared to transfers @xmath10 as @xmath955 mev for the pd(dmit)@xmath2 salt and @xmath965 mev for the @xmath3-bedt - ttf one , and the resulting large correlation strength ( @xmath7@xmath11@xmath8)/@xmath12 indicates that the present compounds are classified as the strongly correlated electron systems . in addition , the validity whether the present multiband model can be reduced to the single - band model for the homo - antibonding state , widely accepted in the literature , is discussed . for this purpose , we estimated the order of vertex corrections ignored in the crpa downfolding to the single band model , which is given by @xmath13/@xmath14 , where @xmath13 is a full - screened - interaction matrix element between the homo - antibonding and other bands away from the fermi level ( namely homo - bonding or lumo - bonding / antibonding bands ) , whereas @xmath14 is the energy distance between the fermi level and the bands away from the fermi level . in the present materials , @xmath13/@xmath14 estimated as 0.3 - 0.5 signals a substantial correction and thus the exchange process between the low - energy homo - antibonding and other bands away from the fermi level may play a key role to the low - energy ground state . this supports that the minimal models to describe the low - energy phenomena of the organic compounds are the multiband models and may not be reduced to the single - band model .
the aim of this paper is to introduce a generalization of several classical knot and link invariants including the linking numbers , the alexander - conway polynomial , and the murasugi - tristram - levine signatures of links in euclidean 3-space . these invariants are generalized to links in so - called quasi - cylindres . a quasi - cylinder over a commutative ring @xmath0 is an oriented 3-manifold @xmath1 endowed with a submodule @xmath2 of the @xmath0-module @xmath3 such that the inclusion homomorphism @xmath4 is an isomorphism . the main example is the cylinder @xmath5 $ ] where @xmath6 is an oriented surface and @xmath7 . ( here , @xmath8 ) . for homologically trivial links in a quasi - cylinder @xmath1 with @xmath9 , we define a generalized seifert form and derive from it several other invariants , namely , a generalized alexander - conway polynomial and generalized signatures . the most interesting feature of our invariants is the appearance of additional parameters which are absent in the classical case . the organization of the paper is as follows . in section [ section : lk ] we introduce generalized linking numbers of links in quasi - cylinders . in section [ section : forms ] we define the generalized seifert form for links in quasi - cylinders . in section [ section : alex ] we study the derived alexander invariants . in section [ section : genus ] we discuss simple estimates of the link genus . in section [ section : concordance ] we study the concordance of links . in section [ section : signature ] we consider the signatures . in section [ section : colored ] we introduce a multivariable extension of the theory . in section [ section : gene ] we discuss various generalizations of our invariants and in particular an extention to homologically non - trivial links . in this paper , all manifolds are smooth . the boundary of an oriented manifold is oriented _ via _ the outward normal vector first " convention . throughout the paper , we fix a commutative ring with unity @xmath0 . by a _ link _ in an oriented 3-manifold @xmath1 , we mean a finite system of disjoint oriented circles embedded in @xmath10 . each link @xmath11 in @xmath1 viewed as a geometric 1-cycle represents a homology class @xmath12\in h_1(m;r)$ ] . a link @xmath11 is _ @xmath0-homologically trivial _ if @xmath12=0 $ ] . for @xmath8 , we say simply _ homologically trivial_. two links @xmath11 and @xmath13 in @xmath1 are said to be _ ambient isotopic _ if there is an ambient isotopy @xmath14 of @xmath1 , keeping @xmath15 fixed , such that @xmath16 , @xmath17 , and @xmath18 is orientation - preserving . a _ knot _ is a link consisting of a single circle . let us stress that all knots and links in this paper are oriented . the classical linking number of disjoint @xmath0-homologically trivial knots @xmath19 in an oriented 3-manifold @xmath1 is defined by @xmath20 where @xmath21 is the standard homological intersection in @xmath1 and @xmath22 is a 2-chain in @xmath1 ( with coefficients in @xmath0 ) such that @xmath23 . the independence of the choice of @xmath24 follows from the fact that given another 2-chain @xmath25 with @xmath26 , one has @xmath27 where @xmath28 is a 2-cycle in @xmath1 . the @xmath0-homological triviality of @xmath29 implies that @xmath30\cdot b=0 $ ] . one easily checks the symmetry @xmath31 . we introduce a _ generalized linking number _ as follows . suppose that @xmath32 and denote by @xmath33 the inclusion homomorphism @xmath34 . fix a submodule @xmath2 of the @xmath0-module @xmath35 such that @xmath36 . for disjoint knots @xmath37 in @xmath1 such that @xmath38,[l ] \in c(v)$ ] , set @xmath39 where @xmath22 is any 2-chain in @xmath1 ( with coefficients in @xmath0 ) such that @xmath40 for a 1-cycle @xmath41 on @xmath15 representing an element of @xmath2 . the homological intersection @xmath42 does not depend on the choice of @xmath24 . indeed , consider another 2-chain @xmath25 in @xmath1 with @xmath43 where @xmath44 is a 1-cycle on @xmath45 representing an element of @xmath2 . then @xmath46 is a relative 2-cycle in @xmath47 in the sense that its boundary lies on @xmath15 . let @xmath48 be a 1-cycle on @xmath15 whose homology class @xmath49\in h_1(\partial m;r)$ ] satisfies @xmath50)=[k]$ ] and let @xmath51 be a 1-cycle obtained by pushing @xmath48 slightly inside @xmath52 . then @xmath53 where @xmath54 is the homological intersection of 1-cycles in @xmath15 . we have @xmath55 since @xmath56=[v - v']\in v\cap{\text{ker}}(c)=0 $ ] . the linking number @xmath57 satisfies @xmath58 where @xmath59 are 1-cycles on @xmath15 representing elements of @xmath2 homological to @xmath37 respectively . indeed , let @xmath51 be a 1-cycle in @xmath52 obtained from @xmath48 as above and let @xmath60 be a 2-chain in @xmath1 with @xmath61 . then @xmath62 is disjoint form @xmath41 and therefore @xmath63 it is clear that @xmath64 is invariant under deformations of @xmath29 and @xmath11 in @xmath1 keeping them disjoint . if @xmath19 are @xmath0-homologically trivial ( in particular , if they lie in a 3-ball inside @xmath1 ) , then @xmath65 . the definition of @xmath64 extends in the obvious way to the case where @xmath19 are disjoint 1-cycles in @xmath1 . by a _ quasi - cylinder _ ( over @xmath0 ) , we mean a pair consisting of a connected oriented 3-manifold @xmath1 with non - empty boundary and a submodule @xmath2 of the @xmath0-module @xmath66 such that the restriction of the inclusion homomorphism @xmath67 to @xmath2 yields an isomorphism @xmath68 . the inverse isomorphism is denoted @xmath69 . the constructions of the previous section show that for a quasi - cylinder @xmath70 over @xmath0 and any disjoint knots @xmath19 in @xmath1 , we have a well - defined linking number @xmath71 satisfying @xmath72 ) \cdot_{\partial m } d_v([l]).\ ] ] we say that a quasi - cylinder @xmath70 has trivial 2-homology if @xmath73 . here and below , the unspecified group of coefficients in homology / cohomology is @xmath74 . note the following lemma . [ lemma : homology ] let @xmath75 be a quasi - cylinder such that @xmath1 is compact . the equality @xmath73 holds if and only if @xmath15 is connected . the components of @xmath15 represent elements of @xmath76 subject to only one relation : their sum is equal to zero . therefore the equality @xmath73 implies that @xmath15 is connected . let us prove the converse . since the inclusion homomorphism @xmath77 is onto and the inclusion homomorphism @xmath78 is an isomorphism , the homology sequence of the pair @xmath47 gives that @xmath79 . observe that @xmath80 . therefore the group @xmath81 is finite and @xmath82 \1 . the pair consising of a 3-ball @xmath83 and @xmath84 is a quasi - cylinder over @xmath85 . clearly , @xmath86 is the usual linking number of knots in the 3-ball . let @xmath6 be a connected oriented surface and @xmath5 $ ] with product orientation . set @xmath87 . it is clear that @xmath70 is a quasi - cylinder . the linking number @xmath88 of knots @xmath89 can be computed as follows . present the link @xmath90 by a link diagram on @xmath6 . let @xmath91 be the components of the diagram representing @xmath29 and @xmath11 , respectively . then @xmath92 where @xmath93 ( resp . @xmath94 ) is the number of positive ( resp . negative ) crossing points on the diagram where @xmath95 passes under @xmath96 . the quasi - cylinder @xmath97 , v)$ ] has trivial 2-homology if and only if @xmath98 . let @xmath99 be an @xmath0-homology 3-sphere , i.e. , a closed oriented 3-manifold with @xmath100 . let @xmath101 be a non - empty finite graph in @xmath99 and @xmath102 be its exterior , that is the complement of an open regular neighborhood of @xmath101 in @xmath99 . let @xmath103 be the submodule generated by the homology classes of the meridians of the edges of @xmath101 . then the pair @xmath70 is a quasi - cylinder . for any knots @xmath89 , we have @xmath104 where the right - hand side is the linking number of @xmath19 in @xmath99 . the quasi - cylinder @xmath70 has trivial 2-homology if and only if @xmath101 is connected . the constructions above suggest that the definition of the milnor numbers of classical links may be extended to links in quasi - cylinders . we shall not pursue this line here . let @xmath70 be a quasi - cylinder over @xmath0 . consider an oriented embedded surface @xmath105 ( possibly , @xmath106 ) . for a 1-cycle @xmath107 on @xmath108 , denote by @xmath109 ( resp . @xmath110 ) the 1-cycle in @xmath111 obtained by pushing @xmath107 along the positive ( resp . negative ) normal direction on @xmath108 in @xmath1 . for 1-cycles @xmath112 on @xmath108 representing homology classes @xmath113 , [ b ] \in h_1(f;r)$ ] , set @xmath114,[b])=lk_v(a^+,b).\ ] ] this number only depends on the homology classes of @xmath107 and @xmath115 in @xmath116 . indeed , if @xmath117 and @xmath118 are @xmath119-cycles on @xmath108 such that @xmath120 and @xmath121 for some @xmath122-chains @xmath62 and @xmath24 in @xmath108 , then @xmath123 where @xmath124 denotes the @xmath122-chain @xmath62 pushed along the positive normal direction on @xmath108 in @xmath1 . since @xmath125 and @xmath126 , @xmath127 , these three intersection number are zero . hence , we have a well - defined bilinear form @xmath128 we call it the _ generalized seifert form _ of @xmath108 . [ lemma : relation ] let @xmath129 be the composition of the inclusion homomorphism @xmath130 with the isomorphism @xmath131 . for all @xmath132 , @xmath133 by abuse of notation , we shall denote 1-cycles representing @xmath134 by the same symbols @xmath134 . consider the bilinear form @xmath135 defined as @xmath136 but using @xmath110 instead of @xmath109 . we claim that @xmath137 indeed , let @xmath24 be a @xmath122-cycle in @xmath1 such that @xmath138 , and let @xmath139 be the @xmath122-cycle @xmath140\times a$ ] in @xmath52 with @xmath141 . we have @xmath142 now , @xmath139 and @xmath143 are disjoint , so @xmath144 . since @xmath145 , this gives ( 2.a ) . we now verify that @xmath146 indeed , @xmath147 combining formulas ( 2.a ) and ( 2.b ) , we obtain the claim of the lemma . let @xmath148 be an arbitrary @xmath0-module . seifert triple _ over @xmath148 is a triple @xmath149 , where @xmath150 is a free @xmath0-module of finite rank , @xmath151 a bilinear form @xmath152 , and @xmath153 a homomorphism @xmath154 . two seifert triples @xmath155 , @xmath156 over @xmath148 are isomorphic if there is an @xmath0-isomorphism @xmath157 such that @xmath158 and @xmath159 . a seifert triple @xmath160 is obtained from a seifert triple @xmath149 by an _ elementary enlargement _ ( and @xmath149 from @xmath160 by an _ elementary reduction _ ) if the following conditions hold : @xmath161 , @xmath162 , @xmath163 , @xmath164 , @xmath165 and either @xmath166 or @xmath167 . if @xmath168 is a basis of @xmath150 , then @xmath169 is a basis of @xmath170 and the matrix @xmath171 of @xmath172 with respect to @xmath169 is computed from the matrix @xmath173 of @xmath151 with respect to @xmath168 by @xmath174 we say that two seifert triples over @xmath148 are _ equivalent _ if they can be related by a finite sequence of isomorphisms and elementary enlargements and reductions . given a quasi - cylinder @xmath75 over @xmath0 and a compact connected oriented surface @xmath175 , the constructions above yield a seifert triple @xmath176 over @xmath2 . note that the @xmath0-module @xmath116 is free of rank @xmath177 , where @xmath178 is the genus of @xmath108 and @xmath179 is the number of connected components of @xmath180 . suppose that a surface @xmath181 is obtained from @xmath108 by surgery along an embedded arc in @xmath52 meeting @xmath108 exactly at its endpoints and approaching @xmath108 either from the positive side or from the negative side at both endpoints . the transformation @xmath182 and the inverse transformation are called _ surgeries_. it is easy to see that the seifert triple of @xmath183 is obtained from the seifert triple of @xmath108 by an elementary enlargement . therefore the equivalence class of the seifert triple of an embedded surface is invariant under surgeries . observe that any given class in @xmath76 can be realized by a closed connected oriented surface and any two such surfaces are related by a sequence of surgeries and isotopies . this leads to algebraic invariants of integral 2-homology classes of quasi - cylinders . we shall however focus on quasi - cylinders with trivial 2-homology . a _ seifert surface _ for a link @xmath11 in a 3-manifold @xmath184 is a compact connected oriented surface in @xmath52 that has @xmath11 as its oriented boundary . clearly , if @xmath11 has a seifert surface , then @xmath12=0 $ ] in @xmath185 . it is well - known that this is the only obstruction . for completeness , we outline a proof . [ prop : surface ] any homologically trivial link in an oriented 3-manifold has a seifert surface . let @xmath11 be a homologically trivial link in an oriented 3-manifold @xmath1 . then @xmath11 is homologically trivial in a compact 3-dimensional submanifold @xmath186 of @xmath1 such that @xmath187 . let @xmath188 be a closed tubular neighborhood of @xmath11 in @xmath189 . set @xmath190 . since @xmath12=0\in h_1(m')$ ] , an appropriate choice of longitudes of @xmath11 gives a link @xmath191 whose class in @xmath192 is equal to @xmath193 . then @xmath194 \in h_1 ( \partial x)$ ] is the boundary of an element of @xmath195 . the latter is the pull - back of a generator of @xmath196 under a map @xmath197 . for an appropriate choice of this map , the preimage of a point of @xmath198 is a compact oriented surface bounded by @xmath13 in @xmath199 . adding if necessary 1-handles to this surface one can make it connected . the resulting surface extends to a seifert surface for @xmath11 in @xmath186 . given two seifert surfaces @xmath200 for a link @xmath11 in an oriented 3-manifold @xmath1 , the union @xmath201 is a closed oriented surface representing an element of @xmath76 . this element is an obstruction to transforming @xmath108 into @xmath183 by surgeries . it is well - known that this is the only obstruction ( see e.g. @xcite ) . in particular , if @xmath73 , then @xmath202 can be related by a finite sequence of surgeries and ambient isotopies in @xmath1 ( which can be chosen to keep @xmath15 fixed ) . combining this fact with the observations above , we obtain the following . [ thm1 ] let @xmath75 be a quasi - cylinder over @xmath0 with @xmath73 . for any homologically trivial link @xmath203 , the equivalence class of the seifert triple of a seifert surface for @xmath11 does not depend on the choice of the surface and provides an isotopy invariant of @xmath11 . throughout this section , @xmath70 is a quasi - cylinder over @xmath0 with @xmath73 . fix a commutative unital ring @xmath204 containing @xmath0 as a subring . we also fix an @xmath0-bilinear pairing @xmath205 . consider a homologically trivial link @xmath11 in @xmath1 . let @xmath206 be the seifert triple associated with a seifert surface for @xmath11 . let @xmath173 and @xmath207 be the matrices of the bilinear forms @xmath151 and @xmath208 with respect to a basis of @xmath150 . the _ alexander module _ @xmath209 of @xmath11 is the @xmath210$]-module presented by the matrix @xmath211 , where the superscript @xmath212 denotes the matrix transposition . [ prop : module ] the alexander module is an isotopy invariant of @xmath11 . obviously , this module does not depend on the choice of a basis of @xmath150 . by theorem [ thm1 ] , we just need to check that if @xmath160 is obtained from @xmath149 by an elementary enlargement , then the corresponding matrices @xmath213 and @xmath214 present isomorphic @xmath210$]-modules . clearly , @xmath215 therefore , @xmath216 in both cases , the corresponding modules over @xmath210 $ ] are isomorphic . for @xmath217 , @xmath218 , @xmath219 and @xmath220 , the module @xmath209 is the usual alexander module . mimicking the standard definitions , we can introduce the alexander ideals and alexander polynomials of @xmath11 ( provided @xmath204 is a unique factorization domain ) . in particular , the first alexander polynomial of @xmath11 can be defined as the determinant of a square presentation matrix of @xmath209 . this polynomial is an element of the ring @xmath221 $ ] defined up to multiplication by units of this ring . as in the classical case , the first alexander polynomial has a canonical normalization which we now discuss . using the notation of the previous subsection , we define the _ ( extended ) alexander - conway polynomial _ of @xmath11 by @xmath222 as in the proof of proposition [ prop : module ] , one checks that this element of @xmath223 $ ] is a well - defined isotopy invariant of @xmath11 . observe that the size of the matrices @xmath224 is equal to @xmath225 , where @xmath178 is the genus of the seifert surface and @xmath179 is the number of components of @xmath11 . therefore @xmath226 thus , @xmath227 $ ] for odd @xmath179 and @xmath228 $ ] for even @xmath179 . we now establish a skein formula for @xmath229 . [ conway ] let @xmath230 , @xmath231 and @xmath232 be homologically trivial links in @xmath1 which coincide everywhere except in a small @xmath233-ball where they are related as illustrated below . 2.5pt at 87 -20 at 340 -20 at 570 -20 then , the corresponding alexander - conway polynomials satisfy the following relation : @xmath234 let @xmath235 be a seifert surface for @xmath232 . then a seifert surface @xmath236 for @xmath230 ( resp . @xmath237 for @xmath231 ) is obtained from @xmath235 by adding a band in the small @xmath233-ball with one negative ( resp . positive ) half - twist . since @xmath235 is connected , a basis for @xmath238 ( resp . for @xmath239 ) is obtained from a basis for @xmath240 by adding a @xmath119-cycle @xmath241 ( resp . @xmath242 ) . clearly , @xmath241 and @xmath242 can be chosen to coincide as @xmath119-cycles in @xmath1 . let @xmath41 be a 1-cycle on @xmath15 with coefficients in @xmath0 such that @xmath243\in v\subset h_1(\partial m;r)$ ] and @xmath41 is homologous to @xmath244 in @xmath1 . let @xmath24 be a @xmath122-cycle in @xmath1 such that @xmath245 . then @xmath246 this leads to the following equalities between the corresponding matrices : @xmath247 for some @xmath248 , column @xmath249 and row @xmath250 over @xmath0 , and column @xmath251 and row @xmath252 over @xmath204 . the skein formula follows . the skein formula implies in particular that @xmath253 is unchanged when one replaces an undercrossing by an overcrossing . hence , it depends only on the homotopy type of the components of @xmath11 . if @xmath13 is a link in an oriented 3-ball @xmath83 and @xmath254 is the image of @xmath13 under an orientation preserving embedding @xmath255 , then @xmath256 is the usual conway - normalized alexander polynomial of @xmath13 . let @xmath257 $ ] be the polynomial ring over @xmath0 generated by @xmath258 commuting variables @xmath259 . let @xmath260 be bilinear forms . we can apply the definitions and results of the previous subsections to the bilinear form @xmath261 this gives a polynomial invariant @xmath262 where @xmath173 and @xmath263 are the matrices of the bilinear forms @xmath151 and @xmath264 with respect to a basis of @xmath150 . the polynomial @xmath265 lies in @xmath266 $ ] for odd @xmath179 and in @xmath267 $ ] for even @xmath179 . the degree in @xmath268 of @xmath269 is bounded from above by a number independent of @xmath11 . namely , this degree is smaller than or equal to the rank of the form @xmath270 . indeed , it follows from the definitions that @xmath271 for a bilinear form @xmath272 we denote by @xmath273 its transpose defined by @xmath274 for @xmath275 . [ prop : properties ] @xmath276 transposing matrices , we obtain @xmath277 for example , if @xmath270 is symmetric for @xmath278 and skew - symmetric for @xmath279 , then @xmath280 @xmath281 the polynomial @xmath265 leads to other polynomial invariants of @xmath11 . first of all , we can expand @xmath282 where @xmath283 belongs to @xmath284 $ ] for odd @xmath179 and to @xmath285 $ ] for even @xmath179 . the sum on the right - hand side is finite since @xmath286 provided @xmath287 for some @xmath288 . for any triple @xmath289 as in proposition [ conway ] and for any @xmath290 , @xmath291 another interesting restriction of @xmath292 is obtained by the substitution @xmath293 . by the skein relation , the resulting polynomial depends only on the homotopy type of the components of @xmath11 . if @xmath2 is a free module , then we can take as @xmath294 a basis in the @xmath0-module of bilinear forms @xmath295 . this results in a link polynomial on @xmath296 variables , where @xmath41 is the rank of @xmath2 . take @xmath297 and let @xmath298 be the homological intersection on @xmath15 restricted to @xmath2 . this gives a polynomial invariant @xmath299 where @xmath173 and @xmath207 are the matrices of the bilinear forms @xmath151 and @xmath300 with respect to a basis of @xmath150 . we leave to the reader to check the following three properties of @xmath301 , where @xmath179 denotes the number of components of @xmath11 : @xmath302 if @xmath303 and @xmath304 otherwise ; @xmath305 ; @xmath306 , where @xmath307 is @xmath11 with opposite orientation . we can sometimes explicitly compute @xmath308 for links @xmath11 represented by simple closed curves on @xmath15 . let @xmath309 be a compact connected surface of genus g with boundary , such that the image of the inclusion homomorphism @xmath310 is contained in @xmath2 . we endow @xmath6 with the orientation induced by the orientation on @xmath15 ( which in its turn is induced by the one on @xmath1 ) . let @xmath311 be the oriented surface obtained by pushing @xmath6 inside @xmath1 and reversing its orientation . clearly , @xmath312 is a homologically trivial link with seifert surface @xmath313 . it is easy to see that the form @xmath151 associated with @xmath313 is identically zero . it follows from the definitions that @xmath314 if @xmath11 is a knot , and @xmath315 else . let @xmath6 be a compact connected oriented surface of genus @xmath178 with @xmath98 . consider the quasi - cylinder @xmath316,v = h_1({\sigma}\times 0)$ ] over @xmath74 . for any knot @xmath29 in @xmath1 , the laurent polynomial @xmath317 $ ] introduced in the previous example satisfies @xmath318 , @xmath319 and @xmath320 . if @xmath321 ( that is , if @xmath6 is a disc with holes ) , then these conditions characterize completely the polynomials @xmath322 which can be realized as the alexander - conway polynomial of a knot in @xmath1 . indeed , in this case @xmath323 $ ] , @xmath324 so that @xmath322 can be realized as the alexander - conway polynomial of a knot in a 3-ball in @xmath1 . we do not know whether the conditions above are sufficient for @xmath325 . the genus of a homologically trivial link @xmath11 in a 3-manifold @xmath1 is defined by @xmath326 if @xmath1 is a 3-ball , then seifert proved that @xmath327 , where @xmath328 is the usual span of a laurent polynomial in one variable @xmath329 and @xmath179 is the number of components of @xmath11 . this result extends to our setting as follows . [ prop : genus1 ] let @xmath11 be a homologically trivial @xmath179-component link in a quasi - cylinder @xmath70 with @xmath73 . let @xmath330 be a pairing as in section [ sec2.1 ] . then @xmath331 let @xmath108 be a seifert surface for @xmath11 realizing the genus @xmath332 , and let @xmath333 be corresponding matrices . by definition of @xmath334 , @xmath335 the inequality follows . consider now a homologically trivial @xmath179-component link @xmath11 in @xmath336 $ ] , where @xmath6 is a compact connected oriented surface of genus @xmath178 . the following algorithm ( due to seifert in the case where @xmath6 is a 2-disc ) produces a seifert surface for @xmath11 from a connected diagram of @xmath11 on @xmath6 . ( a link diagram is connected if it can not be presented as a union of disjoint non - empty link diagrams . ) let @xmath258 be the number of crossings on the diagram . smoothing these crossings in the unique way compatible with the orientation of @xmath11 , one obtains a closed oriented 1-manifold @xmath337 consisting of @xmath338 disjoint simple closed curves on @xmath6 . note that @xmath339=[l]=0\in h_1({\sigma})$ ] . therefore , there is a finite collection of oriented connected subsurfaces @xmath340 of @xmath341 whose boundaries are disjoint and @xmath342 . a seifert surface @xmath108 for @xmath11 can be obtained from the @xmath343 by pushing their interiors into @xmath344 $ ] and adding a half - twisted band at each crossing . [ prop : genus2 ] let @xmath345 be the number of discs among the surfaces @xmath340 . then @xmath346 and @xmath347 we have @xmath348 where @xmath349 is the genus of @xmath343 . clearly @xmath350 and @xmath351 if @xmath352 is a disc . hence @xmath353 the inequalities @xmath354 now give the result . note that if @xmath321 , then @xmath355 and we obtain seifert s inequality @xmath356 for links in the 3-ball . combining propositions [ prop : genus1 ] and [ prop : genus2 ] , we obtain in the case @xmath357 that @xmath358 two links @xmath359 in a 3-manifold @xmath1 are _ concordant _ if there is a smooth oriented surface @xmath360 $ ] such that @xmath361 and each component of @xmath362 is an annulus with one boundary component on @xmath363 and the other one on @xmath364 . cobordant links have the same number of components . [ lemma : conc ] assume that @xmath0 is a principal ideal domain . let @xmath70 be a quasi - cylinder over @xmath0 such that @xmath1 is compact and @xmath73 . let @xmath330 be a bilinear pairing with values in an integral domain @xmath204 containing @xmath0 as a subring . let @xmath359 be concordant homologically trivial links in @xmath1 and @xmath365 be their seifert surfaces with associated seifert triples @xmath366 , @xmath367 . then there is a basis @xmath368 of the @xmath0-module @xmath369 such that the bilinear forms @xmath370 satisfy @xmath371 for all @xmath372 . let @xmath373 $ ] be a surface as in the definition of the link concordance . then @xmath374 is a closed connected oriented surface in @xmath375 $ ] . _ claim 1 : _ there is a compact oriented @xmath233-manifold @xmath376 $ ] such that @xmath377 . indeed , let @xmath378 be a closed tubular neighborhood of @xmath379 in @xmath380 for @xmath381 . let @xmath382 be the closure of @xmath383 . deforming if necessary @xmath362 , we can assume that @xmath362 meets @xmath384)$ ] precisely along @xmath361 . let @xmath385 be a closed tubular neighborhood of @xmath362 in @xmath375 $ ] . deforming if necessary @xmath386 , we can assume that @xmath387 for @xmath381 . let @xmath388 be the closure of @xmath389 ) \setminus u$ ] . then @xmath388 is a compact oriented 4-manifold with boundary and @xmath390 for @xmath381 . we define a continuous map @xmath391 as follows . for @xmath381 , let @xmath392 $ ] be a closed tubular neighborhood of @xmath393 in @xmath394 . then , @xmath395 restricted to @xmath392 $ ] is given by @xmath396 for @xmath397 $ ] . on @xmath398 , the map @xmath395 is such that @xmath399 for some @xmath400 . finally , @xmath401 for all @xmath402 $ ] and all @xmath403)$ ] where @xmath381 . by elementary obstruction theory , the map @xmath391 extends to @xmath388 if and only if there is a homomorphism @xmath404 such that @xmath405 , where @xmath406 is the inclusion @xmath407 . using the exact homology sequence of the pair @xmath408,y)$ ] , the excision theorem , and the assumption @xmath73 , we obtain that @xmath409 and @xmath410 where @xmath179 is the number of components of @xmath232 ( and of @xmath411 ) . a basis of @xmath412 is given by the homology classes of @xmath179 tori @xmath413 forming @xmath414 . we have @xmath415 and @xmath416 where @xmath101 is a finite abelian group . the summand @xmath417 has a basis @xmath418 dual to the basis @xmath419,\dots,[t_m]$ ] of @xmath412 . the homological sequence of the pair @xmath420 yields @xmath421 clearly , @xmath422 . using the assumption @xmath423 , it is easy to construct for each @xmath424 , a loop in @xmath425 piercing @xmath426 once and disjoint from the other @xmath427 tori . this loop represents @xmath428 mod @xmath429 . therefore , @xmath430 for all @xmath431 . thus , the obstruction to the extension of @xmath395 to @xmath388 mentioned above is @xmath193 . let @xmath432 be a continuous extension of @xmath395 . deform @xmath433 so that @xmath119 is one of its regular values . then the 3-manifold @xmath434 satisfies the conditions of claim 1 . set @xmath435 , which we identify with @xmath369 via the inclusion homomorphism . let @xmath436 ( resp . @xmath29 ) be the kernel of the inclusion homomorphism @xmath437 ( resp . @xmath438 ) , where @xmath439 and @xmath440 denotes the field of fractions of @xmath0 . by the standard argument using the poincar - lefschetz duality , the dimension of @xmath436 is half of the dimension of @xmath441 . furthermore , one easily checks that both the kernel and the cokernel of the inclusion homomorphism @xmath442 have dimension @xmath427 . therefore , @xmath443 we now use this fact to show a second claim . the proof is adapted from @xcite . _ claim 2 : _ there is an @xmath0-basis @xmath368 of @xmath150 such that @xmath444 maps to zero in @xmath445 for all @xmath446 . observe first that @xmath150 is a free @xmath0-module of rang @xmath447 where @xmath178 is the genus of @xmath448 . then @xmath449 is a vector space over @xmath450 of dimension @xmath447 and @xmath451 . pick a @xmath178-dimensional subspace @xmath452 of @xmath29 . clearly , @xmath452 admits a @xmath450-basis consisting of elements in @xmath150 : just take any @xmath450-basis of @xmath452 and multiply its vectors by non - zero scalars . let @xmath453 be the @xmath0-span of these elements in @xmath150 . since @xmath0 is a principal ideal domain , @xmath454 where @xmath108 is a free @xmath0-module of rank @xmath178 and @xmath212 a torsion @xmath0-module . let @xmath455 be the pre - image of @xmath212 under the projection @xmath456 . then @xmath457 and @xmath458 . since @xmath0 is a principal ideal domain and @xmath150 is free , @xmath459 is free as well . since the sequence @xmath460 is exact and @xmath108 is free , a basis for @xmath459 can be completed to an @xmath0-basis of @xmath461 which satisfies the conditions of claim 2 . the lemma now follows from one last claim . _ claim 3 : _ if @xmath462 map to zero in @xmath445 , then @xmath463 . indeed , if @xmath462 map to zero in @xmath464 , then @xmath465 and @xmath466 map to zero in @xmath467 for some non - zero @xmath468 . by @xmath0-bilinearity of @xmath469 and @xmath151 and the assumption that @xmath204 is an integral domain , it is enough to consider the case where @xmath462 map to zero in @xmath470 . we have @xmath471 and @xmath472 with @xmath473 and @xmath474 . consider the following inclusion homomorphisms @xmath475;r)=h_1(m;r)=v .\ ] ] clearly , the composition is given by @xmath476 . since @xmath112 are in the kernel of this composition , @xmath477 . hence , @xmath478 by the assumptions on @xmath112 , there are 2-chains @xmath479 in @xmath99 such that @xmath480 and @xmath481 . let @xmath482 be a 2-cycle in @xmath483 such that @xmath484 for @xmath381 . then @xmath485 the equality @xmath486 implies that there is a 2-chain @xmath487 in @xmath488 $ ] such that @xmath489 . since @xmath487 is disjoint from @xmath490 and @xmath491 , and @xmath492 is disjoint from @xmath493 for @xmath494 , @xmath495 ) } ( b_0+b_1+z).\ ] ] here we used the fact that the orientation on @xmath496)$ ] matches the one on @xmath364 and is opposite to the one on @xmath363 . there is a map @xmath497)\setminus n$ ] extending the push in the positive normal direction @xmath498 for @xmath381 . let @xmath499 be the image of @xmath139 under this map . then @xmath500}(b_0+b_1+z)= \alpha^+\cdot_{m\times[0,1]}\beta,\ ] ] since @xmath501 is a @xmath122-cycle , and therefore a @xmath122-boundary , in @xmath502 $ ] . finally , @xmath503 and @xmath504)\setminus n$ ] are disjoint , so @xmath505}\beta=0 $ ] . this concludes the proof . the following theorem generalizes the results of fox - milnor @xcite for knots in @xmath506 . [ irg ] let @xmath359 be concordant homologically trivial links in a quasi - cylinder @xmath70 over a principal ideal domain @xmath0 such that @xmath1 is compact and @xmath73 . let @xmath260 be bilinear forms such that @xmath507 is symmetric for @xmath508 and skew - symmetric for @xmath509 . then for some @xmath510 $ ] , @xmath511 @xmath512 @xmath513 by lemma [ lemma : conc ] , the matrices of the bilinear pairings @xmath514 and @xmath515 with respect to a certain basis of @xmath369 have the form @xmath516 where @xmath517 are square matrices over @xmath0 of equal size . note that @xmath518 for @xmath508 and @xmath519 for @xmath509 . let @xmath179 be the number of components of @xmath232 ( and of @xmath411 ) . by proposition [ prop : properties ] , @xmath520 @xmath521 @xmath522 where @xmath523 . ( the sign @xmath524 disappears because of the minuses in the definition of the forms @xmath525 and @xmath526 . ) the classical murasugi - tristram - levine signature of a link @xmath11 in the 3-ball is the function @xmath527 whose value on @xmath528 is the signature of the hermitian matrix @xmath529 , where @xmath173 is a seifert matrix of @xmath11 . this function is a well - defined invariant of @xmath11 . it is a concordance invariant away from the roots of @xmath530 on @xmath198 . we now extend these results to our setting . consider a quasi - cylinder @xmath70 over @xmath531 . fix @xmath532 symmetric bilinear forms @xmath533 and @xmath534 skew - symmetric bilinear forms @xmath535 . let @xmath11 be a homologically trivial link in @xmath1 and @xmath536 be the seifert triple associated with a seifert surface for @xmath11 . the _ signature of @xmath11 _ is the function @xmath537 sending a tuple @xmath538 to the signature of the hermitian form @xmath539 on @xmath540 . using theorem [ thm1 ] , one easily checks that @xmath541 does not depend on the choice of the seifert surface ( see e.g. ( * ? ? ? * chapter 8) for a proof which extends to our setting ) . thus , it is a well - defined isotopy invariant of @xmath11 . [ thm : sign ] let @xmath359 be concordant homologically trivial links in a quasi - cylinder @xmath70 over @xmath542 such that @xmath1 is compact and @xmath73 . then @xmath543 for all @xmath544 and @xmath545 such that both @xmath546 and @xmath547 do not vanish on @xmath548 where @xmath549 . we shall use the notation introduced in the proof of theorem [ irg ] . clearly , @xmath550 , where @xmath551 with @xmath552 square matrices over @xmath553 of equal size . therefore , @xmath554 unless @xmath555 is degenerate . we have @xmath556 where @xmath557 and @xmath558 are the matrices of the forms @xmath559 and @xmath560 on @xmath561 . for @xmath381 , the @xmath95-th determinant on the right - hand side is equal to @xmath562 where @xmath563 . this proves the theorem . we assume in this subsection that the ground ring @xmath0 is a field and @xmath148 is a vector space over @xmath0 . more invariants of seifert triples can be obtained using the following construction . a seifert triple @xmath149 over @xmath148 gives a seifert triple @xmath160 over any submodule @xmath564 of @xmath148 by @xmath565 , @xmath566 , and @xmath567 . the latter triple is said to be a _ restriction _ of @xmath149 . note that equivalent seifert triples may give non - equivalent restrictions . to handle this , we introduce a notion of stable equivalence for seifert triples . we say that a seifert triple @xmath156 over @xmath148 is obtained from a seifert triple @xmath155 over @xmath148 by a _ trivial enlargement _ ( and @xmath155 is obtained from @xmath156 by a _ trivial reduction _ ) if @xmath568 , @xmath569 , @xmath570 , @xmath571 , @xmath572 . thus , a matrix of @xmath573 is obtained from a matrix of @xmath574 by adding a zero row and a zero column . two seifert triples over @xmath148 are _ stably equivalent _ if they can be related by ( a finite sequence of ) isomorphisms , elementary enlargements and reductions , and trivial enlargements and reductions . it is easy to check that stably equivalent seifert triples over @xmath148 restrict to stably equivalent seifert triples over submodules of @xmath148 . therefore a stable equivalence invariant of seifert triples generates a family of such invariants by applying it to all possible restrictions of a given seifert triple . given a seifert triple @xmath149 over @xmath148 , the associated polynomial @xmath575 as in section [ sec2.2 ] is not preserved under trivial enlargements . the module presented by the matrix @xmath211 is preserved up to taking direct sums with free @xmath221$]-modules of finite rank . the sequence of elementary ideals of this module is preserved up to shifts of the index . the signatures of seifert triples are easily seen to be invariant under stable equivalence . this generates a family of stable equivalence invariants obtained by taking the signatures of the restrictions . applying the constructions above to homologically trivial links in a quasi - cylinder @xmath70 over @xmath542 , we obtain _ derived signatures _ numerated by the subspaces of @xmath2 . they are isotopy invariants . we do not know whether they are concordance invariants or not . the classical theory of seifert surfaces for oriented links in @xmath506 has been extended to @xmath576-colored links in @xmath506 using ` c - complexes ' ( see @xcite ) . the aim of the present section is to sketch a further extension of this theory to @xmath576-colored links in quasi - cylinders . let @xmath576 be a fixed positive integer . @xmath576-colored link _ @xmath577 in an oriented @xmath233-manifold @xmath1 is an oriented link in the interior of @xmath1 together with a surjective map assigning to each component of @xmath11 a color in @xmath578 . the sublink @xmath579 is constitued by the components of @xmath11 with color @xmath406 for @xmath580 . we shall say that two colored links @xmath581 in @xmath1 are _ isotopic _ if there is an ambient isotopy between @xmath11 and @xmath13 , fixing @xmath15 , and preserving the orientation and color of every component . a @xmath576-colored link @xmath577 is _ homologically trivial _ if @xmath582=0 $ ] in @xmath185 for all @xmath580 . note that a @xmath119-colored link is an ordinary link , as defined in section [ section : lk ] . setting @xmath583 in the present section , we obtain the theory developed in the previous sections . 2.5pt at 15 75 at 380 75 a _ c - complex _ for a @xmath576-colored link @xmath577 in an oriented @xmath233-manifold @xmath1 is a union @xmath584 of surfaces in @xmath1 such that @xmath108 is connected , and the following conditions hold : for all @xmath406 , @xmath585 is a seifert surface for @xmath579 ; for all @xmath586 , @xmath587 is either empty or a union of clasps ( see figure [ fig : clasp ] ) ; for all @xmath588 pairwise distinct , @xmath589 is empty . in the case @xmath583 , a c - complex for @xmath11 is simply a seifert surface for @xmath11 . in order to have a c - complex , a @xmath576-colored link clearly needs to be homologically trivial . one easily checks that it is the only obstruction : every homologically trivial @xmath576-colored link @xmath577 in an oriented @xmath233-manifold has a c - complex . indeed , by proposition [ prop : surface ] , every sublink @xmath579 admits a seifert surface @xmath585 . then , by ( * ? ? ? * lemma 1 ) , each @xmath585 can be isotoped keeping its boundary fixed to give a c - complex for @xmath11 . [ prop : cmoves ] let @xmath108 and @xmath183 be c - complexes for isotopic colored links in a quasi - cylinder @xmath70 over @xmath0 . if @xmath73 , then @xmath108 and @xmath183 can be transformed into each other by a finite number of the following operations and their inverses : ambient isotopy keeping @xmath15 fixed ; surgery on one surface ; addition of a ribbon intersection , followed by a ` push along an arc ' through this intersection ( see figure [ fig : sequiv ] ) ; the transformation described in figure [ fig : sequiv ] . by the first move , it may be assumed that @xmath590 for all @xmath406 . since @xmath73 , @xmath585 and @xmath591 are related by ambient isotopies ( keeping @xmath579 fixed ) and surgeries . clearly , a surgery on @xmath585 can be performed avoiding @xmath592 , giving move @xmath593 . now , for every ambient isotopy between @xmath585 and @xmath591 , we can apply ( * ? ? ? * lemma 3 ) , whose proof extends to our setting : such an ambient isotopy can be induced by a finite sequence of moves @xmath594 , @xmath595 , @xmath596 and their inverses . 2.5pt at 330 200 at 1140 200 let us now define the corresponding generalization of the seifert form . let as above @xmath0 be an arbitrary commutative ring with unit . let @xmath597 $ ] be a bicollar neighborhood of @xmath585 in the interior of @xmath1 . given a sign @xmath598 , let @xmath599 be the translated surface @xmath600 . also , let @xmath601 be a tubular neighborhood of @xmath579 in @xmath52 , and let @xmath388 be the complement of @xmath602 in @xmath1 . given a sequence @xmath603 of @xmath604 , set @xmath605 see figure [ fig : eps ] for an illustration of @xmath606 near a clasp . since all the intersections are clasps , there is an obvious homotopy equivalence between @xmath108 and @xmath606 inducing an isomorphism @xmath607 , @xmath608 . note also that @xmath606 is a smooth surface , endowed with a canonical orientation : the orientation that matches the one on @xmath585 if and only if @xmath609 . hence , we have a well - defined seifert form @xmath610 on @xmath611 as in section [ section : forms ] . therefore , each choice of signs @xmath603 leads to a seifert form @xmath612 and to an intersection form @xmath613 on @xmath116 defined by @xmath614 for all @xmath112 in @xmath116 . these forms are related as follows . [ lemma : crelation ] for all @xmath112 in @xmath116 and all signs @xmath603 , @xmath615 where @xmath54 is the intersection pairing on @xmath15 and @xmath616 the composition of the inclusion homomorphism @xmath130 with the isomorphism @xmath131 . let @xmath617 denote the isomorphism given by @xmath608 . as an oriented smooth surface , @xmath606 is diffeomorphic to @xmath618 , the surface @xmath619 with the opposite orientation . this leads to a canonical isomorphism @xmath620 such that @xmath621 and @xmath622 . ( recall that the bilinear form @xmath623 is defined as @xmath624 but using @xmath110 instead of @xmath109 . ) therefore : @xmath625 by formula ( 2.a ) applied to @xmath606 , this is equal to @xmath626 giving the result . the second equality follows from formula ( 2.b ) in a similar way . 2.5pt at 85 160 at 300 170 at 33 111 at 265 -10 at 465 170 this result leads to the following definition . @xmath576-colored seifert triple _ over an @xmath0-module @xmath148 is a triple @xmath627 , where @xmath150 is a free @xmath0-module of finite rank , @xmath628 a family of @xmath629 bilinear forms on @xmath150 indexed by the set @xmath630 and @xmath153 a homomorphism @xmath631 . ( note that we do nt consider the forms @xmath612 with @xmath632 since they can be recovered from the other forms via lemma [ lemma : crelation ] . ) a @xmath576-colored seifert triple @xmath633 is obtained from another @xmath576-colored seifert triple @xmath634 by a _ type i elementary enlargement _ if the following conditions hold : @xmath635 , @xmath636 , @xmath637 , and there is some index @xmath406 and some sign @xmath638 such that for all @xmath639 , the matrix @xmath640 for @xmath612 with respect to a basis @xmath168 of @xmath150 is related to the matrix @xmath641 with respect to the basis @xmath642 of @xmath643 by @xmath644 where @xmath645 is the kronecker symbol . similarly , one speaks of _ type ii elementary enlargement _ if the following conditions hold : @xmath635 , @xmath636 , @xmath637 , and there is some indices @xmath586 and some signs @xmath646 such that @xmath647 we shall say that two @xmath576-colored seifert triples over @xmath148 are _ equivalent _ if they can be related by a finite number of type i and ii elementary enlargements ( and reductions ) . [ thmc ] let @xmath70 be a quasi - cylinder over @xmath0 with @xmath73 . for any homologically trivial @xmath576-colored link @xmath11 in @xmath1 , the equivalence class of the @xmath576-colored seifert triple of a c - complex for @xmath11 does not depend on the choice of the c - complex and provides an isotopy invariant of the @xmath576-colored link @xmath11 . by proposition [ prop : cmoves ] , we are left with the proof that if two c - complexes are related by transformations @xmath594 to @xmath596 , then the corresponding seifert triples are equivalent . obviously , transformation @xmath594 does not change the seifert triple . it is an easy exercice to check that if a c - complex @xmath648 is obtained from a c - complex @xmath108 via surgery on @xmath585 , then the corresponding seifert triples are related by a type i elementary enlargement with index @xmath406 . ( the sign @xmath649 is determined by the side of @xmath585 along which the surgery is performed . ) also , one verifies that transformation @xmath595 involving surfaces @xmath585 and @xmath650 corresponds to a type ii elementary enlargement with indices @xmath651 , and some signs @xmath646 given by the orientations of @xmath585 and @xmath650 . finally , consider two c - complexes related by a @xmath596 transformation . then , the two corresponding seifert triples can be understood as two distinct type ii elementary enlargements of some fixed seifert triple . this concludes the proof . fix a commutative unital ring @xmath204 containing @xmath0 as a subring , and an @xmath0-bilinear pairing @xmath652 . consider a homologically trivial @xmath576-colored link @xmath11 in @xmath1 , and let @xmath653 be the @xmath576-colored seifert triple associated with a c - complex @xmath108 for @xmath11 . let @xmath654 and @xmath207 be the matrices of the bilinear forms @xmath655 and @xmath208 with respect to a basis of @xmath150 . let @xmath656 denote the localization of the ring @xmath657 $ ] with respect to the multiplicative system generated by @xmath658 . the _ ( extended ) conway function _ of @xmath11 is the element of @xmath656 defined by @xmath659 where @xmath33 is the number of clasps in @xmath108 , @xmath660 , and @xmath661.\ ] ] the extended conway function is an isotopy invariant of the @xmath576-colored link @xmath11 . by proposition [ prop : cmoves ] and the proof of theorem [ thmc ] , we just need to check that @xmath662 remains unchanged if the c - complex @xmath108 is transformed via moves @xmath593 and @xmath595 . so , let @xmath663 be a c - complex obtained from @xmath108 by a surgery on @xmath380 . clearly , the number of clasps @xmath33 remains the same , while @xmath664 furthermore , the corresponding @xmath576-colored seifert triples are related by a type i elementary enlargement ( with index @xmath665 ) . using the equality @xmath666=\ ] ] @xmath667 we get @xmath668 therefore , @xmath669 . the equality follows . now , let @xmath663 be a c - complex obtained from @xmath108 by a move @xmath595 involving @xmath380 and @xmath670 . the number of clasps @xmath671 of @xmath663 is given by @xmath672 , and @xmath673 the corresponding @xmath576-colored seifert triples are related by a type ii elementary enlargement with indices @xmath674 . by the equality @xmath675=\ ] ] @xmath676 we get @xmath677 the invariance follows . in the case @xmath583 , @xmath108 is a seifert surface for @xmath11 , and the unique seifert matrix coincides with the matrix @xmath173 constructed in section [ section : forms ] . furthermore , we have @xmath678 , @xmath679 . hence , the conway function is given by @xmath680 where @xmath179 is the number of components of @xmath11 . if @xmath13 is a @xmath576-colored link in an oriented @xmath233-ball @xmath83 and @xmath11 is the image of @xmath13 under an orientation preserving embedding @xmath255 , then @xmath681 is the usual conway function of @xmath13 , as constructed in @xcite . let us conclude this paragraph with a list of properties of @xmath662 generalizing well - known properties of the conway function of colored links in @xmath506 . we refer to @xcite for the proofs which easily extend to our setting . @xmath682 let @xmath230 , @xmath231 and @xmath232 be homologically trivial @xmath576-colored links which coincide everywhere except in a small @xmath233-ball where they are related as illustrated below . ( here , @xmath406 denotes the color of the strands in the @xmath233-ball . ) 2.5pt at -10 150 at 170 150 at 245 150 at 423 150 at 483 150 at 655 150 at 87 -20 at 340 -20 at 570 -20 then , the corresponding conway functions satisfy the following relation : @xmath683 @xmath684 similarly , if @xmath685 , @xmath686 and @xmath687 are homologically trivial @xmath576-colored links which differ by the following local operation , 2.5pt at -20 205 at 110 205 at 250 205 at 383 205 at 505 205 at 655 205 at 45 -20 at 315 -20 at 575 -20 then we have the equality @xmath688 @xmath689 for any homologically trivial @xmath576-colored link @xmath11 with @xmath179 components , @xmath690 where @xmath691 is the bilinear form given by @xmath692 . as in section [ section : signature ] , consider a quasi - cylinder @xmath70 over @xmath531 , and fix @xmath532 symmetric bilinear forms @xmath693 and @xmath694 skew - symmetric bilinear forms @xmath695 . let @xmath11 be a @xmath576-colored homologically trivial link in @xmath1 and @xmath627 be the @xmath576-colored seifert triple associated with a c - complex for @xmath11 . finally , let @xmath696 denote the @xmath576-dimensional torus @xmath697 . the _ ( extended ) signature of @xmath11 _ is the function @xmath698 sending a tuple @xmath699 to the signature of the hermitian form @xmath700+\psi\ ] ] on @xmath540 , where @xmath701 . the extended signature is an isotopy invariant of the @xmath576-colored link @xmath11 . note that if @xmath702 for some @xmath406 , then the signature is equal to zero . therefore , it may be assumed that @xmath703 for all @xmath406 . by theorem [ thmc ] , we just need to check that the signatures corresponding to equivalent @xmath576-colored seifert triples are equal . so , let us assume that a seifert triple @xmath633 is obtained from another seifert triple @xmath634 by a type i elementary enlargement ( with index @xmath665 ) . using the equality @xmath704=\ ] ] @xmath705 we see that the corresponding hermitian matrices @xmath706 and @xmath1 are related by @xmath707 since @xmath703 for all @xmath406 , the signatures of @xmath706 and @xmath1 coincide by the usual argument . the invariance of the signature under elementary enlargement of type ii follows from the equality @xmath708=\ ] ] @xmath709 in the same way . in the case @xmath583 , we obviously get back the extended signatures defined in section [ section : signature ] . if @xmath13 is a @xmath576-colored link in an oriented @xmath233-ball @xmath83 and @xmath11 is the image of @xmath13 under an orientation preserving embedding @xmath255 , then @xmath710 is the multivariable signature of the @xmath576-colored link @xmath13 , as constructed in @xcite . we do nt know to which extent the concordance properties of these two special cases ( see theorem [ thm : sign ] and ( * ? ? ? * section 7 ) ) hold in the general case considered here . our invariants of links are defined under rather strong assumptions : the links are supposed to be homologically trivial ; the ambient manifold , @xmath1 , is supposed to have trivial 2-homology and the inclusion homomorphism @xmath67 is supposed to be surjective and to have a section . we explain how to weaken these conditions . let @xmath70 be a quasi - cylinder over @xmath0 with @xmath73 . let @xmath711 belong to the image of the inclusion homomorphism @xmath712 . to construct non - trivial invariants of links in @xmath1 representing @xmath168 , one can proceed as follows . pick a link @xmath713 in a cylinder neighborhood @xmath714 of @xmath15 such that @xmath715=-h$ ] . any link @xmath203 may be isotopically deformed in @xmath716 uniquely up to isotopy in @xmath716 . if @xmath717 and @xmath12=h $ ] , then @xmath718 is a homologically trivial link in @xmath1 . the isotopy type of @xmath719 is entirely determined by the isotopy type of @xmath11 and the isotopy type of @xmath713 in @xmath386 . the invariants of homologically trivial links in @xmath1 defined above may be applied to @xmath719 . this yields isotopy invariants of @xmath11 depending on @xmath2 and @xmath713 . in particular , concordance invariants of homologically trivial links yield concordance invariants of @xmath11 . indeed , if two links @xmath359 in @xmath1 are concordant , then @xmath720 and @xmath721 are concordant . a _ generalized quasi - cylinder _ over @xmath0 is a pair consisting of an oriented 3-manifold @xmath1 and a submodule @xmath2 of @xmath66 such that the inclusion homomorphism @xmath722 is injective . the theory of seifert triples associated with surfaces in quasi - cylinders extend to generalized quasi - cylinders as follows . given an oriented surface @xmath108 in the interior of @xmath1 , set @xmath723 where @xmath431 is the inclusion homomorphism @xmath130 . for 1-cycles @xmath112 on @xmath108 representing homology classes @xmath113 , [ b ] \in h $ ] , set @xmath724,[b])=lk_v(a^+,b)$ ] . this yields a well - defined bilinear form @xmath725 . applying this construction to the seifert surface for a link @xmath11 in @xmath1 , we obtain the seifert triple @xmath726 of @xmath11 . if @xmath73 and @xmath0 is a field , then the stable equivalence class of @xmath149 does not depend on the choice of @xmath108 and yields an isotopy invariant of @xmath11 . the definitions of linking numbers and generalized seifert forms given in sections [ section : lk ] and [ section : forms ] make perfect sense whether @xmath76 is trivial or not ( that is , whether @xmath15 is connected or not ) . however , the triviality of @xmath76 is needed for theorem [ thm1 ] to hold . indeed , this result is based on the fact that two seifert surfaces for a link in @xmath1 can be related by surgeries . this is clearly not true if @xmath727 . therefore , the general theory of sections [ section : alex ] to [ section : colored ] does not hold if the boundary of @xmath1 is non - connected , and it is very unlikely that any seifert type invariant can be constructed in this general setting . nevertheless , parts of the theory can be developed in the following special case . let @xmath70 be a quasi - cylinder over @xmath0 , and let us assume that @xmath1 has exactly two boundary components @xmath6 and @xmath728 , with @xmath729 . this is a natural class of quasi - cylinders , as it contains the prototypical example @xmath5 $ ] with @xmath6 closed . let @xmath108 be a seifert surface in such a quasi - cylinder @xmath70 , and let @xmath730 denote a parallel copy of @xmath6 obtained by pushing @xmath731 in @xmath732 . suppose that there is a solid cylinder @xmath733\times d^2 $ ] in the interior of @xmath1 such that @xmath734\times d^2)\cap f=\{0\}\times d^2 $ ] and @xmath734\times d^2)\cap\widetilde{\sigma}=\{1\}\times d^2 $ ] . then we shall say that the surface @xmath735\times\partial d^2)\cup(\widetilde{\sigma}\setminus ( \{1\}\times d^2))\ ] ] is obtained from @xmath108 by _ adding @xmath730 along the arc @xmath733\times\{0\}$]_. here , the orientation of @xmath730 is chosen so that the orientation of @xmath108 extends to @xmath183 . [ prop : closed ] let @xmath70 be a compact quasi - cylinder over @xmath0 with @xmath736 and @xmath729 . any two seifert surfaces @xmath200 for isotopic links in a @xmath70 can be related by a finite number of ambient isotopies keeping @xmath15 fixed , surgeries , and additions of parallel copies of @xmath6 along embedded arcs in @xmath52 . consider a path @xmath737\to m$ ] such that @xmath738)\cap{\sigma}=\gamma(0)$ ] , @xmath738)\cap{\sigma}'=\gamma(1)$ ] , and such that @xmath739 intersects @xmath6 , @xmath728 , @xmath108 and @xmath183 transversally . let us assume that @xmath108 intersects @xmath739 in @xmath258 points . let @xmath730 be a parallel copy of @xmath6 pushed into @xmath1 , disjoint from @xmath108 , and which intersects @xmath739 transversally in @xmath740 . let @xmath741 be the smallest number such that @xmath742 . consider the surface @xmath743 obtained from @xmath108 by adding @xmath730 along the arc @xmath744)$ ] . clearly , @xmath743 intersects @xmath739 in @xmath745 points . iterating this construction , we obtain a seifert surface @xmath746 for @xmath11 disjoint from @xmath739 . similarly , we obtain a seifert surface @xmath747 from @xmath183 disjoint from @xmath739 . now , consider the compact manifold @xmath748 given by the complement in @xmath1 of an open tubular neighborhood of @xmath739 . also , let @xmath749 be the surface with boundary given by @xmath750 . by excision , @xmath751 , so the homological sequence of @xmath752 reads @xmath753 since @xmath15 has exactly two components , one of which is @xmath6 , the inclusion homomorphism @xmath754 is an isomorphism , as well as the composition @xmath755 . therefore , @xmath756 is an isomorphism , and @xmath757 . so , we have two seifert surfaces @xmath758 and @xmath747 in @xmath748 for a fixed link @xmath11 in @xmath759 , with @xmath757 . by the standard argument , @xmath758 and @xmath747 are related by surgeries in @xmath760 and by isotopies of @xmath748 keeping its boundary fixed . such an isotopy obviously extends to an isotopy of @xmath1 fixing @xmath15 . this concludes the proof . let @xmath148 be a free @xmath0-module of finite rank equipped with bilinear form @xmath762 . let @xmath149 and @xmath160 be two seifert triples over @xmath148 . we shall say that @xmath160 is obtained from @xmath149 by a _ @xmath763-enlargement _ ( and @xmath149 from @xmath160 by a _ @xmath763-reduction _ ) if the following conditions hold : @xmath764 , @xmath162 , @xmath765 , @xmath164 , @xmath766 , @xmath767 and @xmath768 or @xmath763 . if @xmath168 is a basis of @xmath150 and @xmath250 a basis of @xmath148 , then @xmath769 is a basis of @xmath170 and the matrix @xmath171 for @xmath172 with respect to @xmath769 is computed from the matrix @xmath173 for @xmath151 with respect to @xmath168 by @xmath770 where @xmath771 is the matrix of @xmath772 , and @xmath773 the matrix of @xmath763 . we shall say that two seifert triples over @xmath148 are _ @xmath763-equivalent _ if they can be related by a finite number of isomorphisms , elementary enlargements , elementary reductions , @xmath763-enlargements and @xmath763-reductions . [ thm : closed ] let @xmath75 be a quasi - cylinder over @xmath0 and let us assume that @xmath1 has exactly two boundary components @xmath6 and @xmath728 , with @xmath729 . finally , let @xmath763 denote the intersection form on @xmath2 . for any homologically trivial link @xmath203 , the @xmath763-equivalence class of the seifert triple of a seifert surface for @xmath11 does not depend on the choice of the surface and provides an isotopy invariant of @xmath11 . by proposition [ prop : closed ] , we just need to check that the addition of a parallel copy of @xmath6 induces a @xmath763-enlargement of the corresponding seifert triple . let @xmath183 denote the seifert surface obtained from @xmath108 by the addition of @xmath730 along an arc , and let @xmath172 denote the corresponding form . clearly , @xmath774 , @xmath775 , @xmath776 and @xmath172 restricted to @xmath777 is equal to @xmath151 . furthermore , @xmath778 for @xmath779 in @xmath780 , since @xmath24 can be chosen to be a thin annulus @xmath781 $ ] disjoint from @xmath109 . for @xmath112 in @xmath782 , @xmath783)= a\cdot_{\partial m}b\ ] ] if the orientation of @xmath730 is induced by the one of @xmath6 and @xmath783)= 0\ ] ] if the orientation of @xmath730 is opposite to the one induced from @xmath6 . finally , for @xmath779 in @xmath784 , lemma [ lemma : relation ] and the above computation give @xmath785 this concludes the proof . the @xmath221$]-module @xmath209 is no longer an invariant of @xmath11 in general . however , it is an invariant in the special case @xmath786 and @xmath787 , where @xmath763 is the intersection form on @xmath2 . indeed , if @xmath160 is obtained from @xmath149 by a @xmath763-enlargement , then the corresponding matrices @xmath788 and @xmath789 are related by @xmath790 since @xmath773 is congruent to the matrix @xmath791 , @xmath792 is an invariant of the link @xmath11 . now , consider the element of @xmath793 $ ] given by @xmath794 it is well - defined up to multiplication by @xmath795 , where @xmath178 denotes the genus of @xmath6 . indeed , if @xmath160 is obtained from @xmath149 by a @xmath763-enlargement , then @xmath796 since @xmath773 is a matrix of the intersection form on @xmath6 , @xmath797 , giving the result . one easily checks the following properties : if @xmath179 is odd , then @xmath798 $ ] . if @xmath179 is even , then @xmath799 $ ] . finally , @xmath800 if @xmath11 is a knot , and @xmath801 else . the first author wishes to thank the uc berkeley department of mathematics for hospitality . he also expresses his thanks to mathieu baillif . the second named author thanks research institute for mathematical sciences ( rims , kyoto ) for hospitality during the preparation of this paper . 10 d. cimasoni : _ a geometric construction of the conway potential function _ , comment . * 79 * ( 2004 ) , no . 1 , 124146 . d. cimasoni & v. florens : _ generalized seifert surfaces and signatures of colored links _ , to appear in trans . d. cooper : _ signatures of surfaces with applications to knot and link cobordism _ , ph . d. thesis , university of warwick , 1982 . d. cooper : _ the universal abelian cover of a link _ , low - dimensional topology ( bangor , 1979 ) , pp . 5166 , london math . lecture note ser . , 48 , cambridge univ . press , cambridge - new york , 1982 . r. fox & j. milnor : _ singularities of @xmath122-spheres in @xmath803-space and cobordism of knots _ , osaka j. math . * 3 * ( 1966 ) , 257267 . a. kawauchi : a survey of knot theory , birkhuser , basel , 1996 . r. lickorish : an introduction to knot theory , springer , new york , 1997 .	we extend several classical invariants of links in the 3-sphere to links in so - called quasi - cylinders . these invariants include the linking number , the seifert form , the alexander module , the alexander - conway polynomial and the murasugi - tristram - levine signatures .
[ [ background . ] ] background . + + + + + + + + + + + a simple average " @xmath17 of an observable @xmath18 @xcite does not depend on which thermodynamic ensemble it is measured in , at least not if the system is sufficiently large and if each ensemble samples indeed the same thermodynamic state point @xcite . an example for such a simple average is the affine shear - elasticity @xmath19 characterizing the energy change of an affinely shear - strained elastic body @xcite as properly defined below in sect . [ sec_algo ] . as one may verify numerically @xcite , it is irrelevant whether one computes @xmath19 in the @xmath20-ensemble at constant particle number @xmath21 , volume @xmath7 , shear strain @xmath2 and temperature @xmath22 or in the conjugated @xmath23-ensemble where the strain is allowed to fluctuate subjected to the constraint that the internal mean shear stress @xmath0 is imposed by the external applied shear stress @xmath24 @xcite . in contrast to this , the fluctuation @xmath25 of two observables @xmath18 and @xmath26 may depend on whether an extensive variable @xmath27 or its conjugated intensive variable @xmath28 is imposed @xcite . focusing in the present work on shear - strained isotropic elastic networks , as sketched in panel ( a ) of fig . [ fig_sketch ] , the relevant extensive variable is the rescaled shear strain @xmath29 and the conjugated intensive variable the shear stress @xmath30 . using the lebowitz - percus - verlet ( lpv ) transform @xcite it is seen for the ( rescaled ) mean - squared fluctuation @xmath31 of the instantaneous shear stress @xmath11 that @xcite @xmath32 with @xmath9 being the equilibrium shear modulus . for later convenience the @xmath23-ensemble is indicated by @xmath1 " and the @xmath20-ensemble by @xmath3 " . generalizing eq . ( [ eq_mutt_trans_old ] ) in the time - domain it has been shown for the stress - stress correlation function @xmath33 that @xcite @xmath34 this assumes that the shear - barostat needed to sample the @xmath23-ensemble must act very slowly on the time - scales used to probe @xmath35 . this may be realized equivalently by averaging over an ensemble of configurations with quenched strains @xmath8 distributed according to the @xmath23-ensemble @xcite . it is due to the ergodicity breaking generated by the averaging over the quenched ensemble that @xmath36 stays finite for @xmath37 , while @xmath38 must vanish for large times ( as one commonly expects for all correlation functions in ergodic systems @xcite ) . by integration by parts it is seen that @xmath35 is equivalent to the experimentally important shear relaxation modulus @xmath39 @xcite . the transform eq . ( [ eq_cttt_trans_old ] ) thus implies @xmath40 , i.e. the relaxation modulus may be measured using the equilibrium modulus @xmath9 and the correlation function @xmath38 both determined in the @xmath41-ensemble @xcite . [ [ generalized - gaussian - ensemble . ] ] generalized gaussian ensemble . + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + as sketched in panel ( b ) of fig . [ fig_sketch ] , it is straightforward to interpolate between the @xmath41-ensemble and the @xmath42-ensemble by imposing an external field @xmath43 with @xmath44 being the spring constant of the external harmonic spring . the standard @xmath23-ensemble at @xmath45 is recovered by setting @xmath46 . note that our approach is conceptually similar to the so - called gaussian ensemble " proposed some years ago by hetherington and others @xcite generalizing the boltzmann weight of the canonical ensemble by an exponential factor @xmath47 of the instantaneous energy @xmath48 . a similar external spring potential has been also used in the recent elastic bath " approach by workum and de pablo @xcite . choosing the reference strain @xmath49 equal to the mean strain @xmath50 of the @xmath23-system at vanishing shear stress @xmath51 allows to work at constant zero mean shear stress irrespective of the strength of the external potential @xcite . all the ensembles considered correspond thus to the _ same _ thermodynamic state , i.e. all first derivatives of the energy or the free energy @xcite and all simple averages are identical . as sketched in panel ( c ) of fig . [ fig_sketch ] , @xmath44 is not necessarily positive , reducing the strain fluctuations , but may even become negative , which thus amplifies the fluctuations . it has been argued @xcite that this may allow a more convenient determination of the elastic modulus . since the external spring is parallel to the system , the combined system and external device have an effective spring constant @xmath52 . defining the dimensionless parameter @xmath53 the strain fluctuations of the combined system are thus given by @xmath54 i.e. must vanish linearly with @xmath4 . @xmath42-ensemble statistics is expected for @xmath55 , while @xmath41-statistics should become relevant in the opposite limit for @xmath56 , i.e. @xmath57 . the system must become unstable in the limit @xmath58 , i.e. @xmath59 . [ [ key - results . ] ] key results . + + + + + + + + + + + + the aim of the present study is to generalize the relations for static fluctuations , eq . ( [ eq_mutt_trans_old ] ) , and dynamical correlation functions , eq . ( [ eq_cttt_trans_old ] ) , to the more general transformations between gaussian ensembles characterized by the continuous parameter @xmath60 . in this way we want to make manifest that these transformation relations are generated by the _ continuous change _ of the constraint imposed on the fluctuations of the extensive variable . focusing on spring networks in two dimensions we show , e.g. , for the stress - stress fluctuations @xmath61 in different @xmath4 ensembles that @xmath62 with @xmath63 being given by the affine shear - elasticity @xmath19 mentioned above . assuming a very slowly acting shear - barostat , which is irrelevant for the system evolution for short times @xmath64 , the above result is then seen to generalize in the time - domain for the stress - stress correlation function @xmath65 since @xmath66 , eq . ( [ eq_keydyna ] ) allows the determination of the relaxation modulus @xmath39 from the equilibrium modulus @xmath9 and the stress - stress correlation function @xmath67 for any @xmath4 and @xmath64 . the upper time limit @xmath68 is seen to be set by the diffusion time @xmath69 of the instantaneous strain @xmath70 over the typical strain fluctuations for @xmath4 . as a consequence from eq . ( [ eq_keystat ] ) and eq . ( [ eq_keydyna ] ) it appears that it is the equilibrium modulus @xmath9 of the system , which generates both transforms @xmath71 with @xmath72 and @xmath73 for @xmath1 . similar relations , allowing also to determine the creep compliance @xmath74 @xcite , are obtained for strain - strain and strain - stress correlations functions . [ [ outline . ] ] outline . + + + + + + + + the stated key relations eq . ( [ eq_keystat ] ) and eq . ( [ eq_keydyna ] ) are justified theoretically in sect . [ sec_theo ] . the static fluctuations are discussed in sect . [ theo_stat ] before we address the dynamical correlation functions in sects . [ theo_dyna_genrel ] , [ theo_dyna_model ] and [ theo_dyna_lambda ] and the macroscopic linear response in sect . [ theo_dyna_gtjt ] . algorithmic details are given in sect . [ sec_algo ] where the specific model system considered is introduced in sect . [ algo_model ] . the canonical affine plane shear transformations used in this work are specified in sect . [ algo_affine ] , the instantaneous shear stress @xmath11 and the instantaneous affine shear - elasticity @xmath75 in sect . [ algo_tau_mua ] . the zero - temperature ground state properties of the two - dimensional elastic network are summarized in sect . [ algo_refer ] . some technical details related to the finite - temperature simulation of the gaussian @xmath4-ensemble using a metropolis monte carlo ( mc ) scheme as a function of the maximum strain displacement rate @xmath76 , the second key operational parameter of this study , are given in sect . [ algo_finitet ] . section [ sec_simu ] presents the numerical results obtained by molecular dynamics ( md ) simulations at one finite ( albeit small ) temperature @xmath77 and one relatively high value of the friction constant @xmath78 of the langevin thermostat used @xcite . the relevant static properties are described in sect . [ simu_stat ] before we turn to the dynamical strain - strain ( sect . [ simu_dyna_gg ] ) , strain - stress ( sect . [ simu_dyna_gt ] ) and stress - stress ( sect . [ simu_dyna_tt ] ) correlation functions . our work is summarized in sect . [ sec_conc ] . [ [ lebowitz - percus - verlet - transform . ] ] lebowitz - percus - verlet transform . + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + we begin by restating the lpv transform in a convenient form assuming that the relevant extensive variable is the ( rescaled ) shear strain @xmath79 and the conjugated intensive variable the shear stress @xmath80 . following lebowitz _ @xcite , one verifies ( see also sect . ii.a of ref . @xcite ) that to leading order @xmath81 for the transformation of @xmath25 with @xmath82 and @xmath83 from the @xmath20-ensemble ( @xmath3 ) to the @xmath23-ensemble ( @xmath1 ) . the more general transformation between arbitrary @xmath4-ensembles can be found by reworking the saddle - point approximation of ref . @xcite taking into account that the fluctuations around the peak of the distribution of the extensive variable is now not set by the modulus @xmath9 of the system but by the effective modulus @xmath52 . how this may be done can be seen in sect . 2.5 of ref . @xcite for the volume @xmath84 as extensive variable and the ( negative ) pressure @xmath85 as intensive variable . rewriting the latter result to the present case we get the generalized lpv transform @xmath86 which reduces for @xmath1 to eq . ( [ eq_lpvold ] ) . [ [ strain - strain - fluctuations . ] ] strain - strain fluctuations . + + + + + + + + + + + + + + + + + + + + + + + + + + + let us check the ensemble dependence of the rescaled strain - strain fluctuations @xmath87 @xcite . the generalized gaussian ensemble corresponds to replacing the shear modulus @xmath9 of the system by the effective modulus @xmath88 . as already stated above , eq . ( [ eq_dgam_l ] ) , this leads to @xcite @xmath89 one verifies readily that the postulated lpv transform for general @xmath4 , eq . ( [ eq_lpvgen ] ) , is consistent with this result . to see this one sets @xmath90 and uses the fact that the strain fluctuations must vanish in the @xmath20-ensemble , i.e. that the first term on the right hand - side of eq . ( [ eq_lpvgen ] ) must vanish . [ [ strain - stress - fluctuations . ] ] strain - stress fluctuations . + + + + + + + + + + + + + + + + + + + + + + + + + + + setting @xmath91 and @xmath92 and using again that the @xmath20-term in eq . ( [ eq_lpvgen ] ) must vanish , it is seen from the lpv transform that the strain - stress fluctuations @xmath93 @xcite should scale as @xmath94 this result can be also obtained directly by replacing in the definition of @xmath95 the fluctuation @xmath96 by @xmath97 and using then the strain - strain relation , eq . ( [ eq_mugg_lambda ] ) . [ [ stress - stress - fluctuations . ] ] stress - stress fluctuations . + + + + + + + + + + + + + + + + + + + + + + + + + + + we turn to the most important stress - stress fluctuation @xmath10 @xcite . we set now @xmath98 in the lpv transform . since the stress fluctuations do _ not _ vanish in the @xmath20-ensemble the corresponding term must now be included . this yields @xcite @xmath99 interestingly , the contribution @xmath100 may be rewritten using the notation @xmath101 for the strain - average of a property @xmath102 with @xmath103 being the normalized strain - distribution for the considered @xmath4-ensemble . the total mean stress @xmath0 of the ensemble is thus given by @xmath104 with @xmath105 being the average shear - stress of all configurations of shear - strain @xmath8 . using that @xmath103 is a gaussian and that @xmath106 with @xmath50 being the maximum of the distribution , it is then seen that @xmath107 using eq . ( [ eq_mutt_lambda_one ] ) this implies in turn @xmath108 as one expects to lowest order from a standard saddle - point approximation . [ [ stress - fluctuation - formula - for - shear - modulus . ] ] stress - fluctuation formula for shear modulus . + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + substracting the transform for @xmath1 from eq . ( [ eq_mutt_lambda_one ] ) confirms the key result eq . ( [ eq_keystat ] ) . as shown in ref . @xcite , @xmath63 can be reduced by integration by parts to the affine shear - elasticity @xmath19 . since the latter expression is a simple average , i.e. the same value @xmath19 is obtained in any ensemble , eq . ( [ eq_keystat ] ) may be further simplified as @xcite @xmath109 it is thus sufficient to compute the material constants @xmath19 and @xmath9 in any ensemble to obtain the stress - stress fluctuations @xmath61 as a function of @xmath4 . note that the special case for @xmath3 corresponds to the well - known stress - fluctuation formula @xcite @xmath110 used in numerous numerical studies to compute the modulus @xmath9 conveniently in the @xmath20-ensemble @xcite . [ [ introduction . ] ] introduction . + + + + + + + + + + + + + we define now several dynamical observables of interest , which will be investigated numerically in sect . [ sec_simu ] , and remind some well - known general relations @xcite . time translational symmetry @xmath111 and time reversal symmetry @xmath112 are assumed . we use @xmath113 and @xmath114 for either the instantaneous shear strain @xmath8 or the shear stress @xmath11 , @xmath115 and @xmath116 for their thermodynamic averages and @xmath117 and @xmath118 for their time - dependent fluctuations . [ [ mean - square - displacements . ] ] mean - square displacements . + + + + + + + + + + + + + + + + + + + + + + + + + + we define ( generalized ) mean - square displacements ( msd ) by @xmath119 where the prefactor @xmath120 has been introduced for convenience @xcite . obviously , @xmath121 , @xmath122 as may be seen using the above - mentioned symmetries , @xmath123 for @xmath124 and @xmath125 with @xmath126 being the characteristic time needed to reach this thermodynamic limit @xcite . [ [ correlation - functions . ] ] correlation functions . + + + + + + + + + + + + + + + + + + + + + + similarly , we define dynamic correlation functions by @xmath127 @xcite . obviously , @xmath128 , @xmath129 and @xmath130 . the latter identity is seen from @xcite @xmath131 where the time translational invariance is used in the first step and the time reversal symmetry in the second step . using again both symmetries one verifies that @xcite @xmath132 this implies that @xmath133 for large times @xmath134 @xcite . albeit @xmath135 and @xmath136 thus contain the same information it will be sometimes better for theoretical or numerical reasons to focus on either @xmath135 or @xmath136 . since @xmath123 for @xmath124 it will be natural , e.g. , to consider the double - logarithmic representation of @xmath135 to clarify the power - law scaling of the correlations at short times . [ [ finite - time - dependence - of - fluctuation - estimation . ] ] finite - time dependence of fluctuation estimation . + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + as defined in sect . [ theo_stat ] , @xmath137 is a thermodynamic ensemble average , hence , a time - independent property . in practice , @xmath137 may , however , often be estimated using a time series @xmath138 with a finite number @xmath139 of more or less correlated entries @xcite . it is supposed here that these series are sampled with a constant time interval of length @xmath140 , i.e. @xmath139 corresponds to a time @xmath141 . with @xmath142 denoting such a finite average , one samples the @xmath139-dependent observable @xcite @xmath143 with @xmath144 standing for an additional ensemble average over different time series of @xmath139 subsequent data points . we have @xmath145 for @xmath146 and @xmath147 for @xmath148 in an ergodic system . interestingly , the detailed time - dependence of @xmath149 can be obtained from a weighted integral of the correlation function @xmath136 @xcite . to see this we note first the identity @xmath150 which can be verified by straightforward algebra . ( see sect . 2.4 of ref . using time translational invariance this implies @xmath151 where the weight @xmath152 stems from the finite length of the trajectory used . we change now to continuous time variables , @xmath153 , and replace the discrete sum by the time integral @xmath154 using eq . ( [ eq_gabt_cabt ] ) this yields the general relation @xmath155 which can be used to obtain @xmath156 if the correlation function @xmath136 is known . [ [ relaxation - time - ensuremaththeta_mathrmab . ] ] relaxation time @xmath157 . + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + interestingly , the time integral in eq . ( [ eq_muabt_cabt ] ) must become constant in all cases considered below where @xmath136 vanishes ultimately for @xmath37 . ( this applies if a finite shear - barostat is switched on . ) defining the characteristic relaxation time @xmath158 this leads to @xmath159 as one expects for the poisson statistics of uncorrelated events @xcite . one may thus determine the relaxation time @xmath157 from the large - time asymptotics of @xmath156 @xcite . let us suppose that the msd @xmath135 is diffusive for short times , i.e. @xmath160 for @xmath161 with @xmath162 being a diffusion constant . matching this short time regime with @xmath163 for @xmath134 gives the possibility to operationally _ define _ @xmath126 as the crossover time @xmath164 different short - time dynamics may suggest of course a different operational definition . let us further assume an exponentially decaying correlation function @xmath165 with @xmath166 . using eq . ( [ eq_gabt_cabt ] ) this is seen to be consistent with a diffusive short - time regime and eq . ( [ eq_tabdab ] ) . it follows by integration from eq . ( [ eq_muabt_cabt ] ) that @xmath167 \label{eq_debye}\ ] ] using the debye function well - known in polymer science @xcite . for large reduced times @xmath168 this leads to @xmath169 , i.e. by comparison with eq . ( [ eq_muabt_asymp ] ) we have @xmath170 for an exponentially decaying correlation function . as we shall see in sect . [ simu_dyna_tt ] , it may also occur that the correlation function is more or less constant between a local ( barostat independent time ) @xmath171 up to a ( barostat dependent ) time @xmath172 , i.e. essentially @xmath173 with @xmath174 being a constant and @xmath175 the heaviside function . it follows from eq . ( [ eq_thetaab_def ] ) that @xmath176 i.e. @xmath177 to leading order . [ [ introduction.-1 ] ] introduction . + + + + + + + + + + + + + we have just stated various general relations applying to all ensembles . these relations do , however , not allow to relate the dynamical correlations between different @xmath4 . since some ensembles are more readily computed than others , it would be useful to have a transformation relation such as the lpv transform for static fluctuations considered in sect . [ theo_stat ] . we emphasize that the correlation functions depend in general on the ensemble , i.e. the parameter @xmath4 , and on the dynamics of the shear - barostat . since in the end we want to describe the intrinsic relaxation dynamics of the system , it is sufficient to focus on the limit where the barostat becomes very slow such that it becomes essentially irrelevant for the system evolution below an upper time @xmath172 . we consider the limit where this time @xmath172 is be much larger than any intrinsic relaxation time of the system . since we want still to consider meaningful thermal averages with respect to the chosen ensemble , we need either trajectories of a time interval @xmath178 much larger than @xmath172 to allow the full sampling of the phase space or , equivalently , we need to average over independent start configurations representing the ensemble . under these two assumptions useful transformation relations can be formulated by reworking the generalization of the lpv transform for dynamical correlations replacing the system modulus @xmath9 by the effective modulus @xmath179 . being beyond the scope of this paper , we proceed by postulating the central scaling relation for the msd @xmath135 and argue briefly why this relation is natural . we discuss then in turn the strain - strain correlations @xmath180 , the strain - stress correlations @xmath181 and the stress - stress correlations @xmath182 for different @xmath4 . alternative , more direct ways to derive the relations are mentioned . [ [ scaling - relations . ] ] scaling relations . + + + + + + + + + + + + + + + + + + we postulate that under the two assumptions made above the msd @xmath135 does not depend on the ensemble @xmath183 i.e. the msd behaves as a simple mean . this scaling postulate is justified by two facts . firstly , the barostat is ( by construction ) too weak to change for @xmath184 the evolution of the system . secondly , that the starting points of the trajectory at @xmath185 are distributed according to the considered ensemble must become an irrelevant higher order effect ( vanishing rapidly with the system volume @xmath7 ) , since the msd probes displacements with respect to the starting points , not their absolute values . the fundamental scaling , eq . ( [ eq_msdscaling ] ) , implies using eq . ( [ eq_gabt_cabt ] ) that @xmath186 where only the first term on the right hand - side depends on @xmath4 . this leads us finally to the general transformation relation between correlation functions @xmath187 where @xmath3 stands for the @xmath20-ensemble with @xmath188 the two static contributions on the right hand - side can be further simplified using results from sect . [ theo_stat ] . this demonstrates the linear relation @xmath189 . [ [ strain - strain - correlations . ] ] strain - strain correlations . + + + + + + + + + + + + + + + + + + + + + + + + + + + since @xmath190 for @xmath191 , the msd @xmath192 must vanish to leading order . using eq . ( [ eq_ctscaling ] ) this implies @xcite @xmath193 this result is directly obtained by setting @xmath194 in the definition of @xmath180 . we emphasize that it is also consistent with the lpv transform , eq . ( [ eq_lpvgen ] ) , as one verifies by setting @xmath80 , @xmath195 , @xmath196 and @xmath197 and using that the strain fluctuation term for @xmath3 must vanish . [ [ strain - stress - correlations . ] ] strain - stress correlations . + + + + + + + + + + + + + + + + + + + + + + + + + + + to determine the strain - stress correlation function @xmath181 one may use again that the msd @xmath198 must vanish since @xmath190 for @xmath184 . using eq . ( [ eq_ctscaling ] ) this implies @xcite @xmath199 this result is also obtained from the lpv transform setting @xmath196 and @xmath200 and using again that @xmath201 . the postulated scaling eq . ( [ eq_msdscaling ] ) has thus led again to a reasonable result . [ [ stress - stress - correlations . ] ] stress - stress correlations . + + + + + + + + + + + + + + + + + + + + + + + + + + + interestingly , as one may see from the @xmath20-ensemble limit considered in @xcite , the stress - stress msd @xmath202 is not expected to simply vanish for @xmath184 as in the two previous cases . ( the instantaneous stress @xmath11 fluctuates even at a fixed strain @xmath8 . ) however , eq . ( [ eq_ctlambda ] ) still holds leading to @xcite @xmath203 where the static term on the right hand - side has been simplified using eq . ( [ eq_keystat ] ) . this confirms the key relation eq . ( [ eq_keydyna ] ) announced in the introduction . note that the correlation function @xmath38 in the @xmath20-ensemble must vanish beyond some local time scale @xmath171 which does depend on the network considered but , of course , not on the shear - barostat . for a sufficiently slow barostat the correlation function thus becomes constant @xmath204 where we have dropped @xmath205 . using eq . ( [ eq_ctscaling ] ) this leads to the remarkable relation @xmath206 which must hold for all @xmath60 . we note finally that while for @xmath207 the msd @xmath208 for @xmath209 , @xmath210 holds for all times @xmath211 for @xmath3 according to stress - fluctuation formula , eq . ( [ eq_gtt ] ) . [ [ introduction.-2 ] ] introduction . + + + + + + + + + + + + + the experimentally important macroscopic linear response measured by the creep compliance @xmath74 and the shear relaxation modulus @xmath39 @xcite may be obtained conveniently in an equilibrium simulation at a given @xmath4 using some of the correlation functions discussed above . please note that being material properties of the given state point ( experimentally obtained using a simple average , not a fluctuation ) both response functions @xmath74 and @xmath39 do , of course , not depend on @xmath4 . [ [ creep - compliance . ] ] creep compliance . + + + + + + + + + + + + + + + + + let us first consider the creep compliance @xmath212 for @xmath213 . it is assumed that for @xmath214 the system is at thermal equilibrium and the internal mean stress @xmath0 equals the applied external stress @xmath24 of the @xmath23-ensemble . after imposing at @xmath215 a small increment @xmath216 , the creep compliance @xmath74 measures the ensuing average strain increment @xmath217 . we note _ en passant _ that the average internal shear stress @xmath218 does neither immediately reach the new equilibrium value @xmath219 but shows a similar time dependence as the strain . reworking the arguments put forward by doi and edwards , ( 3.67 ) of ref . @xcite , it is seen that @xmath220 i.e. the creep compliance is most readily computed using the strain - strain msd @xmath192 in the @xmath23-ensemble . as described in sect . [ algo_finitet ] , we shall change the shear - strain @xmath70 using a mc shear - barostat which corresponds to a perfectly viscous dynamics . one thus expects @xmath221 for the strain - strain correlations . using @xmath222 for @xmath1 and eq . ( [ eq_gabt_cabt ] ) this suggest @xmath223 \label{eq_kelvinvoigt}\ ] ] in agreement with the kelvin - voigt model @xcite representing a purely viscous damper and a purely elastic spring connected in parallel . [ [ shear - relaxation - modulus . ] ] shear relaxation modulus . + + + + + + + + + + + + + + + + + + + + + + + + + the shear relaxation modulus @xmath224 may be obtained from the stress increment @xmath225 for @xmath226 after a small step strain with @xmath227 has been imposed at time @xmath185 . it is well known that the components of the fourier transformed relaxation modulus @xmath39 , the storage modulus @xmath228 and the loss modulus @xmath229 , are directly measurable in an oscillatory shear strain experiment @xcite . as seen , e.g. , by eq . ( 32 ) in ref . @xcite , it can be demonstrated by integration by parts that @xmath230 i.e. the barostat should be irrelevant on the time scales considered . using the transformation relation between different ensembles , eq . ( [ eq_cttt_scaling ] ) , the relaxation modulus may equivalently be obtained for other @xmath4 according to @xmath231 since for large times @xmath232 , this implies @xmath233 consistently with eq . ( [ eq_cttt_plateau ] ) . for the specific case @xmath3 eq . ( [ eq_gt_lambda ] ) yields @xmath234 which holds for all times @xmath235 since the barostat becomes irrelevant for @xmath57 . note that eq . ( [ eq_gt_nvgt ] ) may also be derived directly ( without using the transformation relation between different ensembles ) using boltzmann s superposition principle for an arbitrary strain history and the standard fluctuation - dissipation theorem for the after - effect function @xcite as shown elsewhere @xcite . two immediate consequences of eq . ( [ eq_gt_nvgt ] ) are _ ( i ) _ that @xmath39 only becomes equivalent to @xmath38 for @xmath226 in the liquid limit where ( trivially ) @xmath236 and _ ( ii ) _ that the shear modulus @xmath9 is only probed by @xmath39 on time scales @xmath211 where @xmath38 must vanish . in principle , it is thus _ impossible _ to obtain the static shear modulus @xmath9 of an elastic body only from @xmath38 as often incorrectly assumed @xcite . to illustrate our key relations we present in sect . [ sec_simu ] numerical data obtained using a periodic two - dimensional ( @xmath237 ) network of @xmath238 harmonic springs connecting @xmath21 vertices . the model hamiltonian is given by the sum @xmath239 of a kinetic energy contribution @xmath240 with @xmath241 being the velocity of vertex @xmath242 and @xmath243 its ( assumed ) monodisperse mass , and an excess potential @xmath244 where @xmath245 denotes the spring constant , @xmath246 the reference length and @xmath247 the length of spring @xmath248 . the sum runs over all springs @xmath248 connecting pairs of beads @xmath242 and @xmath249 with @xmath250 at positions @xmath251 and @xmath252 . the vertex mass @xmath243 and boltzmann s constant @xmath253 are set to unity and lennard - jones ( lj ) units @xcite are assumed . while the box volume @xmath7 is kept constant throughout this work , we shall frequently change the shape of the simulation box . as sketched in panel ( a ) of fig . [ fig_sketch ] , we perform plane shear transformations of the instantaneous shear strain @xmath254 with an essentially infinitesimal strain increment @xmath255 . we assume that not only the box shape is changed but that the particle positions @xmath256 ( using the principal box convention @xcite ) follow the imposed macroscopic constraint in an _ affine _ manner according to @xmath257 with all other coordinates remaining unchanged . albeit not strictly necessary for the demonstration of our key relations , we assume , moreover , that this shear transformation is also _ canonical _ @xcite . this implies that the @xmath258-component of the velocity must transform as @xcite @xmath259 we emphasize the negative sign in eq . ( [ eq_p_cantrans ] ) which assures that liouville s theorem is obeyed @xcite . let @xmath260 denote the system hamiltonian of a configuration originally at @xmath8 strained using a canonical affine transformation to @xmath261 compactly written as a function of the strain increment @xmath255 . the instantaneous shear stress @xmath11 and the instantaneous affine shear - elasticity @xmath75 may be defined as the expansion coefficients associated to the energy change @xmath262 with @xmath8 being the reference , i.e. @xcite @xmath263 the derivatives @xmath264 and @xmath265 with respect to @xmath255 may be computed as shown in sect . 2.1 of @xcite . similar relations apply for the corresponding contributions @xmath266 and @xmath267 to @xmath268 and for the contributions @xmath269 and @xmath270 to @xmath271 . using eq . ( [ eq_eid ] ) this implies @xcite @xmath272 for the ideal contributions to the shear stress and the affine shear - elasticity . note that the minus sign for the shear stress is due to the minus sign in eq . ( [ eq_p_cantrans ] ) required for a canonical transformation . for the excess contributions one obtains @xcite @xmath273 with @xmath274 being the normalized distance vector @xmath275 between the particles @xmath242 and @xmath249 . interestingly , eq . ( [ eq_tauexhat ] ) is strictly identical to the corresponding off - diagonal term of the kirkwood stress tensor @xcite . the last term in eq . ( [ eq_muaexhat ] ) automatically takes into account the finite normal pressure of the system . [ [ specific - network . ] ] specific network . + + + + + + + + + + + + + + + + + as explained elsewhere @xcite , the specific elastic network used in this work has been constructed using the dynamical matrix of a quenched polydisperse lj bead glass , i.e. at low temperatures our network has exactly the same mechanical and vibrational properties as the original discrete particle system . prior to forming the network the bead system was cooled down to @xmath276 using a constant quenching rate and imposing a normal pressure @xmath277 . the original lj beads are represented in the snapshot shown in fig . [ fig_algo ] by grey polydisperse circles , the permanent spring network created from the quenched bead system by lines between vertices . the dark ( black ) lines indicate repulsive forces between the vertices , while the light ( red ) lines represent tensile forces . the line width is proportional to the tension ( repulsion ) . note that the force network is strongly inhomogeneous with zones of weak attractive links embedded within a strong repulsive skeleton as already discussed in refs . only a small subvolume of the network is represented . the total periodic box of linear length @xmath278 contains @xmath279 vertices and @xmath280 springs . the monomer density @xmath281 is close to unity . [ [ finite - shear - stress - tau_0 . ] ] finite shear stress @xmath282 . + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + due to the construction of the network the total force acting on each vertex of the reference network must vanish at @xmath276 . as seen in the snapshot , this does not imply that the repulsive and/or tensile forces transmitted along the springs must also vanish . due to the periodic boundary conditions and the constant - strain constraint ( @xmath283 ) the shear stress @xmath0 does not necessarily vanish . for a pair potential such as eq . ( [ eq_enet ] ) the relevant excess contribution @xmath267 of the shear stress is readily computed using the kirkwook expression , eq . ( [ eq_tauexhat ] ) . as shown in the main panel of fig . [ fig_algo ] ( horizontal arrow ) , it turns out that for the specific network we use throughout this work we have a finite shear stress @xmath284 . [ [ affine - shear - elasticity - ensuremathmu_mathrma . ] ] affine shear - elasticity @xmath19 . + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + let us consider a small affine shear strain according to eq . ( [ eq_r_cantrans ] ) . as show in panel ( b ) of fig . [ fig_algo ] , one may now compute using eq . ( [ eq_tauexhat ] ) the shear stress @xmath285 for different @xmath2 ( squares ) . as one expects from the definition of the affine shear - elasticity coefficient @xmath19 , eq . ( [ eq_haffine ] ) , this yields the linear relation @xmath286 with @xmath287 as indicated by the dashed line . the same coefficient @xmath19 is also obtained directly from the unstrained configuration using eq . ( [ eq_muaexhat ] ) . [ [ equilibrium - shear - modulus - g_mathrmeq . ] ] equilibrium shear modulus @xmath9 . + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + the forces @xmath288 acting on the vertices @xmath242 of an _ affinely _ strained network do not vanish in general and the system is normally _ not _ at mechanical equilibrium . as described elsewhere @xcite , we relax these forces by first applying a steepest descend algorithm , i.e. by imposing displacements proportional to the force , and then by means of the conjugate gradient method @xcite . the ensuing non - affine displacements of the vertices decrease the energy @xcite and the magnitude of the shear stress @xmath285 as may be seen from the large circles indicated in fig . [ fig_algo ] . as indicated by the bold solid line these final stresses @xmath285 scale again linearly as @xmath289 with @xmath290 and @xmath291 . the affine coefficient @xmath19 has thus been replaced by the much smaller equilibrium shear modulus @xmath9 of the groundstate network . note that shear stress vanishes at a strain @xmath292 as indicated by the vertical arrow . if the strain @xmath2 is allowed to change freely ( e.g. , using a steepest descend scheme ) without any external force @xmath24 applied , the system relaxes to @xmath293 . [ [ molecular - dynamics - scheme . ] ] molecular dynamics scheme . + + + + + + + + + + + + + + + + + + + + + + + + + + as discussed in sect . [ sec_simu ] , this network is investigated numerically basically by means of a molecular dynamics ( md ) simulation @xcite at constant particle number @xmath279 , box volume @xmath294 and a small , but finite mean temperature @xmath295 . newton s equations are integrated using a velocity - verlet algorithm with a tiny time step @xmath296 . the temperature @xmath77 is fixed using a langevin thermostat with a relatively large friction constant @xmath297 . this was done to suppress artificial long - range correlations in the two - dimensional periodic simulation box . [ [ thermal - averages . ] ] thermal averages . + + + + + + + + + + + + + + + + + using the measured instantaneous shear stress @xmath11 and affine shear - elasticity @xmath75 we compute , e.g. , the thermal averages @xmath298 , @xmath10 and @xmath299 . it can be shown for the respective ideal contributions that @xcite @xmath300 with @xmath301 being the ideal normal pressure contribution . this holds irrespective of the considered @xmath4-ensemble . the ideal contributions are thus negligible . due to the equivalence of the different axes it is also seen for the excess contributions that @xcite @xmath302 being the born - lam coefficient @xcite and @xmath303 the excess part of the normal pressure @xmath304 . .summary of some properties as a function of the external spring constant @xmath44 : @xmath305 with @xmath306 , strain - strain fluctuation @xmath307 ( fig . [ fig_dgam ] ) , stress - stress fluctuation @xmath61 ( fig . [ fig_muf_l ] ) , acceptance rate @xmath308 of mc shear - barostat ( fig . [ fig_aeta ] ) , @xmath309-parameter defined by eq . ( [ eq_eta_def ] ) , strain - strain crossover time @xmath310 ( fig . [ fig_gggt ] ) , strain - stress crossover time @xmath311 ( fig . [ fig_ggtt ] ) , stress - stress relaxation time @xmath312 ( fig . [ fig_thetatt ] ) . the dynamical properties ( columns 5 - 9 ) are only given for @xmath313 . the time scales @xmath69 , @xmath314 and @xmath312 should be compared to the intrinsic relaxation time @xmath315 of the stress - stress correlations for @xmath316 ( fig . [ fig_gttt ] ) . [ tab ] [ cols="^,^,^,^,^,^,^,^,^",options="header " , ] [ [ monte - carlo - shear - barostat . ] ] monte carlo shear - barostat . + + + + + + + + + + + + + + + + + + + + + + + + + + + the md algorithm for the particles is coupled for @xmath317 with a monte carlo ( mc ) scheme @xcite attempting every md time step a canonical affine strain increment @xmath318 ( sect . [ algo_affine ] ) . first , a strain increment @xmath255 is randomly chosen from a uniformly distributed interval @xmath319 $ ] . in order to determine the metropolis weight @xcite we compute next the energy change @xmath320 of the network which comprises both an excess contribution due to eq . ( [ eq_r_cantrans ] ) and an ideal contribution due to eq . ( [ eq_p_cantrans ] ) . since our system is subjected to an external field @xmath321 , eq . ( [ eq_uext ] ) , this gives rise to an additional contribution @xmath322 . the suggested strain move @xmath255 is accepted if @xmath323 \label{eq_metropol}\ ] ] with @xmath324 being a uniformly distributed random variable with @xmath325 @xcite . [ [ operational - parameters - lambda - and - kappa . ] ] operational parameters @xmath4 and @xmath76 . + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + we assume for the external field @xmath321 that @xmath326 and @xmath327 , i.e. the average shear stress @xmath0 is imposed to vanish for all ensembles studied and all ensembles compare the same thermodynamic state point . ( this was explicitly checked . ) the only remaining operational parameter from the static point of view is the external spring constant @xmath44 or , equivalently , @xmath328 as sketched in panel ( c ) of fig . [ fig_sketch ] . as seen in table [ tab ] or fig . [ fig_aeta ] , we vary @xmath44 from @xmath329 , i.e. @xmath330 , over @xmath331 ( @xmath23-ensemble ) up to @xmath332 , i.e. @xmath333 . the latter case is essentially equivalent to the standard @xmath20-ensemble , i.e. the strain fluctuations become irrelevant for most properties . the second operational parameter of this study is the maximum attempted strain displacement @xmath334 which determines the impact of the shear - barostat on the relaxation dynamics . since the metropolis mc move for the strain is performed every md time step of length @xmath335 , it is convenient to use instead of @xmath334 the maximum strain increment _ rate _ @xmath336 . ( since all simulations are performed with the same @xmath335 , @xmath334 and @xmath76 are strictly equivalent . ) we compare below the dynamical strain and stress correlations for the five rates @xmath337 and @xmath338 for a broad range of @xmath4 . for @xmath339 we have computed in addition the values @xmath340 and @xmath341 . some properties for @xmath313 are summarized in the columns 5 - 9 of table [ tab ] . the simulations become increasingly time consuming with decreasing @xmath76 and data obtained for @xmath342 have to be taken with care . [ [ free - strain - diffusion - limit . ] ] free strain diffusion limit . + + + + + + + + + + + + + + + + + + + + + + + + + + + + the impact of @xmath76 can be best judged from the simple limit where neither the system nor the external field restricts the barostat . this is the case for sufficiently small @xmath76 depending on @xmath4 . in this limit one expects the free diffusion of the strain @xmath70 , i.e. @xmath343 for @xmath344 . since the ( attempted and accepted ) strain increment @xmath255 is uniformly distributed in @xmath319 $ ] , we have a mean - square strain step @xmath345 every @xmath335 . this implies a diffusion constant @xmath346 in the free - diffusion limit . using eq . ( [ eq_tabdab ] ) and eq . ( [ eq_dgam_l ] ) this yields in turn the corresponding crossover time @xmath347 where we have defined the dimensionless variable @xmath348 since @xmath9 and @xmath335 are kept constant in all our simulations , the parameter @xmath349 determines the dynamical regime for a system of operational parameters @xmath4 and @xmath76 . as seen in fig . [ fig_aeta ] using @xmath349 as a scaling variable the acceptance rate @xmath308 of the mc shear - barostat of a broad range of @xmath4 and @xmath76 can be brought to collapse . for @xmath350 the mc barostat is most efficient for static properties . for dynamical properties we shall focus below on @xmath76-values where @xmath351 and thus @xmath352 as emphasized by the solid horizontal line [ [ strain - strain - fluctuations.-1 ] ] strain - strain fluctuations . + + + + + + + + + + + + + + + + + + + + + + + + + + + the equilibrium fluctuations of the strain @xmath8 at a finite temperature @xmath295 for different values of @xmath44 or , equivalently , @xmath305 with @xmath353 are presented in fig . [ fig_histol ] and fig . [ fig_dgam ] . the given data refer to a maximum strain rate @xmath354 for the metropolis mc step @xmath254 attempted after each md time step of @xmath355 ( sect . [ algo_finitet ] ) . that this leads indeed to the desired gaussian distributions can be seen from fig . [ fig_histol ] where the normalized histogram @xmath356 is plotted as a function of @xmath357 with @xmath358 . as one expects , @xmath359 for @xmath57 . the rescaled second moment @xmath360 of the distribution is further analyzed in fig . [ fig_dgam ] . as one may see from panel ( a ) , @xmath361 decreases linearly with @xmath4 ( bold solid line ) and vanishes for @xmath362 as it should according to eq . ( [ eq_mugg_lambda ] ) . in turn this implies that one may determine @xmath9 from @xmath363 as can be seen from panel ( b ) of fig . [ fig_dgam ] , using this relation one confirms that the groundstate value @xmath290 remains unchanged at a finite temperature @xmath364 . the accurate determination of @xmath365 becomes of course more delicate with increasing @xmath4 since the mc acceptance rate becomes eventually too small . the data quality in this limit can be readily improved ( not shown ) by either increasing the sampling time @xmath178 or by decreasing @xmath366 . ( the sampling becomes again inefficient if @xmath349 gets too small . ) [ [ strain - stress - correlations.-1 ] ] strain - stress correlations . + + + + + + + + + + + + + + + + + + + + + + + + + + + since the shear modulus @xmath9 is finite , this implies that shear strain and shear stress fluctuations must be correlated @xcite . these correlations are addressed in fig . [ fig_tau_gam ] . the main panel presents the pre - averaged stress @xmath367 for an instantaneous strain @xmath368 for several @xmath4 @xcite . irrespective of @xmath258 or @xmath4 all data collapse on the linear slope indicated by the bold line with @xmath306 . naturally , it follows from fig . [ fig_histol ] that the statistics must strongly decrease for @xmath369 , i.e. a larger linear - slope window is visible in fig . [ fig_tau_gam ] for smaller @xmath4 . panel ( b ) of fig . [ fig_tau_gam ] presents two scatter plots of @xmath370 for @xmath371 ( @xmath372 ) . the small circles correspond to a time series of 100 subsequent and , hence , strongly correlated configurations with @xmath373 . on these time scales the configuration has no time to relax the affine displacements imposed by the shear - barostat . as emphasized by the dashed slope this leads to a linear stress - strain relation with a coefficient @xmath374 . other short - time sequences yield similar , but horizontally shifted linear slopes ( not shown ) . a @xmath140-independent representation of the static strain - stress correlations for large times is obtained if @xmath370 is indicated for 100 configurations with a time interval @xmath375 ( squares ) . the scatter plot of the latter data is already in nice agreement with a coefficient @xmath306 ( bold line ) . sampling over _ all _ data tuples @xmath370 of a given @xmath4-ensemble one verifies that the shear modulus @xmath9 is accurately determined using the linear regression coefficient @xmath376 since , as shown in sect . [ theo_stat ] , @xmath377 and @xmath378 , this ratio must yield @xmath9 for all @xmath317 . confirming @xmath306 the values of @xmath379 for different @xmath4 are identical to @xmath365 as shown in fig . [ fig_muf_l ] . [ [ stress - stress - fluctuations.-1 ] ] stress - stress fluctuations . + + + + + + + + + + + + + + + + + + + + + + + + + + + the stress - stress fluctuations @xmath380 presented in fig . [ fig_muf_l ] ( circles ) decrease linearly with @xmath4 as stated in the introduction , eq . ( [ eq_keystat ] ) . being a simple mean @xmath19 ( small squares ) is found to be strictly @xmath4-independent . as expected , @xmath381 for @xmath55 . using eq . ( [ eq_keystat ] ) the generalized stress - fluctuation formula for @xmath9 reads @xmath382 for @xmath57 , i.e. @xmath56 , this reduces to eq . ( [ eq_gtt ] ) . equation ( [ eq_gttgen ] ) thus generalizes the stress - fluctuation formula for the @xmath20-ensemble @xcite to general @xmath4 . as seen from fig . [ fig_muf_l ] ( diamonds ) , it is thus possible to determine @xmath9 from @xmath19 and @xmath61 for all @xmath383 . [ [ strain - strain - msd . ] ] strain - strain msd . + + + + + + + + + + + + + + + + + + the strain histograms @xmath356 for different @xmath4 and their second moments shown , respectively , in fig . [ fig_histol ] and fig . [ fig_dgam ] have been rapidly sampled by averaging over short time series from trajectories of length @xmath384 . this is demonstrated in panel ( c ) of fig . [ fig_dgam ] presenting the msd @xmath192 of the instantaneous strain @xmath70 for @xmath385 . the statistics is improved by performing a gliding average over the @xmath386 data tuples sampled @xcite . it is seen that @xmath192 converges rapidly after a time @xmath69 of order unity to its asymptotic limit @xmath387 indicated by horizontal lines . free diffusion is observed for short times , i.e. @xmath388 for @xmath344 ( dash - dotted line ) , and this essentially irrespective of the ensemble in agreement with eq . ( [ eq_dgg_fdl ] ) . note that free strain diffusion is observed for all @xmath4 and sufficiently small @xmath76 if the parameter @xmath389 . as shown in the main panel of fig . [ fig_gggt ] for different @xmath76 and @xmath1 ( open symbols ) , this behavior may be used to determine first the short - time diffusion coefficient @xmath390 for systems with sufficiently small @xmath349 and then using eq . ( [ eq_tabdab ] ) the crossover time @xmath69 of the strain fluctuations . the rescaled times @xmath391 are represented in panel ( b ) of fig . [ fig_gggt ] as a function of the scaling variable @xmath349 for several @xmath4 . the dash - dotted line indicates the power - law slope , eq . ( [ eq_tgg_fdl ] ) . a perfect data collapse on this line is observed for all @xmath392 . a successful data collapse can be also obtained for @xmath192 over a broad range of @xmath4 and @xmath76 by plotting @xmath393 as a function of the reduced time @xmath394 ( not shown ) . as seen by the thin solid line in the main panel of fig . [ fig_gggt ] , one verifies that the scaling function @xmath395 is given by @xmath396 . this is , of course , consistent with an exponentially decaying strain - strain relaxation function @xmath397 as we have also checked directly . [ [ creep - compliance - jt . ] ] creep compliance @xmath74 . + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + as discussed in sect . [ theo_dyna_gtjt ] , a msd @xmath192 computed in the @xmath23-ensemble corresponds to a creep compliance @xmath398 measured by imposing at @xmath185 a step stress increment @xmath399 to the external shear stress @xmath24 applied to the system . using @xmath400 and averaging over 100 configurations this yields the data shown by the filled symbols in panel ( a ) of fig . [ fig_gggt ] for two values of @xmath76 . an excellent data collapse @xmath401 is observed . while it is natural to use the @xmath42-ensemble for obtaining the creep compliance from the equilibrium fluctuations , it is worthwhile to note that due to the scaling @xmath402 , @xmath74 may be also determined from other @xmath4-values . [ [ time - dependent - strain - strain - fluctuations . ] ] time - dependent strain - strain fluctuations . + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + as expected from sect . [ theo_dyna_genrel ] , the strain - strain fluctuations @xmath403 computed according to eq . ( [ eq_muabn_two ] ) by gliding average from a finite time series increase monotonously with time @xmath235 from zero ( if only one configuration is sampled ) to the asymptotic thermodynamic limit @xmath307 ( not shown ) . note that one determines first the strain fluctuations in an interval @xmath404 $ ] and averages then over all possible @xmath405 . if the estimate @xmath365 , eq . ( [ eq_ggggen ] ) , of the modulus @xmath9 is determined using @xmath403 instead of @xmath307 , it becomes a time dependent quantity called @xmath406 . as can be seen from fig . [ fig_geq_t ] for three different @xmath4 values , @xmath406 thus becomes a monotonously decreasing function of time ( thin lines ) . interestingly , @xmath406 appears not to depend on @xmath4 . we note finally that using either the large - time asymptotics eq . ( [ eq_muabt_asymp ] ) or , more conveniently , the debye relation eq . ( [ eq_debye ] ) , one may determine the relaxation time @xmath407 of the strain - strain fluctuations @xmath403 . one verifies that @xmath408 holds as expected for an exponentially decaying correlation function ( sect . [ theo_dyna_model ] ) . [ [ strain - stress - msd . ] ] strain - stress msd . + + + + + + + + + + + + + + + + + + the scaling of the strain - stress msd @xmath198 is investigated in fig . [ fig_ggtt ] . the relaxation time @xmath314 presented in panel ( b ) has been determined , as before for @xmath69 , by matching the short - time diffusive regime with the long - time limit @xmath409 . as indicated by the dash - dotted line , the data are well described by @xmath410 that this scaling must hold can be seen by replacing in the definition of @xmath198 the stress fluctuation @xmath411 by @xmath412 which implies @xmath413 for the diffusion coefficients and in turn eq . ( [ eq_tgttgg ] ) using again eq . ( [ eq_tabdab ] ) . we stress that in agreement with the inset of fig . [ fig_tau_gam ] the system has not enough time to relax the affine strain imposed by the shear - barostat for the short times @xmath414 considered here . it is thus @xmath19 and not @xmath9 which has to be used as the linear slope coefficient . the main panel ( a ) presents the collapse of @xmath415 as a function of @xmath416 for @xmath313 . minor deviations are visible for @xmath417 . the quality of the collapse improves by decreasing @xmath309 , whereas the crossover becomes more sudden ( even a hump may occur ) if @xmath349 is too large ( not shown ) . [ [ time - dependent - correlation - coefficient - g_mathrmgammataut . ] ] time - dependent correlation coefficient @xmath418 . + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + since the strain - shear correlations are dominated for @xmath419 by the _ affine _ shearing , one expects @xmath418 computed using the ratio of @xmath420 and @xmath403 to be similar to @xmath19 for @xmath124 . as one sees from @xmath418 presented in fig . [ fig_geq_t ] ( filled symbols ) , this is indeed the case . interestingly , @xmath418 is seen to be independent of @xmath4 for _ all _ times @xmath235 , i.e. the @xmath4-dependences of @xmath420 and @xmath403 do cancel for all times just as they cancel for the ratio of the equilibrium moments @xmath421 . [ [ introduction.-3 ] ] introduction . + + + + + + + + + + + + + we turn finally to the more intricate characterization of the stress - stress correlations as a function of our two operational parameters @xmath4 and @xmath76 . we focus again on the regime with @xmath389 . as above we begin by discussing the msd @xmath202 . since according to eq . ( [ eq_gabt_cabt ] ) @xmath422 the msd @xmath202 and the correlation function @xmath182 contain in principal the same information . from the presentational point of view @xmath202 has the advantage that the short - time power - law behavior of the correlations can be made manifest using a double - logarithmic plot as shown in fig . [ fig_gttt ] . we describe then the large time behavior of the correlation function @xmath182 and demonstrate that @xmath172 is given by @xmath69 . finally , we turn to the time - dependent stress fluctuation @xmath423 which is used to determine the stress - stress relaxation time @xmath312 . [ [ stress - stress - msd . ] ] stress - stress msd . + + + + + + + + + + + + + + + + + + let us concentrate first on the data for @xmath424 presented by open symbols in fig . [ fig_gttt ] . if @xmath76 is large , all internal dynamics is destroyed and we find immediately @xmath425 as seen for @xmath426 ( small circles ) . for smaller , but not too small @xmath76 one observes a free - diffusion regime with @xmath427 as shown by the bold dash - dotted lines on the left . as above for @xmath198 in sect . [ simu_dyna_gt ] , @xmath411 in the definition of @xmath202 can be replaced by @xmath412 . this argument yields the diffusion coefficient @xmath428 successfully used in the plot for @xmath385 and @xmath313 . if @xmath76 is further reduced , @xmath202 becomes @xmath76-independent for short times . as seen for @xmath429 and @xmath338 the dynamics becomes deterministic " in the sense that @xmath430 with @xmath431 as shown by the thin solid line . albeit @xmath61 depends on @xmath4 , the ratio @xmath432 does not , in agreement with the fundamental postulate eq . ( [ eq_msdscaling ] ) . this implies that @xmath433 depends somewhat on @xmath4 . the @xmath434-scaling in eq . ( [ eq_gttt_determinist ] ) is expected @xcite since the associated correlation function @xmath182 must be a continuous and symmetric function at @xmath185 which , moreover , should be an _ analytic _ function if barostat and thermostat effects can be ignored @xcite . this implies that @xmath202 must be an even expansion in terms of @xmath434 which to leading order leads to eq . ( [ eq_gttt_determinist ] ) . since there are several dynamical regimes it is not meaningful to determine a crossover time @xmath435 using eq . ( [ eq_tabdab ] ) as before for @xmath69 and @xmath314 over the full range of @xmath4 and @xmath76 . we shall see at the end of this section how a longest relaxation time @xmath312 for the stress - stress correlations might be obtained . for times @xmath211 an intermediate plateau appears ( bold solid line ) as predicted by eq . ( [ eq_cttt_plateau ] ) . as can be seen for four different @xmath4 values this intermediate plateau does _ not _ depend on the ensemble provided that @xmath76 is sufficiently weak . ( as one expects from fig . [ fig_aeta ] deviations from this asymptotic limit arise again for large @xmath76 if @xmath436 . ) we emphasize that the two solid lines indicated in fig . [ fig_gttt ] give the lower envelope for all parameters @xmath4 and @xmath76 we have simulated . the msd @xmath202 behaves thus as an ensemble - independent simple average as we have argued in sect . [ theo_dyna_lambda ] . one may operationally define the crossover time @xmath171 by matching eq . ( [ eq_gttt_determinist ] ) and eq . ( [ eq_gttt_plateau ] ) . this yields @xmath437 as indicated by the vertical line . note that @xmath438 for @xmath3 . we stress that while @xmath433 is a function of @xmath4 , @xmath171 is an ensemble - independent constant for the given elastic network . ( moreover , it can be shown that it barely depends on the friction constant @xmath78 of the langevin thermostat @xcite . ) using eq . ( [ eq_taua ] ) one may reformulate eq . ( [ eq_gttt_determinist ] ) in a manifest @xmath4-independent form : @xmath439 . the msd ultimately approaches @xmath440 for even larger times @xmath209 as indicated by the dashed horizontal line for @xmath424 . remembering that @xmath441 for @xmath3 it is clear that the @xmath41-data stays constant for all times . [ [ relaxation - modulus . ] ] relaxation modulus . + + + + + + + + + + + + + + + + + + + the large-@xmath235 scaling is further addressed in fig . [ fig_cttt ] and fig . [ fig_cttt_gxymax ] presenting the stress - stress correlation function @xmath182 . the most central result of this work stated by eq . ( [ eq_keydyna ] ) is demonstrated in fig . [ fig_cttt ] . we compare the directly measured relaxation modulus @xmath39 with the equilibrium correlation function @xmath182 assuming an asymptotically slow barostat @xmath442 ) . averaging over 1000 independent configurations from an @xmath20-ensemble with @xmath283 for @xmath214 , the relaxation modulus @xmath443 measures the stress response after a ( canonical and affine ) shear strain @xmath444 was been applied at @xmath185 ( stars ) . the correlation functions @xmath182 for @xmath317 have been obtained by averaging over 1000 equilibrated configurations from a @xmath4-ensemble as indicated . switching off the shear - barostat the strain @xmath8 of each configuration is quenched . note that @xmath445 holds due to the ensemble averaging . as expected from eq . ( [ eq_cttt_plateau ] ) , @xmath182 decreases monotonously from @xmath61 down to @xmath446 for @xmath211 . this corresponds to the @xmath4-independent intermediate plateau @xmath447 in fig . [ fig_gttt ] . it is seen that @xmath448 for @xmath1 ( large circles ) and that for all @xmath4 one may obtain @xmath39 by vertically shifting the correlation functions @xmath449 using the already determined shear modulus @xmath9 . note also that @xmath450 for large times while all other @xmath182 remain finite . as already emphasized at the end of sect . [ theo_dyna_gtjt ] , it is thus impossible to obtain the modulus @xmath9 _ solely _ from @xmath38 . please note that the observed short - time value @xmath451 ( bold dashed line ) for @xmath3 is quite different from the equilibrium modulus @xmath452 ( bold solid line ) . [ [ finite - kappa - effects . ] ] finite-@xmath76 effects . + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + focusing on one ensemble with @xmath339 we investigate in fig . [ fig_cttt_gxymax ] the effect of a switched - on shear barostat . the stars refer to data already seen in fig . [ fig_cttt ] using the quenched - strain ensemble ( @xmath453 , @xmath454 ) . as one expects , the decay of @xmath182 becomes systematically slower with decreasing @xmath76 . for large @xmath455 we have computed @xmath182 by gliding averaging over a single trajectory of @xmath384 . for smaller @xmath76 it becomes increasingly harder to have a sufficiently long trajectory of @xmath456 probing the phase space , i.e. for which @xmath445 , and to store at the same time the shear stresses with a sufficiently fine time resolution . the data for @xmath457 have thus been obtained by using an ensemble of @xmath458 configurations of different @xmath8 and averaging over the trajectories obtained for each configuration using a finite @xmath76 . as expected from eq . ( [ eq_cttt_plateau ] ) , for sufficiently small @xmath76 all correlation functions have an intermediate plateau for @xmath459 ( bold solid line ) . that the barostat is not completely switched off can be seen from the final decay . the vertical arrows indicate the crossover times @xmath69 for the four smallest @xmath76 . as one expects , this time scale appears to coincide roughly ( up to a constant prefactor ) with the upper limit @xmath172 of the respective plateau . in other words , the plateau disappears when the ergodicity breaking associated with the averaging over a quenched - strain ensemble is lifted by the shear - barostat . [ [ time - dependent - stress - fluctuation - formula . ] ] time - dependent stress - fluctuation formula . + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + let us return to fig . [ fig_geq_t ] . according to the generalized stress - fluctuation formula @xmath460 , eq . ( [ eq_gttgen ] ) , the shear modulus @xmath9 may be obtained from the affine shear - elasticity @xmath19 and the equilibrium stress - stress fluctuation @xmath61 for @xmath60 and @xmath383 ( fig . [ fig_muf_l ] ) . if instead of @xmath61 a time - averaged fluctuation @xmath423 is used , this leads again to a time - dependent monotonously decreasing estimate @xmath461 as seen from the open symbols in fig . [ fig_geq_t ] . no shear - barostat is of course applied for the @xmath20-ensemble indicated by the large squares . we emphasize the remarkable finding that within numerical accuracy @xmath462 for all times and a broad range of @xmath76 for @xmath418 . as seen for @xmath463 , for small @xmath4 and not too short times @xmath461 and @xmath406 are generally similar . since @xmath423 must vanish for small times , this implies that @xmath461 must become finite @xmath464 and , hence , @xmath465 for @xmath466 . this limit is indicated by the bold dashed line for @xmath3 . [ [ time - dependence - of - stress - stress - fluctuations . ] ] time - dependence of stress - stress fluctuations . + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + the time - dependence of @xmath423 for several @xmath4 is explicitly represented in fig . [ fig_muf_t ] assuming @xmath354 for @xmath317 . while the simple mean @xmath19 converges immediately with time , @xmath423 increases monotonously from zero to the large - time limit @xmath61 ( horizontal lines ) . as already pointed out in sect . [ theo_dyna_genrel ] , assuming time - translational invariance the time - dependence of @xmath423 is a consequence of the time - dependence of the @xmath182 . to emphasize this point we have integrated @xmath182 as suggested by eq . ( [ eq_muabt_cabt ] ) . as seen by the thin lines indicated in fig . [ fig_muf_t ] , this leads to consistent results . if one has thus characterized @xmath182 , this implies @xmath423 and then in turn @xmath461 . note that for all @xmath4 the asymptotic limit is reached at about @xmath467 which is of the same order as @xmath468 for @xmath1 and @xmath385 . [ [ relaxation - time - ensuremaththeta_mathrmtautau . ] ] relaxation time @xmath312 . + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + the convergence of @xmath423 is systematically analyzed in fig . [ fig_thetatt ] . the main panel presents the dimensionless deviation @xmath469 for several @xmath76 and @xmath339 . as emphasized by the thin dashed lines , a correlation time @xmath312 may be measured from the large - time @xmath470-decay following eq . ( [ eq_muabt_asymp ] ) . ( unfortunately , the sampling time is yet insufficient for our smallest @xmath76 . ) with decreasing @xmath76 the decay becomes systematically slower and a horizontal plateau with @xmath471 ( bold solid line ) appears in an intermediate time window . this plateau is expected from eq . ( [ eq_intplat ] ) and eq . ( [ eq_cttt_plateau ] ) . following the discussion in sect . [ theo_dyna_genrel ] , this implies that for sufficiently small @xmath76 and not too large @xmath4 the relaxation time @xmath312 must scale ( up to a prefactor of order unity ) as @xmath472 to leading order where we have used eq . ( [ eq_tgg_fdl ] ) for @xmath473 . that the @xmath76-scaling holds can be seen in panel ( b ) of fig . [ fig_thetatt ] . ( for @xmath339 we have computed @xmath312 also for the additional values @xmath340 and @xmath341 . ) the same representation as in the insets of figs . [ fig_gggt ] and [ fig_ggtt ] has been chosen for comparison . it should be stressed , however , that for @xmath312 this is for two reasons not a rigorous scaling plot over the full range of operational parameters considered . firstly , due to the @xmath474-factor in eq . ( [ eq_thetatt ] ) there is an additional @xmath4-dependence even for @xmath392 and not too large @xmath4 . ( this can be accounted for using a different representation . ) more importantly , for @xmath475 barostat effects naturally become irrelevant and the intrinsic relaxation time @xmath171 essentially sets @xmath312 and not @xmath476 . we thus observe @xmath477 for all @xmath478 as seen in table [ tab ] for @xmath313 . basically , since we have _ two _ time scales @xmath171 and @xmath69 , it is not possible to put all @xmath479 on _ one _ master curve . [ [ static - properties . ] ] static properties . + + + + + + + + + + + + + + + + + + focusing on two - dimensional elastic networks ( sect . [ sec_algo ] ) we have characterized theoretically ( sect . [ sec_theo ] ) and numerically ( sect . [ sec_simu ] ) the static and dynamical fluctuations of shear - strain and shear - stress in generalized gaussian ensembles as a function of the dimensionless parameter @xmath480 , fig . [ fig_sketch ] . monitoring various properties we interpolate between the @xmath20-ensemble ( @xmath3 ) and the @xmath23-ensemble ( @xmath1 ) and consider even negative @xmath4 . the static strain - strain fluctuations @xmath307 ( fig . [ fig_dgam ] ) , the strain - stress fluctuations @xmath95 and the stress - stress fluctuations @xmath61 ( fig . [ fig_muf_l ] ) have been shown to decrease linearly with increasing @xmath4 @xcite . as a consequence , the static shear modulus @xmath9 may be obtained ( fig . [ fig_muf_l ] ) from either the strain - strain fluctuation formula @xmath365 , eq . ( [ eq_ggggen ] ) , or from the strain - stress formula @xmath379 , eq . ( [ eq_ggtgen ] ) , for @xmath317 or the stress - stress fluctuation formula @xmath460 , eq . ( [ eq_gttgen ] ) , for @xmath383 . the latter formula is a generalization of the well - known stress - fluctuation formula eq . ( [ eq_gtt ] ) for @xmath481 . when sampled from a _ finite _ time series @xmath482 , @xmath406 , @xmath418 and @xmath461 behave similarly with time @xmath235 for all @xmath4 ( fig . [ fig_geq_t ] ) . interestingly , @xmath483 and @xmath484 are identical for _ all _ times . the influence of the parameter @xmath76 of the mc shear - barostat ( sect . [ algo_finitet ] ) has been investigated for various dynamical properties , especially for the stress - stress correlation function @xmath182 ( fig . [ fig_cttt_gxymax ] ) . the upper time limit @xmath68 , below which barostat effects can be ignored , is set by the strain - strain crossover time @xmath485 ( fig . [ fig_gggt ] ) . while a rapid barostat may be of advantage for static properties , a sufficiently slow barostat with a large @xmath172 corresponds to the fundamentally important linear - response limit where the barostat is only relevant for the distribution of the start points of the trajectories ( and , hence , @xmath486 ) but _ not _ for their dynamical pathways for @xmath184 . we have argued that this implies the fundamental scaling @xmath487 for @xmath64 as demonstrated explicitly for @xmath202 in fig . [ fig_gttt ] . assuming such a slow barostat one may obtain @xmath9 using eq . ( [ eq_keydyna ] ) from @xmath488 for @xmath317 ( fig . [ fig_cttt ] ) . as emphasized elsewhere @xcite , it is impossible , however , to obtain the modulus @xmath9 _ alone _ from the autocorrelation function @xmath38 . the experimentally important shear - stress relaxation modulus @xmath39 may be most readily determined at @xmath3 calculating first @xmath489 and then @xmath490 @xcite . we have commented briefly on the creep compliance @xmath74 . being well - described by the kelvin - voigt model , eq . ( [ eq_kelvinvoigt ] ) , @xmath74 is best obtained from the strain - strain msd @xmath192 at @xmath1 ( fig . [ fig_gggt ] ) . most of the theoretical relations and numerical techniques discussed above , especially the key formulae eq . ( [ eq_keystat ] ) and eq . ( [ eq_keydyna ] ) , generalize readily for any pair of _ continuous _ extensive and intensive variables @xmath27 and @xmath28 with @xmath491 being the equilibrium modulus . with @xmath492 being the spring constant of the external potential controlling the fluctuations of the extensive variable and @xmath493 one obtains , e.g. , the relaxation modulus @xmath494 with @xmath495 being the relevant autocorrelation function . the associated msd @xmath496 is expected to be strictly @xmath4-independent in the same time window . these properties must be computed again either using a slow barostat " with a large , albeit finite @xmath497 or , equivalently , by averaging over an equilibrium ensemble of configurations with frozen @xmath498 . for times @xmath209 a switched - on barostat " ultimately restores the ergodicity and @xmath499 must vanish while @xmath500 approaches its finite thermodynamic value @xmath501 .	shear - strain and shear - stress correlations in isotropic elastic bodies are investigated both theoretically and numerically at either imposed mean shear - stress @xmath0 ( @xmath1 ) or shear - strain @xmath2 ( @xmath3 ) and for more general values of a dimensionless parameter @xmath4 characterizing the generalized gaussian ensemble . it allows to tune the strain fluctuations @xmath5 with @xmath6 being the inverse temperature , @xmath7 the volume , @xmath8 the instantaneous strain and @xmath9 the equilibrium shear modulus . focusing on spring networks in two dimensions we show , e.g. , for the stress fluctuations @xmath10 ( @xmath11 being the instantaneous stress ) that @xmath12 with @xmath13 being the affine shear - elasticity . for the stress autocorrelation function @xmath14 this result is then seen ( assuming a sufficiently slow shear - stress barostat ) to generalize to @xmath15 with @xmath16 being the shear - stress relaxation modulus .
in the standard model , b quarks produced in e@xmath8e@xmath9 collisions with @xmath10 have an average longitudinal polarization of @xmath11 ; the sign indicates that the b quarks are predominantly left - handed . the variation of this polarization with the angle of emission of the b quark @xmath12 axis points along the e@xmath9 beam and the @xmath13 axis points towards the center for the lep ring . the polar and azimuthal angles are denoted by @xmath14 and @xmath15 , and the origin is taken to be the center of the detector . ] is only @xmath162% , and the dependence on @xmath17 is fairly weak : @xmath18 . this average polarization of b quarks is the quantity @xmath19 measured by sld @xcite , and is an ingredient to the @xmath20 measurements from lep @xcite . in the heavy - quark limit , the spin of the b quark is expected to decouple from the light degrees of freedom produced by the fragmentation processes . the process of stable hadron formation has a profound effect on the polarization retained in the final state . the pseudoscalar mesons b@xmath0 and b@xmath21 carry no polarization information . the vector meson b@xmath22 decays by photon radiation which couples mainly to the light degrees of freedom within the meson , and hence the radiated photons are very nearly isotropic @xcite . when directly produced , the @xmath2 is expected to retain nearly the full polarization of the parent b quark @xcite , where the light valence quarks form a spin - zero system . some @xmath2 mesons are produced from strong decays of the @xmath23 and the @xmath24 . the energy difference between these states is the hyperfine splitting induced by the relative orientation of the b quark spin and the spin-1 diquark . these states are produced in a coherent superposition in the hadronization process . if the @xmath23 and the @xmath24 have narrower widths than their mass splittings , then they decay incoherently ; their lifetime is long enough to flip the spin of the b quark . if the states are broader than the splitting , then the b polarization retention is higher . the widths of these states are predicted to be of the order of the mass splitting between them , hence substantial depolarization is expected for these cascade decays @xcite . a measurement of the average polarization of @xmath2 baryons tests both heavy quark effective theory ( hqet ) and also models of baryon formation . the polarization of @xmath2 baryons is most easily measured using their semileptonic decays @xmath25 . to conserve angular momentum in the maximally parity - violating decay , the charged leptons @xmath26 are preferentially emitted antiparallel to the spin of the @xmath2 , while the @xmath27 s are preferentially emitted parallel to the spin . when boosted into the laboratory frame , the charged lepton energy is on average higher than the neutrino energy . the spectra depend on the @xmath2 momentum spectrum and the details of the semileptonic decay , which are influenced by @xmath28 , qcd corrections , and form factors . the ratio of the average charged lepton energy to the average neutrino energy partially cancels these effects and is statistically more powerful than either energy alone @xcite . this analysis is based on the event - by - event distribution of the ratio @xmath29 . the ingredients needed for the analysis are a purified sample of semileptonic @xmath2 decays , an estimation of the background contribution , and measurements of the lepton and neutrino energies . the opal measurement is described in detail in reference @xcite and is summarized here . the selection uses correlations between reconstructed @xmath30 decays and high-@xmath31 leptons similar to the selection of @xcite , but with improved electron identification , different lepton kinematic cuts , and the inclusion of the requirement of a @xmath32 , whose charge is required to be correlated with the reconstructed @xmath33 to take advantage of the decay chain @xmath34 . the @xmath32 is additionally required to be consistent with coming from the common @xmath33-lepton vertex . the invariant mass distribution of the selected p@xmath35 combinations is shown in figure [ ppifig ] , separately for the @xmath36 ( right - sign ) and @xmath37 ( wrong - sign ) combinations . from a total of 4.3 million hadronic z@xmath0 decays used in all three analyses presented in this article , a total of 912 right - sign @xmath38 combinations are selected with an overall b baryon purity of 69% . a total of 316 wrong - sign combinations are also selected . the wrong - sign combinations were used to estimate the level of background as well as the shape of the background distribution of @xmath29 . part of the wrong - sign background consists of real leptons from @xmath2 decay combined with fragmentation @xmath33 s , which are preferentially produced with the wrong sign because of conservation of baryon number ; these events contain the same polarization information as right - sign events . 3.0 truein to estimate the neutrino energy , the missing energy in the signal hemisphere is computed as the difference between the expected total energy in the signal hemisphere and the measured energy in the same hemisphere : @xmath39 where @xmath40 is the beam energy , @xmath41 is the measured visible energy in the signal hemisphere , @xmath42 is the invariant mass of the observed particles in the signal hemisphere and @xmath43 is the invariant mass of the opposite hemisphere . the estimated resolution of the neutrino energy is 3.5 gev . the measurement error on the charged lepton momentum is negligible by comparison , although the momentum scale and spectrum modeling introduce a systematic uncertainty . the distribution of the measured @xmath44 is compared between data and monte carlo for events with a high-@xmath45 lepton and also for those without , providing samples enriched in high - energy neutrinos and depleted in them . differences in the @xmath44 spectra are used to estimate corrections and systematic errors . the binned fit to the measured @xmath29 distribution was performed by comparing it to fully simulated monte carlo samples generated with various values of the average polarization @xmath46 . the right - sign and wrong - sign spectra were fit simultaneously . these distributions and the fit are shown in figure [ plfit ] . the @xmath47 per degree of freedom is 0.91 and the fit value is @xmath48 , where the errors are statistical . 3.0 truein table [ lbpsyst ] summarizes the sources of systematic uncertainty considered . added together in quadrature , the total systematic error is estimated to be @xmath49 . the average longitudinal polarization of b baryons is measured to be @xmath50 including the effects of both statistical and systematic uncertainties , the 95% confidence level ( cl ) acceptance region is @xmath51 , excluding for the first time zero polarization at the 95% cl , and also disfavoring full preservation of the original polarization of @xmath11 through the fragmentation and hadronization processes . lc\|lc source of uncertainty & @xmath52 & source of uncertainty & @xmath52 + @xmath44 resolution & @xmath53 & @xmath54 polarization & @xmath53 + @xmath44 reconstruction & @xmath55 & b@xmath56 & @xmath57 + @xmath58 scale and shape & @xmath59 & fitting method & @xmath59 + selection criteria & @xmath53 & theoretical uncertainty ( form factor & @xmath59 + background fraction and shape & @xmath60 & modeling , qcd corrections , @xmath61 & + b fragmentation & @xmath59 & & + + in february 1998 , the delphi collaboration reported evidence @xcite for a radially excited charm meson , labeled the d@xmath3 . the lowest - energy radially excited state is called the d@xmath63 , but the d@xmath62 is easier to select because it is expected to decay to d@xmath64 with a significant branching ratio . delphi s measured signal rate is expressed in terms of the rates of production of d@xmath65 and d@xmath66 mesons in z@xmath0 decay : @xmath67 the 95% cl upper limit on the width of the observed resonance is 15 mev and is limited by the detector resolution . the observed mass is 2637@xmath166 mev . the expected mass of the d@xmath3 is 2629@xmath1620 mev @xcite , consistent with the delphi observation . an estimate of the decay width @xmath68 using a harmonic oscillator approximation @xcite is @xmath69 mev , although a more recent computation by melikhov and pne @xcite of the same partial width using heavy - quark symmetry is between 120 and 160 mev . melikhov and pne also estimate the width @xmath70 to be between 45 and 450 mev , indicating some trouble in interpreting the observation . in addition , cleo has searched for the d@xmath71 in an analysis sensitive only to e@xmath8e@xmath72d@xmath71 @xcite and has placed a limit of @xmath73 at 90% cl . opal has searched for the d@xmath71 and the selection procedure is summarized here . charged d@xmath74 candidates are selected in the decay mode d@xmath75d@xmath76d@xmath77k@xmath78 . the k@xmath78 combination is required to have its invariant mass within @xmath79 of the expected d@xmath0 mass ; the resolution is 25 mev in both data and monte carlo . the d@xmath74 is formed with a pion of opposite sign to the kaon and the scaled energy of the d@xmath74 , @xmath80 is required to exceed 0.2 . the mass difference @xmath81 must be between 142 and 149 mev , corresponding to @xmath82 in the resolution . additional background rejection is obtained by placing requirements on the helicity angle and the angle between the kaon and the d@xmath0 flight direction in the d@xmath0 rest frame . opal s excellent d@xmath83/d@xmath13 capability @xcite is additionally used to purify the kaon sample . the d@xmath71 candidates are formed by combining the d@xmath74 candidates with pairs of oppositely charged tracks , in order to explore delphi s observation in the d@xmath64 decay mode . these pions have loose d@xmath83/d@xmath13 requirements placed on them . the reconstructed mass of the d@xmath71 is corrected by using the pdg value for the d@xmath74 mass and the measured difference between the d@xmath71 and the d@xmath74 candidates . the analysis proceeds separately for events tagged as bottom and charm , in order to suppress background from z@xmath84 , and @xmath85 events . to select candidates from a sample enriched in z@xmath86 events , the d@xmath74 candidate is required to have @xmath87 , and the magnitude of the vector sum of the momenta of the two pions added to form the d@xmath71 candidate is required to be greater than 3.6 gev . to select candidates in z@xmath88 events , the separation of the d@xmath0 vertex from the primary divided by its error is required to be larger than 4.0 . a loose requirement is made on the separation between the d@xmath0 vertex and the d@xmath71 pion vertex to reduce the combinatoric contamination . using a monte carlo simulation , the efficiency for selecting d@xmath71 mesons in z@xmath86 events is estimated to be ( [email protected])% , while for z@xmath88 it is estimated to be ( [email protected])% . the combined signal expectation in opal is a factor of [email protected] times that of delphi , although the mass resolution in opal is 15 mev and the background level is about 50% higher , when the b and c analyses are combined . the combined result of the bottom and charm analyses is shown in figure [ opaldstarprimefig ] , where events that pass either of the two analyses enter the histogram once . the monte carlo expectation for a signal the size of delphi s observation is also shown . limits have been obtained on the d@xmath71 production rate by defining a mass window between 2.59 and 2.67 gev ( @xmath79 ) , and comparing the observed candidate count against a background from a smooth parameterization shown also in figure [ opaldstarprimefig ] . systematic uncertainties , summarized in table [ dstarprimesysttable ] , are incorporated in the limits . the 95% cl upper limit on the number of d@xmath71 candidates in the combined analysis in the mass window is 32.8 , with 443 events observed and 475 expected from the background fit . this can be converted into a limit on the production fraction : @xmath89 no signal is seen separately in the tagged b or c samples , and the upper limits are @xmath90 and @xmath91 3.0 truein the analysis has been checked by reconstructing the orbitally excited d@xmath92 and d@xmath93 , following a very similar selection procedure , except only one pion is added to the d@xmath74 candidate . the rates and widths of these are consistent with monte carlo expectations . combining the limits above with the measurements of these rates , which share common systematic uncertainties , opal sets a limit of @xmath94 at the 95% cl , where @xmath95 is defined in equation [ delphidstarprimeequation ] . lrrr error source & + & c sample & b sample & combined sample + + monte carlo statistics & @xmath96 & @xmath97 & @xmath98 + detector resolution & @xmath99 & @xmath100 & @xmath101 + kaon d@xmath83d@xmath13 & @xmath102 & @xmath103 & @xmath100 + pion d@xmath83d@xmath13 & @xmath104 & @xmath104 & @xmath104 + d@xmath71 width & @xmath105 & @xmath106 & @xmath107 + fragmentation modeling & @xmath108 & @xmath109 & @xmath110 + production rates in b / c & none & none & @xmath111 + total efficiency error & @xmath112 & @xmath113 & @xmath97 + + branching ratio d@xmath75d@xmath114 , d@xmath77k@xmath78 & @xmath115 & @xmath115 & @xmath115 + error on @xmath116 & @xmath117 & @xmath118 & none + total branching ratio error & @xmath119 & @xmath115 & @xmath115 + the measurement of @xmath121 has been carried out mainly because of its interest in interpreting measurements of @xmath122 to extract the quantity @xmath123 , which is an approximate measure of semileptonic branching fraction of the @xmath2 . the average lifetime of the @xmath2 has been measured to be approximately 80% that of the b mesons , while theoretical predictions can accommodate ratios between 90% and 100% @xcite . it is therefore important to check the shortened lifetime with a measurement of the semileptonic branching ratio , which is expected to be proportional to the lifetime within the b hadron family , provided that the leptonic decay currents are the same for the b baryons and the b mesons . a second reason for measuring this product branching ratio is that it allows one to extract @xmath124 . this analysis is fully described in reference @xcite and a summary is provided here . the procedure is to identify @xmath125p@xmath35 decays in b - tagged events , and then to use the momentum , @xmath126 , and transverse momentum , @xmath127 of each @xmath33 with respect to its jet , to separate the contributions from b baryon decay , b meson decay , and @xmath33 s produced by the fragmentation process . events are tagged as z@xmath88 if at least one jet tags as a b using displaced secondary vertices as described in @xcite . the estimated overall efficiency is about 21% per jet and the purity is about 96% per event . the @xmath33 selection requires that the reconstructed mass of the p@xmath35 combination be within 8 mev of @xmath128 , that it be farther than 6 mev from the k@xmath0 mass when interpreted as a @xmath129 pair , that the flight distance be at least 8.0 cm from the interaction point when projected into the @xmath130 plane , that the momentum points back at the interaction point , that it have a momentum exceeding 5 gev , and that it be within 0.2 radians of its jet axis . in addition , opal s d@xmath83/d@xmath13 capability is used to require that the proton s candidate track s ionization is consistent with that expected from a proton . the monte carlo is used to provide distributions of @xmath126 and @xmath127 of the @xmath33 is measured with respect to the nearest jet axis , the calculation of which includes the @xmath33 momentum . ] for @xmath33 s from each of the three sources , b baryon decay , b meson decay , and fragmentation . other backgrounds contribute 3% of the sample , dominated by d@xmath8 mesons tagging as b mesons because of their long lifetime . because the d@xmath8 does not have enough mass to decay into a @xmath33 and another baryon , the main contribution comes from fragmentation @xmath33 s in events with tagged d@xmath8 mesons . these are incorporated into the fragmentation lambda portion of the fit . these distributions of @xmath126 and @xmath127 are jointly fit for the fractions of @xmath33 s in each of the three categories , and the results are shown in figure [ lsfitfig ] and summarized in table [ lambdabrtab1 ] . the momentum is a powerful discriminator between b baryon decay and fragmentation , while the transverse momentum is effective for separating b baryon decays from b meson decays , although this separation is more difficult , which is reflected in the correlations shown in table [ lambdabrtab1 ] . a summary of the systematic errors on the fit fractions is given in table [ lambdabrtab2 ] . checks of systematic biases in the @xmath126 and @xmath127 spectra were made by comparing events with @xmath33-lepton pairs and comparing the averages of @xmath126 and @xmath127 in the right- and wrong - sign samples to obtain estimations for a pure sample of semileptonic @xmath33 decays . 4.0 truein .results of the fit for the fractional contributions to the @xmath33 sample in b - tagged events . the correlations between the fit fractions are also shown . [ cols="<,^,^,^,^ " , ] opal has produced a wide spectrum of heavy - flavor physics results in the past year . in addition to the measurements described here of the b baryon longitudinal polarization , the search for the radial excitation d@xmath71 , and the measurement of the product branching ratio @xmath121 , opal has completed its final @xmath131 measurement @xcite , studied cp violation in the @xmath132-k@xmath133 channel @xcite , measured the @xmath134 and b@xmath0 lifetimes using a topological tag @xcite , and measured the semileptonic branching fraction of inclusive b hadrons @xcite . each analysis uses opal s full lep1 data sample and are among the most precise published measurements of their respective quantities .	a selection of recent heavy - flavor results from opal using the lep1 data sample are presented . the average polarization of b baryons in hadronic z@xmath0 decay has been measured to be @xmath1 using semileptonic decays of @xmath2 baryons . a search has been conducted for the radially excited d@xmath3 and has produced a 95% cl upper limit on its production of @xmath4 . finally , the measurement of the product branching ratio @xmath5 has been made . this measurement , along with an earlier measurement of the product branching ratio @xmath6 , has been used to compute an updated @xmath7 , consistent with the expected low semileptonic branching fraction of the @xmath2 inferred from its short lifetime compared to the other b hadrons .
the galactic plane monitoring performed by the integral satellite has led to the discovery of a number of new high mass x ray binaries ( hmxbs ) in the last 5 years @xcite . several of these new sources are transients associated with ob supergiants and show short outbursts ( a few hours , as observed with integral , @xcite , @xcite ) . these sources have been called supergiant fast x ray transients ( sfxts ) , and their x ray transient behavior is quite surprising since neutron stars accreting from the winds of supergiant companions were known to be persistent . since their bright x ray emission , reaching 10@xmath0 erg s@xmath1 , is concentrated in very short outbursts , they are difficult to discover , but sfxts may be a large ( and probably predominant ) population of massive x ray binaries . their quiescent level has been observed so far only in a few sources : igr j17544@xmath42619 @xcite and igr j08408@xmath44503 @xcite ) , and is at about 10@xmath2 erg s@xmath1 , thus making sfxts a class of transients with a large dynamic range ( @xmath3@xmath8 ) . among them , particularly interesting is the case of igr j112155952 , which is the only sfxt to date which displays periodic outbursts @xcite with a period of 329 days ( or a half of this , as recently discovered with _ swift_/xrt ; @xcite , @xcite ) . the _ swift _ monitoring during the 2007 february outburst of igr j11215@xmath45952 represents the deepest and most complete set of x ray observations of an outburst from an sfxt , allowed thanks to the predictable behavior of the recurrent outbursts @xcite . these observations demonstrated that the accretion phase during the bright outburst lasts longer than previously thought : a few days instead of hours , with only the brightest phase lasting less than one day , and being characterized by a large variability with several flares lasting from a few minutes to a few hours @xcite . these observations allowed us to propose an alternative explanation for the sfxt outburst mechanism , which accounts for both the narrow shape of the igr j112155952 x ray lightcurve and the periodicity in the outburst recurrence . this model suggests the possible presence of a preferential plane in the wind mass loss from the supergiant , an equatorial wind `` disk '' , which should also be inclined with respect to the orbital plane of the binary system to explain the shortness of the outburst @xcite : in this framework , x ray outbursts are produced when the neutron star crosses this equatorial disk component , which is denser and slower than the `` polar '' wind component . thus , depending on the truncation of the disk , its orientation and inclination with respect to the orbital plane , together with the system eccentricity , the neutron star will cross once or twice the disk , resulting in periodic or quasi - periodic outbursts . this implies that basically _ all _ sfxts should display a periodicity in the outburst recurrence , or a double periodicity , depending on the different possible geometries discussed in @xcite . we note that this scenario attempts to describe in a coherent way not only sfxt outbursts but also hmxbs as a class . stimulated by these _ swift _ results on igr j11215@xmath45952 we are performing the first sensitive wide - band x ray monitoring campaign of the activity of a sample of sfxts . a sensitive monitoring campaign carried out as frequently as possible and spanning very long timescales is crucial to fully characterize the behavior of sfxts , to determine the properties of their quiescent state ( where the accumulation of large observing time is needed to allow a meaningful spectral analysis of this faintest emission ) , to monitor the onset of the outbursts and measure the outburst recurrence period(s ) . determination of the outburst pattern is fundamental in order to discover the supposed periodicities ( or quasi - periodicities ) implied by our proposed model . a positive ( or negative ) result will be possible only after several months of observations ( for example , the period for the outburst recurrence in igr j112155952 is @xmath3165 days ) this will allow us to reach a firm conclusion about periodic or non - periodic behavior . a monitoring with these aims is now possible only with the _ swift _ satellite , which offers a good sensitivity combined with a broad - band coverage ( from optical / uv to hard x rays ) and , most importantly , a unique flexibility . this latter becomes crucial when observing transients with short outbursts , in order to rapidly modify the original schedule when they show indications of an imminent outburst ( possible `` bounces '' in the light curve , which would normally be missed ) , to catch not only the ourburst but also its onset . here we report on the results of the first four months of this on - going campaign with _ swift _ , which targeted the following sources : igr j16479@xmath44514 , xte j1739@xmath4302 , igr j17544@xmath42619 and igr j18410@xmath40535 ( see @xcite and references therein , for an updated review of the parameters of these systems ) . the targets have been selected considering sources which , among several candidates of this newly discovered class of sources , are confirmed sfxts , i.e. they display both a `` short '' transient ( and recurrent ) x ray activity and they have been optically identified with supergiant companions ( see references in @xcite ) . in particular , xte j1739 - 302 and igr j17544@xmath42619 are generally considered prototypical sfxts : xte j1739@xmath4302 was the first transient which showed an unusual x ray behavior @xcite , only recently optically associated with a blue supergiant @xcite . igr j16479@xmath44514 has displayed in the past a more frequent x ray outburst occurrence than other sfxts , ( see , e.g. walter et al . 2007 ) , and offers an _ a priori _ better chance to be caught during an outburst . ax j1841.0@xmath40536/igr j18410@xmath40535 , on the other hand , is an interesting source which , together with igr j112155952 , is the only sfxt where a pulsar has been detected @xcite and which may offer the opportunity to determine the orbital parameters from the pulsar timing on long timescales . the results we are reporting here are concentrated on the _ out - of - outburst _ behavior of the four sfxts monitored with _ swift _ between 2007 october 26 and 2008 february 28 . an old outburst from igr j16479@xmath44514 observed in 2005 in archival _ swift _ data is also discussed . table [ sfxts : tab : alldata ] reports the log of the _ swift _ observations , which include both the new and the old ones , retrieved from the _ swift _ archive . the xrt data were processed with standard procedures ( xrtpipeline v0.11.6 ) , filtering , and screening criteria by using ftools in the heasoft package ( v.6.4 ) . we considered both wt and pc data , depending on the count rate of the sources , and selected event grades 02 and 012 , respectively . when appropriate , we corrected for pile - up . to account for the background , we also extracted events within source - free regions . ancillary response files were generated with xrtmkarf , and they account for different extraction regions , vignetting , and psf corrections . we used the latest spectral redistribution matrices ( v010 ) in the calibration database maintained by heasarc . for timing analysis , the arrival times of xrt events were converted to the solar system barycentre with the task barycorr . during the observations of our monitoring campaign bat observed the sources simultaneously with xrt , and survey data products , in the form of detector plane histograms ( dph ) , are available . igr j16479@xmath44514 @xcite and xte j1739@xmath4302 also triggered the bat in the past ( see table [ sfxts : tab : alldata ] , observations labelled with ( a ) ) and in those cases bat events are also available . the optical counterparts of the sources are expected to display a flux at a level of v@xmath3 1620 mag , and have been observed by uvot . the bat and uvot data will be included in a forthcoming paper . figures [ sfxts : fig : lcv_all](a ) to [ sfxts : fig : lcv_all](f ) show the results as lightcurves in the 0.210kev band , which were corrected for pile - up , psf losses , and vignetting and background - subtracted . each point in the lightcurves of figure [ sfxts : fig : lcv_all](b ) to [ sfxts : fig : lcv_all](f ) refers to the average flux observed during each snapshot observation performed with _ swift_/xrt and reported in table [ sfxts : tab : alldata ] ; each point in figure [ sfxts : fig : lcv_all](a ) has at least 25 source counts per bin and minimum time bin of 15s , while the insets in figures [ sfxts : fig : lcv_all](b ) and figure [ sfxts : fig : lcv_all](e ) have at least 18 source counts per bin and minimum time bin of 30s . these insets show details of specific observations . the downward - pointing arrows are 3-@xmath9 upper limits . the lack of observations occurring roughly from december 2007 to january 2008 , depending on the target coordinates , is due to the sources being sun - constrained . we show both archival data as black points ( triangles ) and the toos we requested in the last four months ( from 2007 october 26 , until till the end of 2008 february ) as red points ( filled circles ) . in particular , figure [ sfxts : fig : lcv_all](a ) shows an intense outburst of igr j16479@xmath44514 which triggered bat in 2005 ( image trigger , trigger number 152652 ) ; in that instance , this source was detected with bat for about 60s , with an average significance level of 8.7 @xmath9 in the 15100kev energy range , and an average count rate of ( 7.7@xmath10@xmath1110@xmath12 counts@xmath1 s@xmath1 det@xmath1 ; unfortunately this source was observed only once during the fall 2007 monitoring campaign because of sun constraints ( figure [ sfxts : fig : lcv_all](b ) ) . figures [ sfxts : fig : lcv_all](c ) and [ sfxts : fig : lcv_all](d ) show the full dataset on igr j17544@xmath42619 . we note that an outburst was discovered in bat monitor data on mjd 54412 @xcite , merely two days after the source had become unobservable by the xrt because it was sun - constrained . [ sfxts : fig : lcv_all](e ) and [ sfxts : fig : lcv_all](f ) show the light curves of xte j1739@xmath4302 and igr j18410@xmath40535 , respectively . bamba et al . ( 2001 ) discovered coherent pulsations with a period of [email protected] s in the brightest flare phase of the transient x ray pulsar ax j1841.0@xmath40536/igr j18410@xmath40535 . in order to search for periodicities , we considered the _ swift_/xrt observations with the best signal to noise ( observations i d 00030988001 , 00030988004 and 00030988005 in table[sfxts : tab : alldata ] ) . we examined a frequency range centered on the above value , performing an epoch folding technique on this subset of data . the best - fit pulse period is [email protected] s. the corresponding pearson statistics for the pulse histogram with 10 bins ( @xmath320 mean source counts per bin ) provides a reduced @xmath14 of 6.3 ( 9 degrees of freedom ) which , even taking into account the @xmath3@xmath15 searched periods , has a virtually null probability of chance occurrence . the long term lightcurves of all sources show a large variability over time scales of hours , days and weeks , with a frequent low level flaring activity . in order to characterize the spectral properties of the sources in this fainter state , we extracted a single spectrum for each source , integrating over all the available observing time when the source is not in outburst . the average spectral parameters are reported in table [ sfxts : tab : spectra ] . they display hard and highly absorbed spectra ( see fig . [ sfxts : fig : spec_all ] ) , with absorption in excess of the total interstellar galactic value towards the sources . the spectrum of the source igr j16479@xmath44514 during the bright outburst reported in fig . [ sfxts : fig : lcv_all](a ) is fit well by a highly absorbed ( column density of 410@xmath1110@xmath16 cm@xmath16 ) hard power law ( photon index , @xmath18 , in the range 0.61.4 ) . during the fainter state monitored in 2008 ( see fig [ sfxts : fig : spec_all](a ) ) , the spectrum appears softer ( @xmath19@xmath20 ) and equally highly absorbed . there is no evidence for a variable absorbing column density in igr j16479@xmath44514 between the long term low level emission and the bright outburst observed in 2005 . we report here on the results of the first sensitive long term x ray monitoring of sfxt lightcurves , performed with _ swift _ , focussing on the out - of - outburst x ray behavior . the first months of these observations of a sample of four transients reveal that they spend most of the time in a low - level x ray activity which is far from being a true quiescent state . all four of these sources display , outside the bright outbursts , a highly variable low level x ray activity , characterized by a flaring behavior with a large dynamic range : more than one order of magnitude for all the sources . the average spectra of this x ray emission are characterized by hard powerlaw photon indexes ( in the range 12 ) , high absorbing column densities , and average fluxes which translate into x ray luminosities of a few 10@xmath5 erg s@xmath1 ( see table [ sfxts : tab : spectra ] ) , assuming the distances optically determined by @xcite ) , up to an average level of 68@xmath1110@xmath6 erg s@xmath1 observed in igr j16479@xmath44514 . previous x ray observations of sfxts outside outbursts consist of a few short exposures with asca , _ swift_/xrt , _ chandra _ , xmm-_newton _ , which were not part of a systematic monitoring campaign . during these observations the true quiescent state was caught in igr j17544@xmath42619 @xcite , in igr j08408@xmath44503 @xcite , and probably in xte j1739@xmath4302 ( @xcite ) . this quiescence is characterized by a luminosity of about 10@xmath2 erg s@xmath1 , and by a very soft spectrum : for example , in igr j17544@xmath42619 , the fit with a power law resulted in a photon index of almost 6 @xcite . the kind of low - level activity we observe now with _ swift _ on long timescales of months , has only been observed before in single short exposures targeted on igr j17544@xmath42619 @xcite and on xte j1739@xmath4302 @xcite , during observations which covered only a few tens of ks . our _ swift _ monitoring campaign , after four months of observations ( on average 12ks of net exposure per source twice or three times a week depending on the source ) , has already demonstrated that accretion at a low level rate is a very frequent and more typical state in sfxts , and that instead the quiescence seems to be a much rarer behavior for these sources than previously thought . these findings firmly establish that sfxts can not be considered as sources which undergo long periods of quiescence only occasionally interrupted by the sudden accretion of matter from the wind of the supergiant companion . instead , they continue accreting matter even when they are not in outburst , over long timescales of months , with a large variability in the x ray flux . we measured the pulsar period in xrt data of the sfxt igr j18410@xmath40535 . compared with the period determined in asca data by @xcite , there is evidence for a spin - up trend , with a pulse period difference of @[email protected]@xmath22 s , and an average period derivative , p , of @xmath23 s s@xmath1 , between the two determinations separated by @xmath32965 days . since the orbital parameters are unknown , it is not possible to correct for the doppler delay due to the orbital motion in the binary system . moreover , a spin - up trend is often observed in hmxbs ( see e.g. @xcite ) induced by the accretion torque @xcite , thus at this stage it is not possible to discriminate between the two effects . to estimate the maximum possible shift on the pulse period due to the orbit , we can assume that the pulsar is orbiting near to the surface of the ob supergiant companion ; in this case the shift is about 0.2% difference between the two period measurements , which is less than what we observe , thus a good fraction of the pulse period change is real . more determinations of the spin period are needed to characterize the orbit , and will hopefully be available in the next months of this observing campaign . on the other hand , besides the large intensity variability and hard spectral properties , this is a further confirmation of the fact that this source is still accreting matter even at much lower rates than in outburst . the main hypotheses proposed to explain the sfxts behavior ( especially their outburst durations ) are based on the structure of the wind of the ob supergiant companions . int zand ( 2005 ) originally proposed that the sudden accretion of material from the clumpy wind of the supergiant could give rise to the bright short outbursts . negueruela et al . ( 2005 ) suggested that sfxts orbits should be highly eccentric to explain the low luminosity in quiescence . more recently , the idea of a clumpy and spherically symmetric wind has been associated with an eccentric and/or wide orbit ( walter & zurita heras 2007 , @xcite ) . for these authors the main difference between the sfxts and the persistent hmxb depends on the number of clumps encountered and accreted by the compact object along its orbit : in persistent systems the rate is very high because the orbit lies within @xmath32 stellar radii from the supergiant , while in the sfxts the orbit probably lies at higher distances , in a region where the wind clumps density is much lower . @xcite suggest that outside this radius the space is effectively void , and the probability for the neutron star to accrete a wind clump very low . in this model the sfxt outbursts are produced by the accretion of a single clump and the x ray luminosity can be used to determine the mass of the clump , knowing the flare duration @xcite . in this case , the mass of the single clumps accreted to explain the low level flaring activity we observe with _ swift _ should be about two orders of magnitude lower than during the bright outburst ( where the luminosity exceeds 10@xmath0 erg s@xmath1 ) , assuming the same flare duration . a second hypothesis has been proposed by @xcite and is based on the shape of the lightcurve observed during the 2007 outburst from the unique sfxt displaying periodic outbursts , igr j11215@xmath45952 . this set of observations clearly demonstrated that the sfxts outbursts can not be produced by the accretion of a single clump , nor by a spherical distribution of clumps . the x ray lightcurve of the 2007 february outburst was too narrow and steep to be explained by an enhanced accretion rate when the neutron star approaches the periastron passage orbiting a supergiant with a spherically symmetric wind , even in a binary system with an extremely high eccentricity . this strongly suggested some degree of anisotropy in the supergiant wind , which could be explained by the presence of a second denser wind component ( besides the polar spherically symmetric wind ) in the form of an equatorial wind `` disk '' from the supergiant donor . the shortness of the outburst is also indicative of an equatorial wind which is inclined with respect to the orbital plane . a flaring activity was detected and could be explained with a wind that is inhomogeneous in addition to being anisotropic @xcite . basically , the main difference between the two hypotheses resides indeed in the spherical ( e.g. @xcite ) vs. anisotropic wind @xcite . in both models the low - level x ray activity could be explained : in the first scenario , it could probably be produced by clumps with a distribution of sizes and masses ( breakup of clumps to smaller sizes at larger distances in the eccentric case , or a distribution of clump sizes in the non - eccentric case ) , while in the anisotropic wind scenario the low level x ray emission is very easily explained by the fact that the neutron star , when not undergoing a bright outburst , is not crossing the `` disk '' wind component from the supergiant donor , but in any case accretes matter from the polar supergiant wind , which is faster and two orders of magnitude less dense than the equatorial wind component . we thank the _ swift _ team for making these observations possible , in particular the duty scientists and science planners . this work was supported by miur grant 2005 - 025417 , contracts asi / inaf i/023/05/0 , i/008/07/0 and i/088/06/0 , and _ swift _ nasa contract nas5 - 00136 . pr thanks inaf - iasfmi , where part of the work was carried out , for their kind hospitality . we thank the anonymous referee for a swift and very constructive report . , i. , smith , d. m. , reig , p. , chaty , s. , & torrejn , j. m. 2006 , in proceedings of the the x - ray universe 2005 , 26 - 30 september 2005 , el escorial , madrid , spain . ed . by a. wilson . esa sp-604 , volume 1 , noordwijk : esa publications division , isbn 92 - 9092 - 915 - 4 , 2006 , ed . a. wilson , 165 llllll igr j16479@xmath44514 & 00152652000 & bat / evt & 2005 - 08 - 30 04:03:49 & 2005 - 08 - 30 04:13:51 & 602 + & 00030296001 & xrt / wt & 2005 - 08 - 30 04:11:00 & 2005 - 09 - 01 13:40:51 & 4541 + & 00030296001 & xrt / pc & 2005 - 08 - 30 04:12:41 & 2005 - 09 - 01 14:01:58 & 11408 + & 00030296002 & xrt / pc & 2005 - 09 - 10 00:15:00 & 2005 - 09 - 12 12:10:59 & 6394 + & 00030296003 & xrt / pc & 2005 - 09 - 14 00:43:34 & 2005 - 09 - 14 10:47:57 & 4182 + & 00030296004 & xrt / pc & 2005 - 10 - 18 09:17:15 & 2005 - 10 - 18 14:35:57 & 5423 + & 00210886000 & bat / evt & 2006 - 05 - 20 17:32:29 & 2006 - 05 - 20 17:33:12 & 43 + & 00215914000 & bat / evt & 2006 - 06 - 24 20:19:49 & 2006 - 06 - 24 20:20:32 & 43 + & 00286412000 & bat / evt & 2007 - 07 - 29 12:07:25 & 2007 - 07 - 29 12:08:08 & 43 + & * 00030296005 * & xrt / pc & 2007 - 10 - 26 08:08:36 & 2007 - 10 - 26 09:42:57 & 1176 + & * 00030296006 * & xrt / pc & 2008 - 01 - 19 22:58:28 & 2008 - 01 - 19 23:13:58 & 927 + & * 00030296007 * & xrt / pc & 2008 - 01 - 22 20:07:57 & 2008 - 01 - 22 20:25:58 & 1079 + & * 00030296008 * & xrt / pc & 2008 - 01 - 26 18:56:04 & 2008 - 01 - 26 19:14:57 & 1133 + & * 00030296009 * & xrt / pc & 2008 - 01 - 29 08:03:13 & 2008 - 01 - 29 08:20:56 & 1062 + & * 00030296010 * & xrt / pc & 2008 - 02 - 05 00:48:59 & 2008 - 02 - 05 16:58:56 & 3056 + & * 00030296011 * & xrt / pc & 2008 - 02 - 09 18:53:39 & 2008 - 02 - 09 23:50:58 & 2749 + & * 00030296012 * & xrt / pc & 2008 - 02 - 10 01:16:40 & 2008 - 02 - 10 04:36:58 & 873 + & * 00030296013 * & xrt / pc & 2008 - 02 - 15 08:12:38 & 2008 - 02 - 15 10:01:25 & 1254 + & * 00030296014 * & xrt / pc & 2008 - 02 - 18 06:49:05 & 2008 - 02 - 18 10:12:57 & 804 + & * 00030296014 * & xrt / wt & 2008 - 02 - 18 06:48:56 & 2008 - 02 - 18 10:10:23 & 37 + & * 00030296015 * & xrt / pc & 2008 - 02 - 20 00:33:38 & 2008 - 02 - 20 00:46:58 & 798 + & * 00030296016 * & xrt / pc & 2008 - 02 - 23 21:44:31 & 2008 - 02 - 23 23:35:57 & 1667 + & * 00030296017 * & xrt / pc & 2008 - 02 - 26 23:38:06 & 2008 - 02 - 26 23:52:56 & 890 + igr j17544@xmath42619 & 00035056001 & xrt / pc & 2005 - 11 - 01 01:26:13 & 2005 - 11 - 01 23:59:59 & 2633 + & * 00035056002 * & xrt / pc & 2007 - 10 - 28 00:20:09 & 2007 - 10 - 29 07:07:56 & 2783 + & * 00035056003 * & xrt / pc & 2007 - 10 - 31 10:19:05 & 2007 - 10 - 31 13:43:35 & 248 + & * 00035056004 * & xrt / pc & 2007 - 11 - 04 00:58:32 & 2007 - 11 - 04 01:16:58 & 1104 + & & bat / mon & 2007 - 11 - 08 01:31:04 & 2007 - 11 - 08 06:07:11 & + & * 00035056005 * & xrt / pc & 2008 - 02 - 07 20:17:58 & 2008 - 02 - 07 23:43:57 & 1331 + & * 00035056006 * & xrt / pc & 2008 - 02 - 10 19:14:07 & 2008 - 02 - 10 20:52:56 & 278 + & * 00035056007 * & xrt / pc & 2008 - 02 - 14 08:17:17 & 2008 - 02 - 14 10:00:56 & 915 + & * 00035056008 * & xrt / pc & 2008 - 02 - 17 00:16:44 & 2008 - 02 - 17 02:04:56 & 1346 + & * 00035056009 * & xrt / pc & 2008 - 02 - 21 10:24:08 & 2008 - 02 - 21 10:40:36 & 960 + & * 00035056010 * & xrt / pc & 2008 - 02 - 24 23:41:09 & 2008 - 02 - 25 01:24:57 & 998 + & * 00035056011 * & xrt / pc & 2008 - 02 - 28 02:41:54 & 2008 - 02 - 28 02:58:57 & 1004 + xte j1739@xmath4302 & 00282535000 & bat / evt & 2007 - 06 - 18 03:10:37 & 2007 - 06 - 18 03:11:20 & 43 + & * 00030987001 * & xrt / pc & 2007 - 10 - 27 09:46:16 & 2007 - 10 - 27 10:05:57 & 1181 + & * 00030987002 * & xrt / pc & 2007 - 10 - 30 10:12:32 & 2007 - 10 - 30 10:25:57 & 805 + & * 00030987003 * & xrt / pc & 2007 - 11 - 01 08:47:14 & 2007 - 11 - 01 09:06:58 & 1181 + & * 00030987004 * & xrt / pc & 2008 - 02 - 04 21:24:07 & 2008 - 02 - 04 22:57:58 & 940 + & * 00030987005 * & xrt / pc & 2008 - 02 - 06 00:48:04 & 2008 - 02 - 06 23:29:58 & 5606 + & * 00030987006 * & xrt / pc & 2008 - 02 - 08 01:10:03 & 2008 - 02 - 08 02:48:58 & 161 + & * 00030987006 * & xrt / wt & 2008 - 02 - 08 01:09:57 & 2008 - 02 - 08 02:46:28 & 11 + & * 00030987007 * & xrt / pc & 2008 - 02 - 11 17:26:01 & 2008 - 02 - 11 19:16:56 & 1295 + & * 00030987008 * & xrt / pc & 2008 - 02 - 13 09:48:13 & 2008 - 02 - 13 11:30:54 & 180 + & * 00030987009 * & xrt / pc & 2008 - 02 - 15 06:32:34 & 2008 - 02 - 15 06:40:58 & 504 + & * 00030987010 * & xrt / pc & 2008 - 02 - 18 03:45:06 & 2008 - 02 - 18 05:26:56 & 498 + & * 00030987011 * & xrt / pc & 2008 - 02 - 21 00:40:01 & 2008 - 02 - 21 00:57:57 & 1076 + & * 00030987012 * & xrt / pc & 2008 - 02 - 22 00:52:08 & 2008 - 02 - 22 01:06:55 & 888 + & * 00030987013 * & xrt / pc & 2008 - 02 - 25 17:19:52 & 2008 - 02 - 25 19:03:57 & 597 + & * 00030987014 * & xrt / pc & 2008 - 02 - 27 01:26:01 & 2008 - 02 - 27 03:05:57 & 937 + & * 00030987015 * & xrt / pc & 2008 - 02 - 28 01:23:19 & 2008 - 02 - 28 01:40:57 & 1056 + igr j18410@xmath40535 & * 00030988001 * & xrt / pc & 2007 - 10 - 26 00:08:53 & 2007 - 10 - 26 06:45:56 & 1384 + & * 00030988002 * & xrt / pc & 2007 - 10 - 28 22:53:11 & 2007 - 10 - 29 23:12:56 & 3199 + & * 00030988003 * & xrt / pc & 2007 - 11 - 03 10:37:52 & 2007 - 11 - 03 12:24:56 & 1122 + & * 00030988004 * & xrt / pc & 2007 - 11 - 05 09:08:39 & 2007 - 11 - 05 09:28:58 & 1218 + & * 00030988005 * & xrt / pc & 2007 - 11 - 09 16:11:30 & 2007 - 11 - 09 16:32:56 & 1286 + & * 00030988006 * & xrt / pc & 2007 - 11 - 12 14:51:25 & 2007 - 11 - 12 14:51:55 & 30 + & * 00030988007 * & xrt / pc & 2007 - 11 - 16 08:40:04 & 2007 - 11 - 16 10:36:58 & 2332 + & * 00030988008 * & xrt / pc & 2007 - 11 - 18 23:22:32 & 2007 - 11 - 18 23:41:58 & 1165 + & * 00030988009 * & xrt / pc & 2008 - 02 - 14 14:44:51 & 2008 - 02 - 14 16:28:58 & 947 + & * 00030988009 * & xrt / wt & 2008 - 02 - 14 14:42:45 & 2008 - 02 - 14 16:20:15 & 127 + & * 00030988010 * & xrt / pc & 2008 - 02 - 18 00:26:32 & 2008 - 02 - 18 02:13:57 & 626 + & * 00030988011 * & xrt / pc & 2008 - 02 - 21 08:46:13 & 2008 - 02 - 21 09:02:58 & 1005 + & * 00030988012 * & xrt / pc & 2008 - 02 - 25 20:32:34 & 2008 - 02 - 25 20:47:57 & 834 + & * 00030988013 * & xrt / pc & 2008 - 02 - 28 03:00:08 & 2008 - 02 - 28 03:16:56 & 1008 ccccccc source & @xmath24 & parameter & av . observed flux & av . luminosity & @xmath25 ( dof ) absorbed power law & & @xmath18 & ( 210 kev ) & ( 210 kev ) & igr j16479@xmath44514 & @xmath26 & @xmath27 & 2.03 & 87 & 0.969 ( 112 ) xte j1739@xmath4302 & @xmath28 & @xmath29 & 0.37 & 3.9 & 1.16 ( 21 ) igr j17544@xmath42619 & @xmath30 & @xmath31 & 0.32 & 6.3 & 0.916 ( 16 ) igr j18410@xmath40535 & @xmath32 & @xmath33 & 0.35 & 13 & 0.59 ( 20 ) absorbed blackbody & & @xmath34 & igr j16479@xmath44514 & @xmath35 & @xmath36 & 1.85 & 62 & 0.954 ( 112 ) xte j1739@xmath4302 & @xmath37 & @xmath38 & 0.32 & 3.0 & 1.08 ( 21 ) igr j17544@xmath42619 & @xmath39 & @xmath40 & 0.26 & 4.4 & 1.109 ( 16 ) igr j18410@xmath40535 & @xmath41 & @xmath38 & 0.30 & 9.6 & 0.72 ( 20 )	supergiant fast x ray transients ( sfxts ) are a new class of high mass x ray binaries ( hmxbs ) discovered thanks to the monitoring of the galactic plane performed with the integral satellite in the last 5 years . these sources display short outbursts ( significantly shorter than typical be / x - ray binaries ) with a peak luminosity of a few 10@xmath0 erg s@xmath1 . the quiescent level , measured only in a few sources , is around 10@xmath2 erg s@xmath1 . the x ray spectral properties are reminiscent of those of accreting pulsars , thus it is likely that all the members of the new class are indeed hmxbs hosting a neutron star , although only two sfxts have a measured pulse period , igr j112155952 ( @xmath3187 s ) and igr j18410@xmath40535 ( @xmath34.7 s ) . several competing mechanisms have been proposed to explain the shortness of these outbursts , mostly involving the structure of the wind from the supergiant companion . to characterize the properties of these sources on timescales of months ( e.g. the quiescent level and the outburst recurrence ) , we are performing a monitoring campaign with _ swift _ of four sfxts ( igr j16479@xmath44514 , xte j1739@xmath4302 , igr j17544@xmath42619 and ax j1841.0@xmath40536/igr j18410@xmath40535 ) . we report on the first four months of _ swift _ observations , started on 2007 october 26 . we detect a low level x ray activity in all four sfxts which demonstrates that these transient sources accrete matter even outside their outbursts . this fainter x ray activity is composed of many flares with a large flux variability , on timescales of thousands of seconds . the lightcurve variability is also evident on larger timescales of days , weeks and months , with a dynamic range of more than one order of magnitude in all four sfxts . the x ray spectra are typically hard , with an average 210 kev luminosity during this monitoring of about 10@xmath510@xmath6 erg s@xmath1 . we detected pulsations from the pulsar ax j1841.0@xmath40536/igr j18410@xmath40535 , with a period of 4.7008@xmath7s . this monitoring demonstrates that these transients spend most of the time accreting matter , although at a much lower level ( @xmath31001000 times lower ) than during the bright outbusts , and that the `` true quiescence '' , characterized by a soft spectrum and a luminosity of a few 10@xmath2 erg s@xmath1 , observed in the past only in a couple of members of this class , is probably a very rare state .
the description and understanding of electronic properties of strongly correlated materials is a very important and long standing problem for ab - initio calculations . widely used approximations for the exchange and correlation energy in density functional theory ( dft ) , mainly based on parametrization of ( nearly ) homogeneous electron gas , miss important features of their physical behavior . for instance both local spin - density approximation ( lsda ) and spin - polarized generalized gradient approximations ( @xmath1gga ) , in their several flavors , fail in predicting the insulating behavior of many simple transition metal oxides ( tmo ) , not only by severely underestimating their electronic band gap but , in most cases , producing a qualitatively wrong metallic ground state . tmos have represented for long time the most notable failure of dft . when the high - t@xmath2 superconductors entered the scene ( their parent materials are also strongly correlated systems ) the quest for new approaches that could describe accurately these systems by first principles received new impulse , and in the last fifteen years many methods were proposed in this direction . among these , lda+u approach , first introduced by anisimov and coworkers @xcite , has allowed to study a large variety of strongly correlated compounds with considerable improvement with respect to lsda or @xmath1gga results . the successes of the method have led to further developments during the last decade which have produced very sophisticated theoretical approaches@xcite and efficient numerical techniques . the formal expression of lda+u energy functional is adapted from model hamiltonians ( hubbard model in particular ) that represent the `` natural '' theoretical framework to deal with strongly correlated materials . as in these models , a small number of localized orbitals is selected and the electronic correlation associated to them is treated in a special way . the obtained results strongly depend on the definition of the localized orbitals and on the choice of the interaction parameters used in the calculation , that should be determined in an internally consistent with . this is not always done and a widespread but , in our opinion , unsatisfactory attitude is to determine the value of the electronic couplings by seeking a good agreement of the calculated properties with the experimental results in a semiempirical way . in this work a critical reexamination of the lda+u approach is proposed , which starting from the formulation of anisimov and coworkers @xcite , and its further improvements @xcite , develops a simpler approximation . this is , in our opinion , the `` minimal '' extension of the usual approximate dft ( lda or gga ) schemes needed when atomic - like features are persistent in the solid environment . in the central part of this work we describe a method , based on a linear response approach , to calculate in an internally consistent way without aprioristic assumption about screening and/or basis set employed in the calculation the interaction parameters entering the lda+u functional used . in this context our plane - wave pseudopotential ( pwpp ) implementation of the lda+u approach is presented and discussed in some details . we stress however that the proposed method is basis - set independent . our methodology is then applied to the study of the electronic properties of some real materials , chosen as representative of `` normal '' ( bulk iron ) and correlated ( bulk cerium ) metals , as well as a few examples of strongly correlated systems ( iron oxide , nickel oxide and fayalite ) . in order to account explicitly for the on - site coulomb interaction responsible for the correlation gap in mott insulators and not treated faithfully within lda , anisimov and coworkers @xcite correct the standard functional adding an on - site hubbard - like interaction , @xmath3 : @xmath4 & = & e_{lda}[n({\bf r } ) ] + \nonumber \\ & & e_{hub}[\{n^{i\sigma}_{m}\ } ] - e_{dc}[\{n^{i\sigma}\}]\end{aligned}\ ] ] where @xmath5 is the electronic density , and @xmath6 are the atomic - orbital occupations for the atom @xmath7 experiencing the `` hubbard '' term . the last term in the above equation is then subtracted in order to avoid double counting of the interactions contained both in @xmath3 and , in some average way , in @xmath8 . in this term the total , spin - projected , occupation of the localized manifold is used : @xmath9 . in its original definition the functional defined in eq . [ simpleu ] was not invariant under rotation of the atomic - orbital basis set used to define the occupancies @xmath6 . a rotationally invariant formulation has then been introduced @xcite where the orbital dependence of @xmath3 is borrowed from atomic hartree - fock with renormalized slater integrals : @xmath10 & = & \frac{1}{2}\sum_{\{m\},\sigma , i}\ { { \langle}m , m''\|v_{ee}\|m',m'''{\rangle}n^{i\sigma}_{mm'}n^{i-\sigma}_{m''m ' '' } \nonumber \\ & & + ( { \langle}m , m''\|v_{ee}\|m',m'''{\rangle}\nonumber \\ & & - { \langle}m , m''\|v_{ee}\|m''',m'{\rangle } ) n^{i\sigma}_{mm'}n^{i\sigma}_{m''m ' '' } \}\end{aligned}\ ] ] with @xmath11 where @xmath12 is the angular moment of the localized ( @xmath13 or @xmath14 ) electrons and @xmath15 the double - counting term @xmath16 is given by : @xmath17 & = & \sum_{i } \frac{u}{2}n^{i}(n^{i}-1 ) \nonumber \\ & -&\sum_{i}\frac{j}{2 } [ n^{i\uparrow}(n^{i\uparrow}-1)+n^{i\downarrow } ( n^{i\downarrow}-1 ) ] .\end{aligned}\ ] ] the radial slater integrals @xmath18 are the parameters of the model ( @xmath19,@xmath20 and @xmath21 for @xmath13 electrons , while also @xmath22 must be specified for @xmath14 states ) and are usually re - expressed in terms of only two parameters , @xmath0 and @xmath23 , describing screened on - site coulomb and exchange interaction , @xmath24 by assuming atomic values for @xmath25 and @xmath26 ratios . to obtain @xmath0 and @xmath23 , anisimov and coworkers @xcite propose to perform lmto calculations in supercells in which the occupation of the localized orbitals of one atom is constrained . the localized orbitals of all atoms in the supercell are decoupled from the remainder of the basis set . this makes the treatment of the local orbitals an atomic - like problem making it easy to fix their occupation numbers and allows to use janak theorem @xcite to identify the shift in the corresponding eigenvalue with the second - order derivative of the lda total energy with respect to orbital occupation . it has however the effect of leaving a rather artificial system to perform the screening , in particular when it is not completely intra - atomic . in elemental metallic iron , for instance , anisimov and gunnarsson @xcite showed that only half of the screening charge is contained in the wigner - seitz cell . this fact , in addition to a sizable error due to the atomic sphere approximation used @xcite , could be at the origin of the severe overestimation of the computed on - site coulomb interaction with respect to estimates based on comparison of spectroscopic data and model calculations@xcite . some aspects of currently used lda+u formulation , and in particular of the determination of the parameters entering the model , have been so far tied to the lmto approach . this is not a very pleasant situation and some efforts have been done recently @xcite to reformulate the method for different basis sets . here we want to elaborate further on these attempts and provide an internally consistent , basis - set independent , method for the calculation of the needed parameters . in order to fully define how the approach works the first thing to do is to select the degrees of freedom on which hubbard @xmath0 " will operate and define the corresponding occupation matrix , @xmath27 . although it is usually straightforward to identify in a given system the atomic levels to be treated in a special way ( the @xmath13 electrons in transition metals and the @xmath14 ones in the rare earths and actinides series ) there is no unique or rigorous way to define occupation of localized atomic levels in a multi - atom system . equally legitimate choices for @xmath27 are @xmath28 projections on normalized atomic orbitals , or @xmath29 projections on wannier functions whenever the relevant orbitals give raise to isolated band manifolds , or @xmath30 mulliken population or @xmath31 integrated values in ( spherical ) regions around the atoms of the angular - momentum - decomposed charge densities . taking into account the arbitrariness in the definition of @xmath27 no particular significance should be attached to any of them ( or other that could be introduced ) and the usefulness and reliability of an approximate dft+u method ( adft+u ) , and of its more recent and involved evolutions like the adft+dmft method , should be judged from its ability to provide a correct physical picture of the systems under study irrespective of the details of the formulation , once all ingredients entering the calculation are determined consistently . all above mentioned definitions for the occupation matrices can be put in the generic form @xmath32 where @xmath33 is the valence electronic wavefunction corresponding to the state ( @xmath34 ) with spin @xmath35 of the system and @xmath36 is the corresponding occupation number . the @xmath37 s are generalized projection operators on the localized - electron manifold that satisfy the following properties : @xmath38 in particular @xmath39 is the projector on the complete manifold of localized states associated with atom at site @xmath7 and therefore @xmath40 is the total localized - states occupation for site @xmath7 . orthogonality of projectors on different sites is _ not _ assumed . in the applications discussed in this work we will define localized - level occupation matrices projecting on atomic pseudo - wavefunctions . the needed projector operators are therefore simply @xmath41 where @xmath42 is the valence atomic orbital with angular momentum component @xmath43 of the atom sitting at site @xmath7 ( the same wavefunctions are used for both spins ) . since we will be using ultrasoft pseudopotentials to describe valence - core interaction , all scalar products between crystal and atomic pseudo - wavefunctions are intended to include the usual s matrix describing orthogonality in presence of charge augmentation @xcite . as already mentioned , other choices could be used as well and different definitions for the occupation matrices will require , in general , different values of the parameter entering the adft+u functional , as it has been pointed out recently also by pickett et al . @xcite where , for instance , the value of hubbard @xmath0 in feo shifts from 4.6 to 7.8 ev when atomic d - orbitals for fe@xmath44 ionic configuration are used instead of those of the neutral atom . in an early study @xcite the u parameter in la@xmath45cuo@xmath46 varies from 6.8 to 7.7 ev upon variation of the atomic sphere radius employed in the lmto calculation . as pointed out in these works it is not fruitful to compare numerical values of u obtained by different methods but rather comparison should be made between results of complete calculations . in order to simplify our analysis and gaining a more transparent physical interpretation of the `` + u '' correction to standard adft functionals we concentrate on the main effect associated to on - site coulomb repulsion . we thus neglect the important but somehow secondary effects associated to non sphericity of the electronic interaction and the proper treatment of magnetic interaction , that in the currently used rotational invariant method is dealt with assuming a screened hartree - fock form . @xcite . we are therefore going to assume in the following that parameter @xmath23 describing these effects can be set to zero , or alternatively that its effects can be mimicked redefining the u parameter as @xmath47 , a practice that have been sometime used in the literature @xcite . the hubbard correction to the energy functional , eqs . [ ub1 ] and [ ub2 ] , greatly simplifies and reads : @xmath48 & = & e_{hub}[\{n^{i}_{mm'}\ } ] - e_{dc}[\{n^{i}\ } ] \nonumber \\ & = & \frac{u}{2}\sum_{i}\sum_{m,\sigma } \{n_{mm}^{i\sigma}-\sum_{m'}n_{mm'}^{i\sigma}n_{m'm}^{i\sigma}\ } \nonumber \\ & = & \frac{u}{2}\sum_{i,\sigma } tr[{\bf n}^{i\sigma}(1-{\bf n}^{i\sigma})].\end{aligned}\ ] ] choosing for the localized orbitals the representation that diagonalizes the occupation matrices @xmath49 with @xmath50 , the energy correction becomes @xmath51 & = & \frac{u}{2}\sum_{i,\sigma } \sum_{i } \lambda^{i\sigma}_i ( 1 -\lambda^{i\sigma}_i ) .\end{aligned}\ ] ] from where it appears clearly that the energy correction introduces a penalty , tuned by the value of the u parameter , for partial occupation of the localized orbitals and thus favors disproportionation in fully occupied ( @xmath52 ) or completely empty ( @xmath53 ) orbitals . this is the basic physical effect built in the adft+u functional and its meaning can be traced back to known deficiencies of lda or gga for atomic systems . an atom in contact with a reservoir of electrons can exchange integer numbers of particles with its environment . the intermediate situation with fractional number of electrons in this open atomic system is described not by a pure state wave function , but rather by a statistical mixture so that , for instance , the total energy of a system with @xmath54 electrons ( where n is an integer and @xmath55 ) is given by : @xmath56 where @xmath57 and @xmath58 are the energies of the system corresponding to states with @xmath59 and @xmath60 particles respectively , while @xmath61 represents the statistical weight of the state with @xmath60 electrons . the total energy of this open atomic system is thus represented by a series of straight - line segments joining states corresponding to integer occupations of the atomic orbitals as depicted in fig . [ parab ] . the slope of the energy vs electron - number curve is instead piece - wise constant , with discontinuity for integer number of electrons , and corresponds to the electron affinity ( ionization potential ) of the n ( n+1 ) electron system . sketch of the total energy profile as a function of number of electrons in a generic atomic system in contact with a reservoir . the bottom curve is simply the difference between the other two ( the lda energy and the `` exact '' result for an open system ) . ] exact dft correctly reproduce this behavior @xcite , which is instead not well described by the lda or gga approach , which produces total energy with unphysical curvatures for non integer occupation and spurious minima in correspondence of fractional occupation of the orbital of the atomic system . this leads to serious problems when one consider the dissociation limit of hetero - polar molecules or an open - shell atom in front of a metallic surface @xcite , and is at the heart of the lda / gga failure in the description of strongly correlated systems@xcite . the unphysical curvature is associated basically to the incorrect treatment by lda or gga of the self - interaction of the partially occupied kohn - sham orbital that gives a non - linear contribution to the total energy with respect to orbital occupation ( with mainly a quadratic term coming from the hartree energy not canceled properly in the exchange - correlation term ) . nevertheless , it is well known @xcite that total energy differences between different states can be reproduced quite accurately by the lda ( or gga ) approach , if the occupation of the orbitals is _ constrained _ to assume integer values . as an alternative , we can recover the physical situation ( an approximately piece - wise linear total energy curve ) by adding a correction to the lda total energy which vanishes for integer number of electrons and eliminates the curvature of the lda energy profile in every interval with fractional occupation ( bottom curve of fig . [ parab ] ) . but this is exactly the kind of correction that is provided by eq . [ our1 ] if the numerical value of the parameter @xmath0 is set equal to the curvature of the lda ( gga ) energy profile . this clarifies the meaning of the interaction parameter @xmath0 as the ( unphysical ) curvature of the lda energy as a function of @xmath59 which is associated with the spurious self - interaction of the fractional electron injected into the system . from this analysis it is clear that the numerical value of @xmath0 will depend in general not only , as noted in the preceding section , on the definition adopted for the occupation matrices but also on the particular approximate exchange - correlation functional to be corrected , and should _ vanish _ if the exact dft functional were used . the situation is of course more complicated in solids where fractional occupations of the atomic orbitals can occur due to hybridization of the localized atomic - like orbitals with the crystal environment and the _ unphysical _ part of the curvature has to be extracted from the total lda / gga energy , which contains also hybridization effects . in the next section this problem is discussed and a linear response approach to evaluate hubbard @xmath0 is proposed . following previous seminal works @xcite we compute @xmath0 by means of constrained - density - functional calculations @xcite . what we need is the total energy as a function of the localized - level occupations of the `` hubbard '' sites : @xmath62 = \min_{n({\bf r}),\alpha_i } \left \ { e [ n({\bf r } ) ] + \sum_i \alpha_i ( n_{i } - q_{i } ) \right \ } , \ ] ] where the constraints on the site occupations , @xmath63 s from eq . [ eq : ni ] , are applied employing the lagrange multipliers , @xmath64 s . from this dependence we can compute numerically the curvature of the total energy with respect to the variation , around the unconstrained values @xmath65 , of the occupation of one isolated site . a supercell approach is adopted in which occupation of one representative site in a sufficiently large supercell is changed leaving unchanged all other site occupations . this curvature contains the energy cost associated to the localization of an electron on the chosen site including all screening effects from the crystal environment , but it is not yet the hubbard @xmath0 we want to compute . in fact , had we computed the same quantity from the total energy of the non - interacting kohn - sham problem associated to the same system , @xmath66 = \min_{n({\bf r}),\alpha_i } \left\ { e^{ks } [ n({\bf r } ) ] + \sum_i \alpha^{ks}_i ( n_i - q_i ) \right \},\ ] ] we would have obtained a non vanishing results as well because by varying the site occupation a rehybridization of the localized orbitals with the other degrees of freedom is induced that gives rise to a non - linear change in the energy of the system . this curvature coming from rehybridization , originating from the non - interacting band structure but present also in the interacting case , has clearly nothing to do with the hubbard @xmath0 of the interacting system and should be subtracted from the total curvature : @xmath67}{\partial q_i^{2 } } - \frac{\partial^{2}e^{ks}[\{q_i\}]}{\partial q_i^{2 } } .\ ] ] in ref . @xcite anisimov and gunnarsson , in order to avoid dealing with the above mentioned non - interacting curvature , exploited the peculiarities of the lmto method , used in their calculation , and decoupled the chosen localized orbitals from the remainder of the crystal by suppressing in the lmto hamiltonian the corresponding hopping terms . this reduced the problem to the one of an isolated atom embedded in an artificially disconnected charge background . thanks to janak theorem @xcite the second order derivative of the total energy in eq . [ eq : eofn ] can then be recast as a first order derivative of the localized - level eigenvalue . in our approach the role played in refs . @xcite by the eigenvalue of the artificially isolated atom is taken by the lagrange multiplier , used to enforce level occupation@xcite : @xmath68}{\partial q_{i } } = - \alpha_{i } , \quad \frac{\partial^{2}e[\{q_{j}\}]}{\partial q_{i}^{2 } } = - \frac{\partial\alpha_{i}}{\partial q_{i}},\\ \nonumber & & \frac{\partial e^{ks}[\{q_{j}\}]}{\partial q_{i } } = - \alpha^{ks}_{i } , \quad \frac{\partial^{2}e^{ks}[\{q_{j}\}]}{\partial q_{i}^{2 } } = - \frac{\partial\alpha^{ks}_{i}}{\partial q_{i}}.\end{aligned}\ ] ] at variance with the original method of refs . @xcite , in our approach we need to compute and subtract the band - structure contribution , @xmath69 , from the total curvature , but , in return , hubbard @xmath0 is computed in exactly the same system to which it is going to be applied and the screening from the environment is more realistically included . the present method was inspired by the linear response scheme proposed by pickett and coworkers @xcite where however the role of the non - interacting curvature was not appreciated . in actual calculations constraining the localized orbital occupations is not very practical and it is easier to pass , via a legendre transform , to a representation where the independent variables are the @xmath64 s @xmath70 & = & \min_{n({\bf r } ) } \left \ { e [ n({\bf r } ) ] + \sum_i \alpha_i \ ; n_{i } \right\ } , \\ \nonumber e^{ks}[\{\alpha^{ks}_i\ } ] & = & \min_{n({\bf r } ) } \left \ { e^{ks } [ n({\bf r } ) ] + \sum_i \alpha^{ks}_i n_{i } \right\}.\end{aligned}\ ] ] variation of these functionals with respect to wavefunctions shows that the effect of the @xmath64 s is to add to the single particle potential a term , @xmath71 ( or @xmath72 for the non - interacting case ) , where localized potential shifts of strength @xmath64 ( @xmath73 ) are applied to the localized levels associated to site @xmath7 . it is useful to introduce the ( interacting and non - interacting ) density response functions of the system with respect to these localized perturbations : @xmath74 using this response - function language , the effective interaction parameter @xmath0 associated to site @xmath7 can be recast as : @xmath75 that is reminiscent of the well known random - phase approximation @xcite in linear response theory giving the interacting density response in terms of the non - interacting one and the coulomb kernel . a similar result is obtained within dft linear response @xcite where the interaction kernel also contains an exchange - correlation part . the response functions , eq . [ chidef ] , needed in eq . [ ueff1 ] are computed taking numerical derivatives . we perform a well converged lda calculation for the unconstrained system ( @xmath76 for all sites in the supercell ) and starting from its self - consistent potential we add small ( positive and negative ) potential shifts on each non equivalent `` hubbard '' site @xmath23 and compute the variation of the occupations , @xmath77 s , for all sites in the supercell in two ways : @xmath28 letting the kohn - sham potential of the system readjust self - consistently to optimally screen the localized perturbation , @xmath78 , and @xmath29 without allowing this screening . this latter result is nothing but the variation computed from the first iteration in the self - consistent cycle leading eventually to the former ( screened ) results . the site - occupation derivatives calculated according to @xmath28 and @xmath29 give the matrices @xmath79 and @xmath80 respectively . before moving to examine some specific examples in the next section , let s end the present one discussing a few additional technical points . as mentioned earlier , hubbard @xmath0 is computed , ideally , from variation of the site occupation of a single site in an infinite crystal and in practice adopting a supercell approach where periodically repeated sites are perturbed coherently . in order to speed up the convergence of the computed @xmath0 with supercell size it may result useful to enforce explicitly charge neutrality for the perturbation , that is to be introduced in the response functions , thus enhancing its local character and reduce the interaction with its periodic images . in this procedure we introduce in the response functions , @xmath81 and @xmath82 , in addition to the degrees of freedom associated to the localized sites also a `` delocalized background '' representing all other degrees of freedom in the system . this translates in one more column and row in the response matrices , whose elements are determined imposing overall charge neutrality of the perturbed system for all localized perturbations , ( @xmath83 ) and absence of any charge density variation upon perturbing the system with a constant potential ( @xmath84 ) . from a mathematical point of view both @xmath81 and @xmath85 acquire a null eigenvalue , corresponding to a constant potential shift , and the needed inversions in eq . [ ueff1 ] must be taken with care . it can be shown that their singularities cancel out when computing the difference @xmath86 and the final result is well defined . we stress that in the limit of infinitely large supercell the coupling with the background gives no contribution to the computed @xmath0 , but we found that this limit is approached more rapidly when this additional degrees of freedom is included . in the same spirit we found that the spatial locality of the response matrices can be rather different from the one of their inverse and a supercell sufficient to decouple the periodically repeated response may be too small to describe correctly the inverse in eq . [ ueff1 ] . as a practical procedure , therefore , after evaluating the response function matrices in a given supercell , we extrapolate the result to much larger supercells assuming that the most important matrix elements in @xmath85 and @xmath81 involve the atoms in the few nearest coordination - shells accessible in the original supercell . the corresponding matrix elements of the larger supercell are filled with the values extracted from the smaller one while all other , more distant , interactions are neglected . again , when a sufficiently large supercell to extract the matrix elements of the response functions is considered , the effect of this extrapolation vanishes , but , as we will see in the following , this scheme capture a large fraction of the system - size dependence of the calculated @xmath0 and it may allow to reach more rapidly the converged result . as a final remark we notice that the electronic structure of a system described within the lda+u approach may largely differ from the one obtained within the lda used to compute @xmath0 . in a more refined approach one might seek internal consistency between the band structure used in the calculation of @xmath0 and the one obtained using it . we have not addressed this issue here , but one can imagine performing the same type of analysis leading to the @xmath0 determination for a functional already containing an lda+u correction . the computed @xmath0 would in that case be a correction to be added to the original @xmath0 and internal consistency would be reached when the correction vanishes . in their seminal paper anisimov and gunnarsson @xcite computed the effective on site coulomb interaction between the localized electrons in metallic fe and ce . for ce the calculated coulomb interaction was about 6 ev in good agreement with empirical and experimental estimates ranging from 5 to 7 ev @xcite , while the result for fe ( also about 6 ev ) was surprisingly high since @xmath0 was expected to be in the range of 1 - 2 ev for elemental transition metals , with the exception of ni @xcite . let us apply the present approach to these two system , starting with iron . in its ground state elemental iron has a ferromagnetic ( fm ) spin arrangement and a body - centered cubic ( bcc ) structure . gradient corrected exchange - correlation functional are needed in order to stabilize the experimental structure as compared with non - magnetic face - centered cubic ( fcc ) structure preferred by lda . the perdew - burke - ernzherof ( pbe ) @xcite gga functional was employed here . iron ions were represented by ultrasoft pseudopotential and kinetic energy cutoffs of 35 ry and 420 ry were adopted for wavefunction and charge density fourier expansion . brillouin zone integrations where performed using 8@xmath878@xmath878 monkhorst and pack special point grids @xcite using methfessel and paxton smearing technique @xcite with a smearing width of 0.005 ry in order to smooth the fermi distribution . the calculation of the effective hubbard @xmath0 followed the procedure outlined in preceding section : a supercell was selected containing a number of inequivalent iron atoms ; then , after a well converged self - consistent calculation , we applied to one of these atoms small , positive and negative , potential shifts , @xmath88 ( with @xmath890.2 - 0.5 ev ) , where @xmath90 is the projector on the localized @xmath13 electron of the selected atom . from the variation of the @xmath13-level occupations of all iron atoms in the cell one column of @xmath81 and @xmath82 response functions was extracted and all other matrix elements were reconstructed by symmetry , including the background as explained previously . hubbard @xmath0 was then calculated from eq . [ ueff1 ] . in order to describe response for an isolated perturbation four supercells were considered : _ i ) _ a simple cubic ( sc ) cell containing two inequivalent iron atoms , the perturbed atom and one of its nearest neighbors ; _ ii ) _ a 2@xmath872@xmath872 bcc supercell containing 8 inequivalent iron atoms , 4 in the nearest - neighbor shell of the perturbed atom and 3 belonging to the second shell of neighbors ; _ iii ) _ a 2@xmath872@xmath872 sc cell containing 16 atoms , including also some third nearest - neighbor atom and _ iv ) _ a 4@xmath874@xmath874 bbc supercell containing 64 inequivalent iron atoms ; we used this largest cell just to extrapolate the results from the smaller ones . the convergence properties of the effective @xmath0 of bulk iron with the size of the used supercell are shown in fig . [ unei ] . in metallic iron for different supercells . lines connect results from the cell - extrapolation procedure described in the text and different symbols correspond to inclusion of screening contributions up to the indicated shell of neighbors of the perturbed atom . ] parameter for iron . ] the hubbard @xmath0 obtained from the sc 2-atom cell , once inserted in the 64-atom supercell , captures most of the effective interaction ; second nearest neighbors shell brings some significant corrections to the final extrapolated result , while third nearest neighbor shell has a smaller effect . we believe that contributions from further neighbor rapidly vanish and that an accurate value of @xmath0 can be extracted from the sc supercell containing 16 atoms . the extrapolation from this cell to larger cells brings only minor variations which are within the finite numerical accuracy that we estimate within a fraction of an ev . from this analysis our estimate for the hubbard @xmath0 in elemental iron at the experimental lattice parameter is therefore 2.2 @xmath91 0.2 ev . this results is in very good agreement with the experimental estimates @xcite , but disagrees with anisimov and gunnarsson result @xcite . we can only recall here that many technical details differ in the two approaches . in particular _ i ) _ in the original approach the perturbed atom is disconnected from the rest of the crystal by removing all hopping terms , thus leaving a rather unphysical environment to perform the screening , while in our approach the actual system is allowed to screen the perturbation , _ ii ) _ the atomic sphere approximation ( asa ) was employed in the original lmto calculation while no shape approximation is made in our case . in order to further test our approach on this element we investigate the dependence of the hubbard parameter on crystal structure . the dependence of the calculated interaction parameter on the lattice spacing of the unit cell is shown in fig . [ uls ] where a marked increase of the hubbard @xmath0 can be observed when the lattice parameter is squeezed below its experimental value . despite this may appear counterintuitive , as correlation effects are expected to become less important when atoms gets closer , one should actually compare the increasing value of @xmath0 with the much steeper increase of bandwidth when reducing the interatomic distance . upon increase of the lattice parameter the hubbard parameter should approach the atomic limit that can be estimated from all - electron atomic calculations where the local neutrality of the metallic system is maintained : @xmath92 ev , in reasonable agreement with the results of fig . [ uls ] . using the calculated volume dependent hubbard @xmath0 parameter we have studied the effect of the lda+u approximation on the structural properties of iron . . comparison between the calculated lattice constant ( @xmath93 ) , bulk modulus ( @xmath94 ) and magnetic moment ( @xmath95 ) within several approximate dft schemes and experimental results quoted from @xcite . [ cols="<,^,^,^",options="header " , ] [ tab : faystrut ] although we did not studied other spin - configurations , magnetic properties seam to improve slightly in the gga+u approximation . the magnetic moment on each iron ( both fe1 and fe2 ) was found to be 3.9 @xmath96 , in closer agreement with the spin - only value ( 4 @xmath96 ) of the experimental result ( 4.4 @xmath96 ) than the one obtained by gga only ( 3.8 @xmath96 ) . this improvement is probably due to the enhanced atomic - like character of iron @xmath13 states , which is consequence of the gap opening . in conclusion , the gga+u provides a quite good description of structural , magnetic _ and _ electronic properties of fayalite , reproducing the observed insulating behavior with a reasonable value for its fundamental band gap . in this work we have reexamined the lda+u approximation to dft and a simplified rotational - invariant form of the functional was adopted . we then developed a method , based on a linear response approach , to calculate in an internally consistent way the interaction parameters entering the lda+u functional , without making aprioristic assumption about screening and/or basis set employed in the calculation . our methodology was then successfully tested on a few systems representative of normal and correlated metals , simple transition metal oxides and iron silicates . in all cases we obtained rather accurate results indicating that our scheme allows us to study both electronic and structural properties of strongly correlated material on equal footing , without resorting to any empirical parameter adjustment . lichtenstein , m.i . katsnelson , _ phys . rev . b _ * 57 * , 6884 ( 1998 ) ; m.i . katsnelson , and a.i . lichtenstein , _ phys . b _ * 61 * , 8906 ( 2000 ) ; a.i.lichtenstein , m.i . katsnelson , and g. kotliar , _ phys . rev.letters_ * 87 * , 067295 ( 2001 ) . a. baldereschi , _ phys . b _ * 7 * , 5212 ( 1973 ) ; d.j . chadi , and m.l . cohen , _ phys . b _ * 8 * , 5747 ( 1973 ) ; h.j . monkhorst , and j.d . pack , _ phys . b _ * 13 * , 5188 ( 1976 ) ; j.d . pack , and h.j . monkhorst _ phys . rev . b _ * 16 * , 1748 ( 1977 ) . b. johansson , _ philos . _ * 30 * , 469 ( 1976 ) ; j.w . allen and r.m . martin , _ phys . rev . letters _ * 49 * , 1106 ( 1982 ) ; j.w . allen and l.z . liu , _ phys . rev.b _ * 46 * , 5047 ( 1992 ) ; i.s . sandalov , o. hjortstam , b. johansson , and o. eriksson , _ phys . b _ * 51 * , 13987 ( 1995 ) . cohen , s. gramsh , g. steinle - neumann and l.stixrude , in _ international school of physics `` enrico fermi '' ( couse 147 , varenna 2001 ) ; high pressure phenomena _ , hemley and g.l . chiarotti editors ( ios press , amsterdam , 2002 ) , p. 215 . h. kuhlenbeck , g. odrfer , r. jaeger , g. illing , m. menges , th . mull , h.j . freund , m. phlchen , v. staemmler , s. witzel , c. scharfschwerdt , k. wennemann , t. liedtke , and m. neumann , _ phys . b _ * 43 * , 1969 ( 1991 ) .	in this work we reexamine the lda+u method of anisimov and coworkers in the framework of a plane - wave pseudopotential approach . a simplified rotational - invariant formulation is adopted . the calculation of the hubbard @xmath0 entering the expression of the functional is discussed and a linear response approach is proposed that is internally consistent with the chosen definition for the occupation matrix of the relevant localized orbitals . in this way we obtain a scheme whose functionality should not depend strongly on the particular implementation of the model in ab - initio calculations . we demonstrate the accuracy of the method , computing structural and electronic properties of a few systems including transition and rare - earth correlated metals , transition metal monoxides and iron - silicate .
non - adiabatic systems , where the quantum mechanics of electrons and that of ions are deeply entangled , are among the most fascinating in physics . in condensed matter , entire classes of phenomena such as superconductivity or charge - density waves are of this nature . dynamical jahn - teller ( djt ) centers ( molecules and ions , impurities , nuclei ) constitute perhaps the simplest `` zero - dimensional '' example of nonadiabatic behavior , and have attracted attention for over half a century@xcite . an amusing anomaly has long been pointed out@xcite in certain simple djt centers , where an @xmath12 electronic doublet is jahn - teller ( jt ) coupled to a vibrational ( `` pseudo - rotational '' ) @xmath13 doublet . in modern language , this anomaly can be seen as a berry phase @xcite . in simple words , what happens in these systems is that there exists a class of closed paths ( loops ) in the @xmath13 coordinate space , regarded as classical , adiabatically transporting the electronic ground state from an initial value @xmath14 to a final value @xmath15 , instead of into @xmath14 itself . there is a berry phase @xmath16 , whose origin lies in the fact that the point of electronic degeneracy ( the `` conical point''@xcite ) is irreducibly surrounded by the trajectory in vibron space . this anomaly which is present at the adiabatic level , where the ionic coordinated are classical , has very interesting consequences when the full joint ion - electron problem is finally quantized . in the @xmath17 case cited above , there appears a new conserved `` pseudo - angular momentum '' quantum number , whose values are half - integer@xcite . consequently , all vibronic levels become twofold degenerate , as appropriate to a rotor coupled to a pseudospin @xmath18@xcite . it is an interesting question therefore , to inquire whether other nontrivial cases of djt are realized in nature , which also possess a berry phase . high - symmetry polyatomic molecules are an obvious place to look at . the fullerene molecule , c@xmath11 , has a very high @xmath19 symmetry , and a nondegenerate ground state . because of high symmetry , however , fullerene ions , both negative and positive have a high orbital degeneracy , and are therefore good candidates . in our previous work , we have studied@xcite the djt problem of negatively - charged fullerene ions , where @xmath20 electrons partly fill the @xmath21 lowest unoccupied molecular orbital ( lumo ) of the neutral molecule . this electronic state couples linearly to the fivefold degenerate @xmath22 vibrations . for odd electron number @xmath23 , the spectrum of this coupled dynamical @xmath24 system is affected by anomalies which can again be usefully described in terms of a berry phase@xcite . among notable consequences , we have predicted the dominance of @xmath25-wave attachment for low - energy electrons@xcite , a negative pairing energy for odd @xmath23@xcite , characteristic spectroscopic anomalies@xcite and a mechanism for kinematic electron pairing@xcite from the mathematical point of view , nevertheless , the djt of c@xmath26 , namely @xmath27 , was not particularly new : the analytical machinery was basically the same as that previously studied by obrien@xcite for the @xmath28 equal coupling case . in the present paper , we move on to analyze the problem of the jt effect in the _ positively _ charged fullerene ions c@xmath29 , inquiring in particular about the possible relevance of a berry phase in these systems too . as it will turn out , the answer is again affirmative ( though with some qualifications ) . moreover , we shall deal in this case with the less trivial jt problems @xmath30 and @xmath31 , where the search for an adiabatic anomaly will now be done afresh . fullerene cations c@xmath29 are obtained by removing @xmath32 electrons from ( i.e. adding @xmath32 holes to ) the fivefold degenerate @xmath8 highest occupied molecular orbital ( homo)@xcite . our approach allows us to treat the one - hole case ( @xmath33 ) as well as the two - holes ( @xmath34 ) singlet case , in a way which is essentially analytical , albeit for a very idealized set of parameters . a hole in the homo interacts with the c@xmath11 vibrational modes . we focus on modes which couple linearly with the hole density fluctuations ( which in turn are of course quadratic in the hole wave function ) . hence , the icosahedral symmetry of the molecule restricts the linearly coupled modes to those of the same symmetry as the irreducible representation of the group ( @xmath19 ) contained in the symmetric product of a hole representation with itself@xcite : @xmath35 the @xmath19 group is a subgroup of the full rotation group o(3 ) . therefore , a general representation of o(3 ) decomposes into a finite number of representations of @xmath19@xcite . in particular , the @xmath4 @xmath36 representations correspond to @xmath37 in icosahedral symmetry , and @xmath5 @xmath38 decomposes into @xmath39 . equation ( [ couplingstructure ] ) may be rewritten , in o(3 ) notation , @xmath40 in order to take advantage of spherical symmetry , and as a simplifying approximation at this initial stage , we neglect the icosahedral details of the molecular shape , and we study the hole - vibron coupling within the o(3 ) scheme , as though c@xmath11 was a perfect sphere@xcite . we study here therefore the problem of @xmath41 holes coupled to @xmath42 , @xmath43 and @xmath44 vibrations . the coupling with a @xmath42 vibration is a trivial polaronic problem , solved exactly as a displaced oscillator , and is irrelevant here , since it shifts energies , but does not cause splittings . we concentrate on quadrupolar ( @xmath43 ) and hexadecapolar ( @xmath44 ) modes , which lead instead to a non - trivial jt effect . in order to simplify the notation , we shall switch from the o(3 ) to the so(3 ) notation , omitting therefore the @xmath45 signs for the inversion . the hamiltonian we consider has the following structure : @xmath46 where @xmath47 describes a free ( uncoupled ) @xmath48 hole and @xmath49 , @xmath50 vibrations , and @xmath51 the hole - vibron interaction , namely @xmath52 c_{m_2,\sigma}^{\dagger } c_{m_1+m_2,\sigma } ~ . \nonumber\end{aligned}\ ] ] here , @xmath53 are the energies of the ( harmonic ) vibrations @xmath54 , @xmath55 are dimensionless coupling parameters , @xmath56 are clebsch - gordan coefficients , and there is no loss of generality in choosing the energy zero so that @xmath57 . for simplicity , we shall assume here a single @xmath4 and a single @xmath5 mode corresponding to @xmath58 and @xmath59 vibrons respectively . in real c@xmath11 , the spectrum is more complex ( see tab . [ experener : table ] ) , involving 8 @xmath22 and 6 @xmath9 modes . of these , only two @xmath22 modes , namely @xmath60 and @xmath61 are actually derived from a @xmath48 spherical vibration , while only two doublets , namely @xmath62 and @xmath63 are of @xmath5 origin@xcite . the other modes derived from different angular momenta ( @xmath64 ) do not belong to a complete @xmath65-multiplet , and can not be simply treated in spherical symmetry . our work is therefore incomplete and of purely qualitative value . nevertheless , we expect some of the features of the present simplified analysis to be of more general relevance . for example , close pairs of modes , such as @xmath66 , or @xmath67 , have no reason to behave qualitatively differently from our @xmath50 mode , in spite of their different @xmath68 and @xmath69 origins . we will also find a `` conservation of berry phases '' : if a jt - coupled problem has a berry phase , the extension of the vibron space with the addition of a mode that does not carry a berry phase , leaves intact the original berry phase . in our case , we will show that the berry phase is coming exclusively from the @xmath70 coupling , the addition of the @xmath22 modes changing the dynamical properties , but not the topological ones . according to the different relative values of the four parameters ( @xmath71 ) in @xmath72 , different regimes can be attained . in particular , when the coupling is extremely strong ( @xmath73 ) , the jt distortion is large , higher order terms ( neglected in ( [ hesplicita ] ) ) will in principle become relevant . in such a case , the final state tends to approach closely a static jt symmetry - broken distortion . when the coupling is instead large but finite , there is a semiclassical region , where quantum fluctuations are able to connect classical equivalent valleys _ via _ tunneling , and so restore the full symmetry while also generating characteristic low - energy tunneling excitations in the spectrum . the intermediate coupling region @xmath74 is , as usual , the less predictable . however , the weak coupling limit ( @xmath75 ) finally simplifies again into a perturbative problem . if either @xmath76 or @xmath77 alternatively vanish , we will have the special cases of @xmath78 or @xmath79 jt respectively . a special case we shall consider in greatest detail below is that of equal coupling of degenerate oscillators @xmath80 , @xmath6 . this paper is organized as follows . the weak - coupling region @xmath81 is discussed in sect . [ perth : sect ] , obtaining approximate analytic expressions for the splittings and shifts in the spectrum . next , we shall move to the opposite , strong - coupling case . here , the classical jt minima of the born - oppenheimer potential energy surfaces are studied first ( sect.[topology : sect ] ) ; then , in sect . [ semiclass : sect ] , the ionic kinetic energy is introduced semiclassically to obtain a description of the low - energy quantum excitations in the strong - coupling limit . exact numerical diagonalizations will be used to shed light on a few particular features of the intermediate - coupling regime . finally , the relevance of these preliminary theoretical results to the physics of fullerene cations is discussed in sect . [ discussion : sect ] . the weak - coupling limit in a jt problem is a coupled situation in which both the electronic and the vibronic degrees of freedom need to be treated on equal footing , as equally `` fast '' . their cooperative dynamics is strictly regulated by quantum mechanics , and no `` classical degree of freedom '' can be singled out in this limit , as opposed to the berry - phase semiclassical approach . nonetheless , the limit of small @xmath82 is easily treated by perturbation theory , and the results are useful in providing an exact match to the semiclassical spectrum on the opposite side of the range of coupling . the jt coupling ( [ hesplicita ] ) splits and shifts the harmonic levels to second - order in @xmath55 . we compute here these shifts for both the fivefold - degenerate ground state and the 70-fold degenerate one - vibron excitation in the 1-hole case and for equal frequencies @xmath83 . as we have two small parameters , @xmath77 and @xmath76 , and two different perturbing hamiltonians , @xmath84 and @xmath85 , there is an infinite number of choices of the small perturbing parameter , depending for example on the ratio @xmath86 . we compute the energy shifts for three special cases : @xmath87 , i.e. perturbing with respect to @xmath84 , @xmath88 , using @xmath85 alone , and finally @xmath89 , the equal coupling case . the coefficients for the three cases are collected in tab . [ perturbcoef : table ] , along with the appropriate so(3 ) symmetry labels . the energy of a state ( in units of @xmath90 ) is thus given by @xmath91 where the number of vibrons is one for all the states except for the ground state at the first line . we note on the second column five states , in the 1-vibron multiplet , whose shifts are @xmath92 , i.e. the same as for the ground state . these states are readily identified as the 45-fold degenerate state corresponding to the @xmath50 vibron fundamental state ( here uncoupled , since @xmath87 ) . parallel considerations apply to the uncoupled @xmath48 vibron in the third column , recognizable by the shifts of @xmath93 . the equal - coupling case is interesting from the point of view of symmetry . pooler studied the hamiltonian ( [ hesplicita ] ) from an algebraic point of view@xcite , discovering that the equal frequencies equal coupling case @xmath80 , @xmath6 , is actually characterized not only by the obvious so(3 ) symmetry , but also by rotation symmetry in a 5-dimensional space ( so(5 ) group ) . in the language of the so(5 ) irreducible representations , characterized by two quantum numbers [ @xmath94,@xmath95@xcite , the @xmath48 hole is classified as a [ 1,0 ] irreducible representation of so(5 ) , while the 14-fold degenerate vibration ( @xmath96 ) is a realization of the [ 2,0 ] representation of so(5)@xcite . in the fourth column of table [ perturbcoef : table ] , we note that the one - vibron level in the equal - coupling case splits into three levels with degeneracies 30 , 35 and 5 respectively . the levels of degeneracy 5 and 30 are readily seen as representations [ 1,0 ] and [ 3,0 ] of so(5)@xcite . the 35-fold degenerate level , by table 10 - 3 of reference@xcite is identified with a [ 1,2 ] representation , as indicated in the last column of table [ perturbcoef : table ] . as this large symmetry is an exact feature of the hamiltonian , it is not limited to the perturbative regime , but will be carried along to all values of ( equal- ) coupling , and to the semiclassical limit ( sect.[semiclass : sect ] ) in particular . to conclude , we note that , while the @xmath97 levels are exactly the same in the three perturbative regimes summarized in tab.[perturbcoef : table ] , the @xmath98 levels ( marked in tab.[perturbcoef : table ] with symbols @xmath99 ) , in the three cases , are instead different linear combinations of the same states : this fact has a very simple explanation . the @xmath100 levels uniquely come from the coupling of the @xmath48 hole to the @xmath48 vibron , and the @xmath101 levels derive from the coupling of the @xmath48 hole to the @xmath50 vibron . the @xmath98 levels can come from either of the previous couplings , and there is no _ a priori _ reason why they should not mix among themselves . in this section we want to understand if the jt hole - vibron coupling in c@xmath11 positive ions possesses a berry phase or not . in order to do that , the canonical steps are the following : ( i ) find the manifold , in the space of classical ionic coordinates , formed by the set of equivalent minima of hamiltonian ( [ hesplicita ] ) ( the minima of the born - oppenheimer ( bo ) energy surface ) ; ( ii ) study the adiabatic transport of the electronic ground state wave - function in this case a unit vector spanning a five - dimensional sphere while the ionic coordinates are taken along a closed path on their manifold . representation ( [ hesplicita ] ) of hamiltonian ( [ hschematica ] ) is inconvenient for this purpose , and we first perform a transformation of the electronic and vibronic operators to a suitable representation . we start defining a vector of hole - destruction operators , @xmath102 in terms of this vector , the interaction hamiltonian can be written as @xmath103 where @xmath104 indicates the adjoint transpose of ( [ vectc ] ) , and @xmath105 and @xmath106 are @xmath107 matrices whose elements are combinations of vibrational coordinates . next , we apply the following transformation to the electronic operators @xmath108 the transformed vector @xmath109 contains the phase factors necessary to give a suitable symmetry to the transformed @xmath110-matrices . the final step converts the vibron operators into their real coordinate representation : @xmath111 where @xmath112 and the masses of both oscillators are taken as unity . we rewrite @xmath47 and @xmath51 after the final transformation as @xmath113 where the matrices @xmath114 and @xmath115 , the transformed of @xmath105 and @xmath106 under ( [ c_transf ] ) , expressed in terms of @xmath116 ( [ b_transf ] ) read @xmath117 and @xmath118 the classical ( purely static ) solution of the jt problem consists in finding the minimum of the born - oppenheimer ( bo ) potential energy defined as the sum of the potential term in ( [ h0 ] ) plus the lowest eigenvalue of the hole - vibron interaction matrix @xmath119 in the ( 5 + 9)-dimensional space of the classical @xmath116 coordinates . this classical minimum is generally not unique . more typically , there will be continuous sets of minima , forming well defined manifolds embedded in the multi - dimensional coordinates space@xcite . we call one such set of minima of the adiabatic potential jahn - teller manifold ( jtm ) . the topology of this manifold crucially affects the possible presence of a berry phase . we study the jtm following the approach of pik and pryce@xcite , more recently reformulated by ceulemans in a clear and useful form@xcite . consider a generic one - fermion state @xmath120 where without loss of generality , the @xmath121 can be taken real , with the normalization constraint @xmath122 the quantum amplitudes @xmath123 live therefore on a @xmath124 sphere in 5-dimensions , which we indicate as the electronic sphere ( es ) . within this notation , it is convenient to rewrite the classical potential energy as @xmath125 at fixed distortion @xmath126 _ and _ electron state @xmath127 . minimization of @xmath128 with respect to @xmath123 yields , by definition , the ( adiabatic ) bo potential . if we keep @xmath123 fixed instead , and minimize ( [ veta ] ) with respect to @xmath129 , then we get the extremal points of the bo potential in configuration space as @xmath130 inserting ( [ q_of_min ] ) in ( [ veta ] ) , we obtain a function @xmath131 of the electronic coordinates alone , hence a function defined on the es , not on the space of distortion like the bo potential . we still have to ensure that the electronic vector is an eigenvector of the interaction hamiltonian . ceulemans@xcite realized that this condition is equivalent to a second minimization procedure , on the es ; moreover , he recognized that the @xmath131 function is isoextremal to the bo surface , that is , its extremal points on the es correspond ( in a 2:1 way ) to extremal points of the bo surface . then the jtm has the same symmetry properties of the set of minima of @xmath131 . we apply this procedure of reverse - order minimization to the present system . we list here the distortions @xmath132 and @xmath133 which minimize @xmath128 at fixed @xmath134 on the es : @xmath135 @xmath136 we can choose a ( hyper - polar ) representation for the electronic coordinates such that the unitarity condition ( [ normalization ] ) is automatically satisfied @xmath137 here , @xmath138 i=1,2,3 are the azimuthal angles and @xmath139 the anomaly . substituting ( [ etas ] ) in ( [ q2 ] ) and ( [ q4 ] ) , and then back into ( [ veta ] ) we obtain an expression for the minimum @xmath140 of @xmath131 : @xmath141 and for the jtm coordinates @xmath142 as eq . ( [ ejt ] ) shows , @xmath140 is independent of the @xmath143 parameters , and is therefore a _ constant on the whole _ this is a nontrivial consequence of the large amount of geometrical freedom which is allowed to the molecule by the large degeneracy of the vibron modes : on the contrary , when the dimension of @xmath144-space is less than that of the es ( like for example in the @xmath145 jt system relevant for the coupling of @xmath8 with the @xmath146 mode , which has no @xmath22 partner ) , @xmath140 is not flat , but is characterized by some amount of corrugation . the isoextremality property of the @xmath147 implies that the jtm , must have locally the same structure as the es , here @xmath148 , a sphere in five dimensions . however , upon closer scrutiny , we note a topological difference between the es and the jtm , due to the 2:1 correspondence mentioned above . in particular , the es is a projective space , as opposite points on the sphere correspond to the same point in the @xmath149-coordinates space ( they belong to the same `` ray '' ) . this is equivalent to the fact that normalized eigenvectors of a symmetric matrix are defined up to a sign . the jtm does not share this property , since there are no pairs of points on the jtm that generate the same @xmath150 . here is the source of the berry phase : to each path in the jtm there corresponds a path on the es , but the two are topologically different . in particular , a loop in the jtm may correspond either to a closed path , or to a path from a point to its opposite ( antipode ) on the es . in general , the latter path can not be reduced to a point , and there will be a topological effect ( berry phase ) . this fact is illustrated in fig . [ es : fig ] on an ordinary @xmath151 sphere , in 3-dimensional space , representing the es . the line @xmath152 corresponds to a closed loop on the jtm : there is no way to make point b coincide with point a in such a way that then we can just shrink the path and have it coincide to a trivial one such as @xmath153 . however , we are going to encounter situations where apparently nontrivial loops become shrinkable to a point _ via _ singularities in the mapping between the es and the jtm . of the most general set of coupling parameters , we consider three special cases : ( a ) @xmath154 , i.e. @xmath155 , @xmath156 ; ( b ) @xmath157 , i.e. @xmath88 , @xmath158 ; and finally ( c ) @xmath147 , i.e. @xmath80 , @xmath159 . \(a ) @xmath154 . here , the @xmath160 mapping is given by eq . ( [ q2 ] ) and ( [ etas ] ) . a class of closed paths on @xmath149-manifold does , as always , transport @xmath150 into its antipode @xmath161 . however , we find in this case that a whole ( `` equatorial '' ) line joining two antipodes on the es , defined by @xmath162 ( i.e. @xmath163 ) and parameterized by @xmath164 , in fact corresponds to a _ point @xmath165 in @xmath149-space . the antipodes are therefore topologically joined , and there are no loops on the jtm that can not be deformed to a point . hence , there is no berry phase in this case . \(b ) @xmath157 . the @xmath160 mapping is now through eq . ( [ q4 ] ) and ( [ etas ] ) . a closed loop on the jtm transports @xmath150 into @xmath161 if it encircles a conical intersection in configuration space . this is clearly the case of the line described in ( a ) , here corresponding to @xmath166 now only the @xmath144 antipodes are topologically equivalent , while the loops that join them are nontrivial : there is a berry phase of @xmath7 in the electronic transport in this case . \(c ) @xmath167 . in this case the @xmath160 mapping passes through both eq.([q2 ] ) and eq . ( [ q4 ] ) . the berry phase of case ( b ) is therefore conserved upon addition of the @xmath48 mode . this case has an instructive analogy to simpler jt - coupled systems . in case ( c ) above , we add to a berry - phase entangled djt system an additional @xmath4 vibron mode , also coupled to the same electronic state . it is interesting to consider what happens to the berry phase when the added mode , on its own , carries no berry phase . the enlargement of the @xmath168-space introduces a large freedom for creating new loops in @xmath168 . one might in principle think that some of these loops could manage to avoid the conical intersections of bo levels , thus healing the connectivity properties of the jtm , and short - circuiting out the berry phase . we found instead that this is not the case , and @xmath169 retains the berry phase of @xmath157 . an example of analogous situation is the @xmath170 problem , extensively studied in the literature@xcite : the @xmath171 problem is berry - phase entangled , while the @xmath172 is not . it is well known that @xmath170 problem retains the same topological properties , including the berry phase of @xmath7 , of the @xmath173 system , the effect of the @xmath13 mode consisting essentially only in flattening the jtm , and removing inter - minima barriers . the reason why new loops seem to be ineffective in destroying the berry phase in the total vibron space is that this space is the cartesian product of the single - mode spaces : the degeneracy points , opening `` holes '' in the space of the first vibron , propagate ( locally , at least ) in the global space along the second - vibron directions . the global conservation of connectivity in the whole product space is a nontrivial topological problem , whose general outcome is not known to us . in both examples above , we simply note that things go as if the berry phase carried by the first mode is strictly conserved . in the present situation , the coupled @xmath48 vibrational mode does not present any topological effect ( a ) , whereas the @xmath50 , as seen in case ( b ) , carries a berry phase . thus the overall @xmath174 system is affected by topological effects , and this remains true even for different couplings @xmath175 . the fullerene cations provide a third example of this behavior . the @xmath176 is naturally resolved in its icosahedral composition @xmath145 plus @xmath30 . in the light of the above considerations , we expect that in this system the berry phase should be carried by the @xmath177 mode@xcite . that this is indeed the case , can thus be seen as a natural consequence of the absence of berry phases in @xmath30 ( i.e. @xmath178 ) , and its presence in @xmath79 . so far we have treated the ions as classical . we study here the consequences of the discussed symmetries of the jtm and of the presence / absence of the berry phase on the quantum spectrum of the coupled system , by quantizing the semiclassical ionic motion . the general form of the ion kinetic energy is @xmath179 it is convenient to use the classical language for the derivation of the kinetic operator restricted to the semiclassical region : @xmath180 in the semiclassical limit , valid when @xmath55 are very large , the ionic motion is confined to the jtm , itself given by eq . ( [ jtm : eqn ] ) . hence , only the tangential part of the kinetic energy ( [ kintot ] ) is relevant , which , upon introduction of @xmath181 can be written @xmath182 where @xmath183 is the metric induced on the jtm by the electronic parametrization : @xmath184 not much progress is possible in this form , due to the general low - symmetry of the kinetic term . however , in the special case of equal frequencies ( @xmath185 : @xmath186 ) and equal couplings , where the true symmetry of the problem is so(5)@xcite , the metric is rewritten as : @xmath187 thus , the kinetic energy corresponds to that of a free particle on a sphere in 5 dimensions : @xmath188 @xmath189 is the inertial momentum of the system @xmath190 the corresponding laplace - beltrami operator ( describing the free motion of a quantum particle on the manifold parametrized by the @xmath138 variables ) is given by the formula@xcite @xmath191 where @xmath177 is the absolute value of the determinant of the metric tensor . this operator is the usual representation of the quadratic casimir operator of so(5 ) on the coset manifold so(5)/so(4 ) , which is actually the sphere @xmath148 : this is a consequence of the exact so(5 ) symmetry of the equal - coupling case . the particle moves freely on the jtm , with harmonic oscillations orthogonal to it . the corresponding energy levels are therefore @xmath192 @xmath140 is the static jt energy ( [ ejt ] ) . the last term is due to the vibrations orthogonal to the jtm . the @xmath193 term corresponds to the eigenvalues of the laplace - beltrami operator on the sphere ; @xmath94 is the label of the angular hyperharmonics on @xmath148 , that realize the @xmath194 $ ] representations of so(5 ) , whose degeneracy is given by @xmath195 the berry phase imposes precise selection rules on these levels : since the one - hole state changes sign after a closed loop around the jtm , and the overall wave function needs to be single - valued , also the vibrational wave function has to change sign . odd-@xmath94 states have this properties , and these are the allowed states that survive the berry phase selection rule . this implies that the ground state is @xmath196 ( fivefold degenerate ) and the first excited state is @xmath197 ( 30-fold degenerate ) . results of exact diagonalization on a truncated basis@xcite @xmath198 shown in fig . [ 2cross2 + 4:fig ] , confirm the spectral structure described here for the semiclassic limit . at all values of the coupling @xmath199 , as expected , the degeneracies are those characteristic of so(5 ) . we have done also numerical diagonalizations in the more general case of different coupling and/or frequencies : the spectra , as expected , display lower so(3 ) symmetry , but a qualitatively similar behavior . in the spin - singlet @xmath23=2 holes case , @xmath2c@xmath3 we find two main differences with the @xmath23=1 @xmath0c@xmath1 case . first , the average distortion @xmath200 is twice as large as in the @xmath23=1 case , @xmath189 and the @xmath140 energy being 4 times as large . secondly , and most importantly , there is a change in the selection rules . the two holes live in the same single - hole eigenvector of @xmath201 , the electronic wave function being the product of two equivalent eigenfunctions for the space coordinates and an antisymmetric part for the spin degrees of freedom . since each hole state changes sign after a closed loop in the jtm , the overall hole state does not . thus the vibrational wave function must also be single - valued in order to have an overall single - valued wave function for the system . as a consequence , only even-@xmath94 states survive : in particular , the ground state is @xmath202 ( non - degenerate ) and the first excited state is @xmath203 ( 14-fold degenerate ) . exact diagonalization results , shown in fig . [ 2cross2 + 4:fig](b ) , confirm that the symmetry and , to a certain degree the spectral structure described here for the semiclassic limit are retained in the weak - coupling limit . in the more general non - equal coupling case @xmath206 , including also the @xmath88 case , we could try to use the metric as for the equal coupling in order to obtain an analogous expression for the kinetic energy . unfortunately , when the couplings are different , we do not get a treatable expression like ( [ ecinetica ] ) , and we can not follow the same route for quantizing the system . we recall instead that the structure of the jtm is the same as that of es , and in particular that so(5 ) symmetry is retained . the expression of the minimum potential energy ( [ ejt ] ) suggests a flat jt trough , for any values of the frequencies and couplings . therefore , it is likely that the free motion of the particle on the jtm is described again by the quadratic casimir operator of so(5 ) , retaining for the low - energy excitations a spectral structure similar to eq . ( [ levels24 ] ) . however , although the full so(5 ) symmetry is retained in the @xmath207 semiclassical limit , for any non - infinite value of the couplings , the so(5 ) representations will be split into representations of the so(3 ) group @xcite . from this point of view we can see the differences between the frequencies and between the coupling strengths as symmetry - reducing perturbations , that become effective when moving away from the strong - coupling limit . this case , corresponding to @xmath208 in a c@xmath209 cation , is very interesting because of the peculiar features described above , i.e. absence of topological singularities and , in particular , absence of a berry phase . furthermore , this case is worth additional analysis , since it provides an alternate limit of our approach for the different coupling case ( of which this is an extreme case ) . we treat again the system as a quantization problem on a five - dimensional sphere . in this case , it is possible to find the extremal eigenvalues of the interaction matrix @xmath210 , as solutions of the characteristic polynomial equation for @xmath114 : @xmath211 where @xmath212 \ . \label{f0}\end{aligned}\ ] ] this polynomial can be decomposed as @xmath213 = 0.\ ] ] the extremal eigenvalues of @xmath114 are @xmath214 in accord with the so(5 ) symmetry of the jtm in the 5-dimensional @xmath49 space . the three remaining intermediate eigenvalues can be expressed as @xmath215 where @xmath216 in order to obtain the eigenenergies of @xmath217 one has to multiply these eigenvalues by a factor @xmath218 . by inspection , @xmath219 are bound in the @xmath220 $ ] interval . however , it is possible to find many points on the jtm such that @xmath221 , for example the `` south pole '' @xmath222 and a one - dimensional manifold on the `` equator '' @xmath223 . since @xmath224 is constant and @xmath225 is analytical in @xmath226 , in particular around the tangency points where the two eigenvalues get degenerate , these degeneracies are not conical intersections of the two adiabatic surfaces . the two surfaces have therefore just a contact of second order at these points . the absence of conical intersections implies the absence of any topological effect@xcite . because of the absence of a berry phase , already discussed in sect.[topology : sect ] , the strong coupling behavior of the system , described in terms of the so(5 ) representations , is such that we expect a [ 0,0 ] ( non degenerate ) ground state , a [ 1,0 ] ( fivefold degenerate ) first excited state , a [ 2,0 ] ( 14-fold degenerate ) second excited state , and so on . moreover , absence of topological effects makes the two - holes case not essentially different from the one - hole problem in the strong coupling limit . of course , the static jt energy and inertial momentum is just four times larger than in the one hole case , but the level sequence in the strong - coupling limit is the same . coming back to the single - hole case , let us now move away from the strong - coupling limit , to discuss the more general case of an arbitrary coupling @xmath77 . we might navely expect , at first , the ground state symmetry to be preserved for arbitrary @xmath77 , as was the case in other djt problems examined previously@xcite . however , a closer inspection reveals that this is not true , and rather that there must be a level crossing as a function of @xmath77 . a first indication comes from the fact that the ground state must in fact be @xmath4 , i.e. fivefold degenerate , as @xmath227 , reflecting just the uncoupled hole ( @xmath8 ) degeneracy . a second hint comes from the tangency of the lowest with the second bo surface , noted above . could this tangency have a role in the expected level crossing from @xmath4 at @xmath228 to @xmath229 at @xmath207 ? to study this possibility , we resort to numerical diagonalization of hamiltonian ( [ hschematica ] ) on a truncated basis ( [ psibasis : eqn ] ) , for finite values of @xmath77 . as pointed out above , for finite coupling the symmetry of the system is strictly so(3 ) . therefore , we expect the large degeneracies characteristic of so(5 ) symmetry to build up from converging so(3 ) states for large @xmath77 . the [ 0,0 ] and [ 1,0 ] states map directly on the @xmath230 and @xmath48 so(3 ) representations respectively @xcite , while the so(3 ) composition of [ 2,0 ] is @xmath231 . therefore , we seek the convergence of two such levels in the strong - coupling limit . the numerical solution of the 1-hole problem ( see fig . [ 2cross2:fig](a ) ) , shows indeed the correct order of the levels for large couplings ( @xmath232 ) , whereas for @xmath2338 the [ 0,0 ] and [ 1,0 ] levels cross and become inverted . this level inversion can in fact be precisely seen as a consequence of the tangency of the two adiabatic potential surfaces described above . the two bo sheets lie close in a significant region of configuration space . treating the upper sheet as a perturbation for the eigenstates of the lower sheet , the energy of all these states is pushed up by this correction . the @xmath202 ( isotropic ) states feel more efficiently this perturbation than the @xmath234 hyperspherical harmonics , since they can arrange their nodes so as to avoid the tangency line . in conclusion , the effect of the tangency is to increase the energy of the @xmath202 level more than that of @xmath196 , so that , for coupling not too large , @xmath235 , the latter is the ground state . in the 2-holes ( spin singlet ) case ( fig . [ 2cross2:fig](b ) ) this inversion is not present , the ground state being nondegenerate ( @xmath236 ) for any coupling @xmath77 . again , from a weak - coupling approach , this is a simple consequence of the order of splitting of the d@xmath0-fermion spin - singlet states upon perturbative jt coupling with a @xmath48 vibration , that is ( from lowest to highest ) @xmath236 , @xmath237 , @xmath48 . from the strong - coupling side , the only observed crossover happens between the @xmath48 and @xmath237 low excitations above the ground - state . this means that the bo picture of the two holes in the same single - hole orbital , although appropriate to the strong - coupling limit , is rather poor for intermediate jt coupling , when the two fermions manage to arrange in a cooperative state preserving the global @xmath236 symmetry despite the tangency of the lowest bo sheets . looking at table [ experener : table ] we can see that there is a single @xmath9 left , that does not pair with any @xmath22 mode . we must therefore treat the coupling of mode @xmath146 according to the true icosahedral picture . actually the static @xmath145 jt problem has been treated in the literature@xcite . the set of absolute minima of the system if made up of ten isolated points in the four - dimensional @xmath9 space . tunnel splitting among these ten bo valleys give rise to vibronic states of symmetry @xmath239 , @xmath177 , and @xmath72 . in ref . @xcite the diagonalization of the corresponding matrix has been performed including the proper provision for the berry phase . the result is that the @xmath72 quintet is the true ground state , followed by the @xmath177 quartet and the @xmath239 singlet in the order . thus , the berry phase again acts in such a way to enforce a fivefold degenerate ground state all the way from weak to strong coupling , in analogy with the @xmath240 jt problem , and in contrast to the @xmath30 case discussed in sect . [ 2cross2 ] . this confirms that , as anticipated , the berry phase in the @xmath241 jt problem comes from @xmath145 sector . we have studied the symmetry aspects and the possible presence of a berry phase in c@xmath209 and c@xmath3 in the linear and spherical approximation . we find that while @xmath2c@xmath3 has no berry phase , @xmath0c@xmath1 does , and is very rich and intriguing . clearly , the present analysis is only a first step . in reality , fullerene is not a sphere , but a discrete icosahedron , and the @xmath50 modes are split into @xmath242 modes . nonetheless the real shape of the molecule can be seen as a perturbation on the spherical picture ( tab.[experener : table ] ) : the so(5 ) and so(3 ) levels are split but the selection rules are still present . the pik - pryce approach can still be applied in order to obtain the jtm in icosahedral symmetry@xcite . quite generally the bo potential will warp in such a way to originate a discrete set of minima , instead of a continuous 4-dimensional manifold . the molecular system will execute tunneling among these minima . however , as shown by ham@xcite , the berry phase imposes the degeneracy of the ground state ( irrespective of the presence and height of barriers between the bo minima ) , which still is fivefold degenerate ( @xmath243 ) . therefore , we expect the ground state symmetry to be @xmath244 for real c@xmath209 , which has a berry phase , and @xmath245 for singlet c@xmath3 , which does not . a quantitative study of the properties of c@xmath209 is clearly beyond the scope of the present work . in order to accomplish that , a full set of e - v couplings @xmath246 and @xmath247 will have to be known ( either from calculations , or fit to experiments ) , and inserted into a diagonalization procedure similar to that previously carried out in the case of c@xmath248@xcite . while that remains to be done , we can list , for the time being , some tentative physical implications of our results . \(i ) _ low - energy electron detachment_. in a way analogous to that proven for the case of c@xmath249 , one can surmize that c@xmath209 can not have any @xmath250 low - energy states . it follows that if an electron is detached from c@xmath11 into a low - energy state , the s - wave cross - section will be zero . and there will be a centrifugal barrier in detachment too , similar in nature to that pointed out for attachment@xcite . this effect could be related to the thermoionic anomalies pointed out by yeretzian _ et al._@xcite \(ii ) _ low - energy tunneling excitation_. in the spherical approximation , the lowest excitation of c@xmath11 is a 30-fold degenerate so(5 ) `` @xmath197 '' tunneling multiplet , as required by the berry phase . split , as it will be by the lower so(3 ) symmetry , as well as by the true @xmath19 effects , this multiplicity might still be identifiable spectroscopically . c@xmath11 photoemission spectrum_. in the valence photoemission spectrum of c@xmath251 c@xmath252 the electron kinetic energy distribution contains very direct information in the so - called final state approximation about vibron and hole - vibron coupling in c@xmath209 . the outgoing electron shakes up multi - vibron excitations of c@xmath209 , with a relative probability which can be calculated in principle , in a manner similar to that used by gunnarsson@xcite for c@xmath26 . very recent data@xcite indicate that these shake - ups do exist and are in fact stronger for c@xmath11 than for c@xmath26 . we hope to return to these topics in subsequent work . p. de los rios is grateful to s. panzeri for many useful discussions . we acknowledge exchange of information and discussion with p. brhwiler and f. iachello . this work was partly sponsored by nato , through crg 920828 , and by eec through erbchrxct 940438 . @xmath254 \end{array } \\ \hline \hline 2^{(gs ) } & -\frac 14 & -\frac 9{20 } & -\frac 7{10 } & [ 1,0 ] \\ \begin{array}{c } 0 \\ 3^{\dagger } \\ 4^{\ddagger } \\ 6 \end{array } & \begin{array}{c } -\frac 12 \\ -\frac{11}{28 } \\ -\frac 9{28 } \\ -\frac 14 \end{array } & \left . \begin{array}{c } -\frac 9{20 } \\ -\frac{121}{280 } \\ -\frac{141}{280 } \\ -\frac 7{10 } \end{array } \right\ } & -\frac{19}{20 } & [ 3,0 ] \\ \begin{array}{c } 1 \\ 2^ { * } \\ 3^{\dagger } \\ 4^{\ddagger } \\ 5 \end{array } & \begin{array}{c } -\frac 18 \\ -\frac 14 \\ -\frac 14 \\ -\frac 14 \\ -\frac 14 \end{array } & \left . \begin{array}{c } -\frac 9{20 } \\ -\frac 9{20 } \\ -\frac 9{20 } \\ -\frac 9{20 } \\ -\frac{13}{40 } \end{array } \right\ } & -\frac{23}{40 } & [ 1,2 ] \\ 2^ { * } & + \frac 3{56 } & -\frac 1{280 } & -\frac 3{40 } & [ 1,0]\\ \hline \end{array } $ ]	we study the jahn - teller effect of positive fullerene ions @xmath0c@xmath1 and @xmath2c@xmath3 . the aim is to discover if this case , in analogy with the negative ion , possesses a berry phase or not , and what are the consequences on dynamical jahn - teller quantization . working in the linear and spherical approximation , we find no berry phase in @xmath2c@xmath3 , and presence / absence of berry phase for coupling of one @xmath4 hole to an @xmath5/@xmath4 vibration . we study in particular the special equal - coupling case ( @xmath6 ) , which is reduced to the motion of a particle on a 5-dimensional sphere . in the icosahedral molecule , the final outcome assesses the presence / absence of a berry phase of @xmath7 for the @xmath8 hole coupled to @xmath9/@xmath10 vibrations . some qualitative consequences on ground - state symmetry , low - lying excitations , and electron emission from c@xmath11 are spelled out .
three active galactic nuclei ( agn ) have been discovered to be very high energy ( vhe , e@xmath2300 gev ) @xmath0-ray sources by the whipple observatory @xmath0-ray collaboration : markarian 421 ( mrk 421 ) ( @xcite ) , markarian 501 ( mrk 501)(@xcite ) and 1es 2344 + 514 ( @xcite ) . these are the three closest bl lacertae objects ( bl lacs ) , with redshifts in the range 0.0308 - 0.044 , and are among the brightest at x - ray energies . they are all classified as x - ray selected bl lacs ( xbls ) as their synchrotron spectra extend into the x - ray range . a fourth bl lac , pks2155 - 304 , has been detected in vhe @xmath0-rays by the university of durham group ( @xcite ) . the energetic gamma - ray experiment telescope ( egret ) on board the compton gamma - ray observatory has detected at least 51 agn at energies @xmath3100 mev ( @xcite ; @xcite ) . they are all members of the blazar class of agn , which include flat spectrum radio quasars and bl lacs . of the egret - detected blazars , 14 are bl lacs with 12 radio selected bl lacs ( rbls ) and only two xbls . mrk 421 is the only vhe source in this catalogue and it is among the weakest . however , mrk 501 has recently been detected at the 4@xmath4 level with egret ( @xcite ) . one of the most striking characteristics of the egret - detected blazars is variability ; 42 of the 51 agn exhibit variability ( @xcite ) . variability time - scales as short as 4 hours have been observed ( @xcite ) . vhe observations of mrk 421 and mrk 501 have also revealed extreme variability ( e.g. @xcite ; @xcite ) . the vhe flux has been measured to vary by nearly a factor of 100 in mrk 421 ( @xcite ) and , as we show in the following sections , mrk 501 has been measured with fluxes ranging from 0.1 to 5 times the crab nebula flux with the whipple observatory atmospheric erenkov telescope . the large collection area ( @xmath5 m@xmath6 ) of the whipple observatory telescope permits sensitive studies of variability on time - scales inaccessible to space - based telescopes . indeed , the shortest observed variability of any blazar at any @xmath0-ray energy , a 30 minute duration flare observed from mrk 421 ( @xcite ) , was measured with this telescope . both mrk 421 and mrk 501 have been closely monitored by the whipple collaboration since their discovery , with an @xmath10.5 hr exposure / night being sufficient for detection of flaring activity . prior to 1997 , the @xmath0-ray emission from mrk 421 was generally observed to have a higher mean flux ( @xcite ) and to have been more frequently variable ( @xcite ) than that of mrk 501 . in spring 1997 , the whipple collaboration observed mrk 501 to be in an unprecedented high emission state at vhe energies , as subsequently confirmed by several independent erenkov imaging groups ( @xcite ) . a public compton gamma - ray observatory target of opportunity was initiated in response to a request by the whipple collaboration in 1997 . evidence of correlated variability in data from the whipple observatory @xmath0-ray telescope , the oriented scintillation spectrometer experiment ( osse ) and the all - sky monitor ( asm ) of the _ rossi x - ray timing explorer _ is presented in catanese et al . ( 1997 ) . during this observing campaign the energy output of mrk 501 in vhe @xmath0-rays was comparable to that in the 2 - 100 kev range but the variability amplitude was larger . the correlations seen may imply some relativistic beaming of the emission , given that the spectrum extends to @xmath27 tev ( @xcite ) . there was also some indication that the optical u - band flux was higher on average in the month of peak @xmath0-ray activity . here , based on 4 years of data , we present a study of the variability of the @xmath0-ray flux above @xmath1350 gev from mrk 501 . details of the observations are given in [ section : observations ] and the analysis methodology , including @xmath0-ray selection criteria and the test for variability , is described in [ section : analysis ] . the results of the analysis are presented in [ section : results ] and their implications briefly discussed in [ section : discussion ] . observations were made with the whipple observatory 10 m atmospheric erenkov imaging telescope ( @xcite ) , located on mt . hopkins in southern arizona . the 10 m reflector images the erenkov radiation from cosmic - ray and @xmath0-ray initiated air - showers onto a high resolution camera mounted in the focal plane . subsequent off - line analysis of the images ( described below ) facilitates the selection of candidate @xmath0-ray events . the high resolution camera utilises fast photomultiplier tubes ( pmts ) arranged in a hexagonal array , with inter - tube spacing of @xmath7 . during 1995 and 1996 the camera consisted of 109 pmts with a resulting field of view ( fov ) of @xmath8 . an event was recorded when any two of the inner 91 pmts registered a signal @xmath240 photoelectrons within an effective resolving time of 15 ns . for the 1997 observations a further 42 pmts were added , resulting in a fov of @xmath9 . the trigger condition remained the same as for the 109 pmt camera . in summer 1997 a new camera , containing 331 pixels , was installed . this camera has a fov of @xmath10 . the trigger condition for this enlarged camera required that any two of the 331 pixels produce a signal @xmath240 photo - electrons within an effective resolving time of 8 ns . the telescope was also triggered artificially , after every 24 events for the 109/151 pixel cameras and once every second for the 331 pixel camera , to determine the background sky - brightness level in each pmt . light - cones , which minimise the dead - space between pmts and reduce the albedo effect , were used on the 109 and 151 pixel cameras , but were not yet installed on the 331 pixel camera . in general , two modes of observation are used : on / off and tracking . with the on / off mode the source is tracked continuously for 28 minutes and then , to estimate the background , a region offset in right ascension ( ra ) by 30 minutes ( allowing 2 minutes slew time ) is tracked . this has the disadvantage that an equivalent amount of observation time is spent looking away from the source . alternatively , the tracking mode , where the background is estimated from the on source run itself ( see [ section : trkanalysis ] ) , can be used . in this case an off source run is not required for each on source run , allowing continuous monitoring of an object . however , off runs are still needed to determine the response of the telescope to background events . observations are typically made when the source zenith angle is less than 35@xmath11 and are referred to as small zenith angle ( sza ) observations . observations at large zenith angles ( lza , typically 55@xmath11 to 70@xmath11 ) may also be made . increasing the zenith angle has the effect of increasing the energy threshold , but has the benefit of increasing the collection area . thus , it is an excellent method for increasing photon statistics to facilitate the determination of the energy spectrum at higher energies . for a detailed description of the lza technique see krennrich et al . ( 1997 ) . since its discovery as a @xmath0-ray source in 1995 ( @xcite ) , the vhe @xmath0-ray emission from mrk 501 has been monitored intensively with the whipple observatory 10 m telescope . only sza observations taken under good sky conditions have been considered in the analysis for variability presented here . our selection includes data from 56 , 50 , 55 and 49 nights of observation in the spring summer periods of 1995 , 1996 , 1997 and 1998 respectively . table [ table : observations ] summarises the resulting database . the vast majority of events detected by erenkov telescopes are cosmic rays . candidate @xmath0-ray events are selected on the basis of the shape and orientation of the erenkov images : @xmath0-ray images are typically more compact and elliptical than background hadronic images and tend to have their major axes aligned with the source location in the fov . background cosmic ray events , on the other hand , have random orientations . each image is first subjected to a cleaning procedure ( @xcite ) which suppresses pixels which are dominated by light from fluctuations of the night - sky background . a moment - fitting routine is then used to calculate various image parameters . the shape of each image is characterised by the parameters _ length _ and _ width _ and the orientation by @xmath12 , the angle between the major axis of the image and the line joining the source location in the fov to the centroid of the image . in addition , for the data taken in 1998 , an _ asymmetry _ cut has been included . @xmath0-ray images have a cometary shape with a tail which points away from the source location in the fov ( @xcite ) and thus their intensity profiles have positive asymmetry . the larger fov of the camera used in 1998 allows this parameter to be accurately determined , something which was not possible with the smaller fov cameras . before the application of shape and orientation cuts a _ software trigger _ cut is also applied to eliminate events close to threshold , some of which are induced by noise fluctuations . the software trigger involves cuts on the image _ size _ ( i.e. the total number of photoelectrons recorded ) , the counts in each of the brightest two tubes ( _ max1 , max2 _ ) , and a requirement that at least three tubes above a low noise threshold ( 2.25@xmath4 , where @xmath4 is the rms sky - noise in a pmt , as determined from the artificially triggered events ) be neighbours ( _ nbr3 _ ) . a _ distance _ cut is applied to eliminate images which are too close to the center of the camera and will have poor @xmath12 reconstruction and also those events which have occurred too close to the edge of the fov and may be truncated . due to the continuous evolution of the high resolution camera , and changes such as deterioration of mirror reflectivity through weathering and the presence or absence of light - cones , the optimum data analysis cuts differ for each year . the cuts used for a given telescope configuration are optimised on an independent data set , usually data taken on the crab nebula or , in the case of the 1998 data , mrk 421 . table [ table : cuts ] lists the cuts used for each years analyses . there are slightly different analysis methods for the on / off and tracking observation modes . for the on / off observations the background is estimated from the off source run , which is assumed to be on a sky region which does not include a @xmath0-ray source . this analysis mode has been discussed at length elsewhere ( e.g. @xcite ; @xcite ) . for tracking observations the background is estimated from the on source run itself . all of the @xmath0-ray selection criteria apart from orientation ( @xmath12 ) are applied to the data . the background is then estimated from events which are not oriented towards the source . in this analysis , background events with values of @xmath12 between 20@xmath11 and 65@xmath11 are used . images having values of @xmath12 between 65@xmath11 and 90@xmath11 are discarded because of possible systematic effects due to truncation at the camera s edge . once the number of events with orientations in the 20@xmath11 to 65@xmath11 range is known then the expected number of background events in the _ signal _ domain ( @xmath12 of 0@xmath11 to 15@xmath11 , or to 10@xmath11 for data taken in 1998 ) can be estimated . off source data recorded for this source and others can be combined to calculate a ratio , @xmath13 , of the number of events in the signal region to those in the 20@xmath11 to 65@xmath11 region in the absence of a source . in this case , the significance of a @xmath0-ray excess , @xmath14 , is given by : @xmath15 and the @xmath0-ray rate ( @xmath16 ) is calculated from : @xmath17 where @xmath18 is the number of counts in the @xmath0-ray domain ( @xmath19 or @xmath20 ) , @xmath21 is the number of counts in the 20@xmath11 to 65@xmath11 @xmath12 range and @xmath22 is the duration of the observation . the inclusion of the statistical error on the tracking ratio effectively limits the amount ( duration ) of tracking data which can be usefully analysed for an excess . once the duration of the tracking data exceeds that of the off source data used in the calculation of the tracking ratio the error on the tracking ratio starts to dominate and to limit the significance of a detection . for the purpose of investigating possible differences in results produced by the on / off and tracking analyses , the crab nebula data were analysed with both methods ( @xcite ) . the results demonstrated that the @xmath0-ray rates derived using both methods were consistent and stable ; the tracking analysis did not introduce any apparent variability in the rate . for the analysis presented in this paper , all of the on source data were combined with the tracking data and the resulting database analysed using the tracking analysis . the data presented here were taken with different telescope configurations having different sensitivities . it is therefore necessary to normalize when comparing the different data sets . to a first approximation , this can be achieved by converting the rates to fractions of the @xmath0-ray rate from the crab nebula taken with the same telescope configuration and analysed with the same cuts . the crab nebula is believed to be a steady source of vhe @xmath0-rays , as has been observed by the whipple collaboration over the past decade ( @xcite ) . to convert a given @xmath0-ray rate to an integral flux , the rate as a fraction of the crab nebula flux was multiplied by @xmath23@xmath24s@xmath25 , which represents the integral crab nebula flux above 350 gev ( @xcite ) , the threshold of the analysis presented here . to search for temporal variability in the @xmath0-ray flux we apply a @xmath26 test for a constant rate . the @xmath26 sum is converted into a probability ( @xmath27 ) that the emission is constant about the mean using the incomplete gamma function _ gammq_(a , x ) ( @xcite ) @xmath28 where n-1 is the number of degrees of freedom . the number of trials is taken into account by calculating the probability ( @xmath29 ) of @xmath27 occurring in n trials from : @xmath30 this method is used to test whether the distribution of measured @xmath0-ray rates is consistent with statistical fluctuations about the mean for a range of time - scales . the data have been analysed using the tracking analysis described above . this differs from that applied to the 1995 data by quinn et al . ( 1996 ) , in that a statistical error on the tracking ratio is now included and that a 10% systematic error is no longer added to tracking results as a careful study showed that the results of the on / off and tracking analysis methods are in close agreement ( @xcite ) . in fact , these changes tend to cancel each other . for the 1995 data set as a whole we obtain a @xmath0-ray rate of approximately 10% of that of the crab nebula ( @xmath31min@xmath25 , giving a 9.1@xmath4 excess ) . this rose to approximately 20% of the crab nebula rate for the following season ( analysis of 1996 data reveals an excess of 11.1@xmath4 and a @xmath0-ray rate of @xmath32min@xmath25 ) , indicating that the average emission level had doubled since the previous year . the monthly and nightly average rates , in fractions of the crab nebula rate , over all four years of mrk 501 observation are shown in figure [ fig - m5all ] . the rate in 1995 appears to have been constant with the exception of one night , mjd 49920 , when the it was approximately 4.6@xmath4 above the average . a @xmath26 test gives a chance probability of @xmath33 ( after accounting for trials ) that the daily averages were constant during that month . the @xmath26 probability that the daily averages are constant over the entire 5 months of observation is 0.17 ( after trials ) . the probability that the emission is constant when averaged on monthly time - scales is 0.06 ( after trials ) . conversely , in the 1996 data set there were no obvious flaring episodes , but when averaged on time - scales of a month there is significant variability . the @xmath26 probability that the emission is constant for the monthly averages is @xmath34 , after accounting for trials . when each month is examined for variability with the rates averaged on time - scales of a day no significant variability is found . however , the @xmath26 probabilities are smaller than for a similar analysis of the 1995 data , suggestive of increased day - scale flickering in the @xmath0-ray emission . the results of the @xmath26 test for variability of the monthly and daily averages are shown in tables [ table : monthly_var ] and [ table : daily_var ] respectively . observations of mrk 501 in 1997 began in january using the lza technique ( @xcite ) . the initial observations suggested that the flux was much higher than in previous years . conventional sza observations commenced as soon as possible in february . other ground - based @xmath0-ray experiments were notified and high emission levels were verified . in march a joint iau circular by the cat , hegra and whipple groups ( @xcite ) announced preliminary results . the whipple observations continued through june and included observations made with the lza technique . the results of the lza data are presented elsewhere ( @xcite ) . our sza observations revealed that the vhe @xmath0-ray emission was very strong throughout spring summer 1997 . significant variability was observed in the monthly averages . the emission appeared to increase steadily from february through may and then to level off in june . the daily rates also exhibit dramatic variability . day - to - day changes in the flux by factors @xmath34 were observed and on eight occasions the flux more than doubled between consecutive nights . on four occasions there were equally rapid decays in the rate . the average flux level for the season is 1.4 times that of the crab nebula - an increase by a factor of 14 from the level in 1995 . the peak rate , observed on mjd 50554 , is 3.7 times the crab rate . the average rates for the 1997 data were calculated from only one run ( the first with elevation above 55@xmath11 ) per night . this was done to remove a bias in the observing strategy whereby observations of mrk 501 on a given night continued only if the source was in a very active state . the strong signal - to - background level in the 1997 data allowed a search to be made for variability on time - scales shorter than one day . for this test , data from each night on which there were three or more runs ( approx . 1.5 hours , see [ section : observations ] ) were analysed to test for run - to - run variability . a total of 24 nights satisfied this criterion . the @xmath26 probability that the emission was constant was calculated for each of the nights ; the distribution of these probabilities is shown in figure [ fig - m597-csqdist]a . for a statistically - variable source this distribution should be flat but there is an excess of nights with small probabilities . assuming that the variations are purely statistical , the probability of getting seven nights in this first bin ( width = 0.025 ) out of 24 trials is 1.45@xmath3510@xmath36 . of these seven nights , two show statistically significant variations within themselves . the probability for constant emission on mjd 50577 is @xmath37 while for mjd 50607 the probability is @xmath38 ( after accounting for trials ) . the flux on mjd 50607 has a doubling time of @xmath12 hours . we thus identify these two nights as exhibiting significant hour - scale variability while the five other nights exhibit marginal variability . figure [ fig - hsvar - strong ] shows the @xmath0-ray rates for the two nights with significant variability . a search of the 1997 data for variability on time - scales of less than half an hour has also been performed . for this test each of the 28 minute runs , 144 in total , was divided into three equal length intervals . each triplet was then analysed for variability . no significant variations were found and the distribution of @xmath26 probabilities ( figure [ fig - m597-csqdist]b ) does not indicate any excess of low probabilities . hence , we see no evidence for significant sub - hour scale variability . the average @xmath0-ray rate for the 1998 data set was [email protected] min@xmath25 , approximately 20% of the rate obtained from the crab nebula i.e. on average the emission was much lower than in 1997 . in fact , the average rates for march , april and may are comparable to the level of the initial detection in 1995 . there were however two significant flaring events . the first occurred in early march where an apparent rise and decay were observed . unfortunately this flare is poorly sampled due to bad weather . the flux was observed to be relatively high ( @xmath11.3 times the crab nebula flux ) on mjd 50876 and on the following night a flux of @xmath15.0 times that of the crab nebula was recorded . this is the largest flux detected to date from mrk 501 by the whipple observatory 10 m telescope . for the next observation on mjd 50880 the measured flux was still relatively high ( @xmath11.3 times the level of the crab nebula flux ) . a second flare occurred in june . the flux increased on two consecutive nights , peaking at @xmath11.1 times the crab nebula flux on mjd 50991 , then decayed on a similar time - scale . the average flux level immediately to either side of this flare was below the sensitivity of the telescope . we have demonstrated that rapid variability , a common characteristic of blazars at all observed energies , is also present in the vhe @xmath0-ray emission from mrk 501 . our 4 year data set spans a remarkable change in the flux level from mrk 501 : the average yearly emission level exhibited a fourteen - fold increase between 1995 and 1997 and the average daily flux varied by a factor of @xmath150 ( see figure [ fig - m5all ] ) . in 1997 large amplitude day - scale flares occurred frequently and were usually followed by equally rapid decays . day - scale changes in flux by factors as large as 4.7 were observed . episodes of significant hour - scale variability were detected , with one having a doubling time of 2 hours . in addition , there is evidence of consistent hour - scale variability which is not resolved in individual episodes . a variable flux from mrk 501 was reported by at least four other atmospheric erenkov observatories in 1997 ( see , e.g. @xcite ) . the data presented here suggest that mrk 501 was more variable when the flux level was higher . however , this effect could also be due to the sensitivity of the telescope . at low flux levels it takes longer to accumulate a significant signal , so the search for short - term variability may be limited by poor statistics . to address this issue we performed a test to see if the day - scale variability observed in 1997 would have been detected in 1996 and/or 1995 and if the month - scale variability seen in 1996 and 1997 would have been detected in 1995 . we calculated the percentage deviations about the mean level from a period where significant variability was observed and then , using the mean signal and background level from another period , calculated the signal ( and statistical error ) which would have been observed given the same percentage deviations about that mean . the results are that neither the 1997 month - scale nor day - scale variability would have been detected at a significant level if present in 1995 . however , the 1996 month - scale variability would have been significant if present in 1995 ( chance probability of @xmath40 ) , while the 1997 degree of day - scale variability , if present in 1996 , would have been detectable ( chance probability @xmath41 ) . thus we conclude that there was a change in the flaring characteristics , in addition to the change in the mean flux level between the different observing seasons . the vhe @xmath0-ray emission from mrk 501 exhibits rapid variability similar to that of the emission from mrk 421 . a major difference between the two objects is that the emission from mrk 501 seems to have a base - level , which changes on monthly and yearly time - scales , whearas the vhe @xmath0-ray emission from mrk 421 has been described as consisting of a series of rapid flares with no underlying baseline ( @xcite ) . the variability of the vhe @xmath0-ray emission of mrk 501 in 1995 and 1996 was similar to that at other wavelengths , with small amplitude , slow variations being more common than fast , large amplitude flares . the increase in the vhe @xmath0-ray power and variability in 1997 was accompanied by an increase in the hard x - ray power and an extension of the synchrotron spectrum to at least 100 kev ( @xcite ) . this is consistent with an inverse - compton mechanism for producing vhe @xmath0-rays . however , our observations of mrk 501 to date can not discriminate between the inverse - compton ( electron ) models and those where the dominant particles producing @xmath0-rays in the jet are protons . more densely sampled light curves , covering as broad a wave - band as possible , are needed to provide more insight into the mechanisms responsible for the vhe radiation from bl lac objects . future , more sensitive @xmath0-ray observations of agn with proposed detectors such as glast , hess , magic and veritas will allow the structure of flares on shorter time - scales to be determined . lrrrr feb & & & 3.3 & 0.9 mar & 3.7 & 18.0 & 12.4 & 5.5 apr & 17.7 & 15.5 & 20.7 & 4.6 may & 20.3 & 10.3 & 21.2 & 20.5 jun & 11.5 & 8.8 & 9.7 & 11.6 jul & 8.3 & & & 8.7 total & 61.5 & 52.6 & 67.3 & 51.8 [ table : observations ] cc & 0.073@xmath11 @xmath42 _ width _ 0.15@xmath11 & 0.16@xmath11 @xmath42 _ length _ @xmath42 0.30@xmath11 1995 & 0.51@xmath11 @xmath42 _ distance _ @xmath42 1.0@xmath11 and & @xmath43 1996 & _ size @xmath3 400 d.c . & _ max1 @xmath3 100 d.c . & _ max2 @xmath3 80 d.c . & 0.073@xmath11 @xmath42 _ width _ 0.16@xmath11 & 0.16@xmath11 @xmath42 _ length _ 0.33@xmath11 & 0.51@xmath11 @xmath42 _ distance _ @xmath42 1.17@xmath11 1997 & @xmath43 & _ size @xmath3 0 d.c . & _ max1 @xmath3 95 d.c . & _ max2 @xmath3 45 d.c . & 0.073@xmath11 @xmath42 _ width _ 0.16@xmath11 & 0.16@xmath11 @xmath42 _ length _ 0.44@xmath11 & 0.51@xmath11 @xmath42 _ distance _ @xmath42 1.25@xmath11 1998 & @xmath44 & _ size @xmath3 0 d.c . & _ max1 @xmath3 75 d.c . & _ max2 @xmath3 65 d.c . & _ asymmetry @xmath3 0 [ table : cuts ] _ _ _ _ _ _ _ _ _ _ lcrr 1995 & 5 & [email protected] & 0.05 1996 & 4 & [email protected] & 3.6@xmath3510@xmath45 1997 & 5 & [email protected] & 8.8@xmath3510@xmath46 1998 & 6 & [email protected] & 8.5@xmath3510@xmath47 [ table : monthly_var ] lcrr & & & mar 28 - apr 07 & 09 & [email protected] & 1.00 apr 22 - may 10 & 15 & [email protected] & 1.00 may 21 - jun 05 & 14 & [email protected] & 1.00 jun 18 - jul 01 & 10 & [email protected] & 0.99 jul 22 - jul 30 & 08 & [email protected] & 1.8@xmath3510@xmath48 total & 56 & [email protected] & 0.22 mar 17 - mar 30 & 13 & [email protected] & 0.33 apr 13 - apr 28 & 15 & [email protected] & 1.00 may 09 - may 25 & 15 & [email protected] & 0.18 jun 06 - jun 21 & 10 & [email protected] & 0.46 total & 53 & [email protected] & 3.8@xmath3510@xmath36 feb 10 - feb 16 & 05 & [email protected] & 0.33 mar 05 - mar 16 & 12 & [email protected] & @xmath49 apr 07 - apr 19 & 12 & [email protected] & @xmath49 apr 30 - may 15 & 12 & [email protected] & @xmath49 may 27 - jun 11 & 14 & [email protected] & @xmath49 total & 55 & [email protected] & @xmath49 feb 28 - mar 09 & 06 & [email protected] & @xmath49 mar 23 - mar 31 & 03 & [email protected] & 1.00 apr 19 - apr 29 & 09 & [email protected] & 1.00 may 20 - may 31 & 10 & [email protected] & 0.13 jun 16 - jul 03 & 15 & [email protected] & @xmath49 jul 15 - jul 27 & 06 & [email protected] & 0.94 total & 49 & [email protected] & @xmath49 [ table : daily_var ]	the bl lacertae object markarian 501 was identified as a source of @xmath0-ray emission at the whipple observatory in march 1995 . here we present a flux variability analysis on several times - scales of the 233 hour data set accumulated over 213 nights ( from march 1995 to july 1998 ) with the whipple observatory 10 m atmospheric erenkov imaging telescope . in 1995 , with the exception of a single night , the flux from markarian 501 was constant on daily and monthly time - scales and had an average flux of only 10% that of the crab nebula , making it the weakest vhe source detected to date . in 1996 , the average flux was approximately twice the 1995 flux and showed significant month - to - month variability . no significant day - scale variations were detected . the average @xmath0-ray flux above @xmath1350 gev in the 1997 observing season rose to 1.4 times that of the crab nebula 14 times the 1995 discovery level allowing a search for variability on time - scales shorter than one day . significant hour - scale variability was present in the 1997 data , with the shortest , observed on mjd 50607 , having a doubling time of @xmath12 hours . in 1998 the average emission level decreased considerably from that of 1997 ( to @xmath120% of the crab nebula flux ) but two significant flaring events were observed . thus , the emission from markarian 501 shows large amplitude and rapid flux variability at very high energies as does markarian 421 . it also shows large mean flux level variations on year - to - year time - scales , behaviour which has not been seen from markarian 421 so far .
at present , the decays @xmath0 and @xmath1 are not well studied in experiments @xcite . however , recently , a number of works devoted to the investigation of these processes in the framework of the different phenomenological models were published @xcite . on the other hand , in @xcite it was shown that different modes of the @xmath5 decay can be satisfactorily described in the njl model @xcite . in the present paper , the njl model is used for the description of the decays @xmath0 and @xmath1 . the probabilities of the transitions @xmath8 and @xmath9 are calculated . these transitions are defined by the mass difference between @xmath2 and @xmath3 quarks and can be calculated in the framework of the njl model without attraction of any arbitrary parameters . our results will be compared with the estimates obtained in @xcite . it is shown that the amplitudes with intermediate vector mesons dominate in the @xmath10 decay . for calculation of this decay we should first calculate two non - diagonal transitions @xmath11 and @xmath12 within the njl model . these transitions go through quark loops containing @xmath2 and @xmath3 quarks ( see fig . [ fig1 ] ) . cc etapi ( 150,30 ) & vtoa0 ( 150,30 ) the amplitude of the transition @xmath8 has the form @xmath13 where @xmath14 , @xmath15 , @xmath16 are masses of @xmath17 , @xmath18 and @xmath19 mesons , respectively , given in pdg @xcite ; @xmath20 and @xmath21 are constituent quark masses , @xmath22 mev . using the last experimental data for the decay @xmath23 @xcite we obtain @xmath24 mev . this decay was described in detail in @xcite . the @xmath18 @xmath19 mixing @xmath25 for the @xmath18 meson and @xmath26 for the @xmath19 meson . the mixing angle @xmath27 was defined in @xcite . the constant @xmath28 and integrals @xmath29 , @xmath30 are defined in @xcite @xmath31\ , , \\ i_2(m ) & = & -i\frac{n_c}{(2\pi)^4}\int^{\lambda_4 } \frac{{\mbox{d}}^4k}{(m^2 - k^2)^2 } = \frac{n_c}{(4\pi)^2}\left[\log\left(\frac{\lambda_4 ^ 2}{m^2 } + 1\right ) - \left(1 + \frac{m^2}{\lambda_4 ^ 2}\right)^{-1}\right]\,,\end{aligned}\ ] ] where @xmath32 is a number of quark colors and @xmath33 mev is a 4-dimentional cut - off parameter in the standard njl model @xcite . the obtained estimates coincide with those used in @xcite . one can see the comparison in table [ tbl1 ] . .comparison of @xmath34 [ cols= " < , < , < , < " , ] [ tbl1 ] the transition @xmath35 takes the form @xmath36 where @xmath37 is momentum of the @xmath38 meson . the @xmath38 meson decay width is defined by two diagrams in figs . [ fig2 ] and [ fig3 ] . rho1 ( 250,100 ) rho2 ( 250,100 ) the first diagram describes the amplitude which contains the @xmath11 transitions in the final state @xmath39 where @xmath40 is defined in @xcite . the second diagram describes the amplitude containing the intermediate @xmath41 meson @xmath42 where the vertex @xmath43 was defined in @xcite @xmath44 and @xmath45 mev is the mass of the @xmath46 meson @xcite . thus , for branching fractions we get @xmath47 where @xmath48 . we note that in these calculations we take into account only the ground state of @xmath49 because the decay with the intermediate @xmath50 is suppresed by a large mass of the radial - excited meson . our estimates coincide with one taken in @xcite . these estimates do not contradict known experimental limits @xcite . the description of the decay @xmath51 was obtained in @xcite with satisfactory agreement with current experimental data . tau1 ( 250,100 ) we use the amplitude from @xcite with the @xmath8 transitions ( [ pietatrans ] ) in the final states ( see fig . [ fig4 ] ) dependence for a rough estimate . ] @xmath52 where the breit - wigner relation @xmath53 and @xmath54 parameter were defined in @xcite . for the processes with the intermediate vector meson we get contributions to branching fractions @xmath55 the @xmath56 transition takes the form @xmath57 where @xmath58 is the electroweak constant . tau2 ( 250,100 ) the amplitude with the intermediate scalar meson ( see fig . [ fig5 ] ) takes the form @xmath59 where @xmath60 is the breit wigner formula for the @xmath61 meson with @xmath62 mev , @xmath63 mev , @xmath64 mev taken from pdg @xcite and @xmath65 mev calculated from ( [ a0njl ] ) which coincides with the upper pdg limit @xcite . for the estimation of the contribution of the radial - excited @xmath66 to the @xmath5 decays we should use the extended njl model @xcite . the amplitudes @xmath67 of the @xmath68 decays can be found in @xcite . the transition @xmath69 takes the form @xmath70 where @xmath71 and @xmath72 are the mixing angles , and @xmath73 . thus , we get the @xmath74 parameter @xmath75 where phase factor @xmath76 are taken similarly @xcite . the values @xmath77 and @xmath78 do not contradict with ones given in @xcite . the contributions to the branching fractions from the amplitude ( [ scalamp ] ) are @xmath79 the expression for the total width is @xmath80 note that there is no interference between the vector and scalar intermediate state contributions . thus , for branchings we get @xmath81 let us note that our estimations for scalar contributions are much less than ones in previous works . our calculations are in qualitative agreement with the previous theoretical estimates obtained in @xcite . however , the njl model allows us to describe the transitions @xmath8 and @xmath82 using the same methods . as the result , we can compare the contribution of amplitudes with intermediate scalar and vector mesons from uniform positions . these calculations show that in the decays @xmath0 and @xmath10 the scalar meson plays a insignificant role . however , in the decay @xmath83 the processes with intermediate @xmath49 and @xmath84 make contributions comparable with the contributions of intermediate vector mesons . it is worth noticing that the width of the decay @xmath85 calculated in the njl model is close to the values obtained in @xcite @xmath86 this confirms the relevance of our expression for the vertex @xmath87 used in ( [ scalamp ] ) . for the vertex @xmath88 the expression was used ( [ a0njl ] ) . we get the amplitude ( [ scalamp ] ) by the matching these expressions through propagator of scalar @xmath49 meson . this contradicts the vdm - like ansatz for the intermediate resonance used in @xcite @xmath89 on the other hand , if we use this ansatz for a vector to scalar transition taken in @xcite and calculate @xmath0 with this ansatz then we get by an order of magnitude @xmath90 this estimate for the branching fraction is close to the current experimental limit @xcite and can be tested in the near future at the high - luminosity @xmath91 colliders in novosibirsk and beijing , for example . therefore , the problem of relevancy of vector scalar transition representation can be clarified . we are grateful to b. a. arbuzov , a. b. arbuzov and e. a. kuraev for useful discussions . this work was supported by the rfbr grant 10 - 02 - 01295-a . k. nakamura _ et al . _ [ particle data group collaboration ] , j. phys . g g * 37 * , 075021 ( 2010 ) . t. ferbel , phys . * 21 * , 111 ( 1966 ) . del amo sanchez _ et al . _ [ the babar collaboration ] , phys . rev . d * 83 * , 032002 ( 2011 ) [ arxiv:1011.3917 [ hep - ex ] ] . s. tisserant and t. n. truong , phys . b * 115 * , 264 ( 1982 ) . h. neufeld and h. rupertsberger , z. phys . c * 68 * , 91 ( 1995 ) . s. nussinov and a. soffer , phys . rev . d * 78 * , 033006 ( 2008 ) [ arxiv:0806.3922 [ hep - ph ] ] . s. nussinov and a. soffer , phys . d * 80 * , 033010 ( 2009 ) [ arxiv:0907.3628 [ hep - ph ] ] . n. paver and riazuddin , phys . d * 82 * , 057301 ( 2010 ) [ arxiv:1005.4001 [ hep - ph ] ] . n. paver and riazuddin , phys . rev . 017302 ( 2011 ) [ arxiv:1105.3595 [ hep - ph ] ] . y. .p . ivanov , a. a. osipov and m. k. volkov , z. phys . c * 49 * , 563 ( 1991 ) . m. k. volkov , y. p. ivanov and a. a. osipov , sov . j. nucl . * 52 * , 82 ( 1990 ) [ yad . 52 * , 129 ( 1990 ) ] . m. k. volkov and d. g. kostunin , arxiv:1202.0506 [ hep - ph ] . m. k. volkov , a. b. arbuzov and d. g. kostunin , arxiv:1204.4537 [ hep - ph ] . d. ebert and m. k. volkov , z. phys . c * 16 * , 205 ( 1983 ) . m. k. volkov , annals phys . * 157 * , 282 ( 1984 ) . m. k. volkov , sov . j. part . * 17 * , 186 ( 1986 ) [ fiz . chast . atom . yadra * 17 * , 433 ( 1986 ) ] . d. ebert and h. reinhardt , nucl . b * 271 * , 188 ( 1986 ) . m. k. volkov , phys . part . * 24 * , 35 ( 1993 ) . d. ebert , h. reinhardt and m. k. volkov , prog . phys . * 33 * , 1 ( 1994 ) . m. k. volkov and a. e. radzhabov , phys . * 49 * , 551 ( 2006 ) . m. k. volkov , m. nagy and v. l. yudichev , nuovo cim . a * 112 * , 225 ( 1999 ) [ hep - ph/9804347 ] . m. k. volkov and c. weiss , phys . rev . d * 56 * , 221 ( 1997 ) [ hep - ph/9608347 ] . m. k. volkov , phys . nucl . * 60 * , 1920 ( 1997 ) [ yad . 60n11 * , 1094 ( 1997 ) ] [ hep - ph/9612456 ] .	the widths of the decays @xmath0 and @xmath1 are calculated in the framework of the njl model . it is shown that these decays are defined by the @xmath2 and @xmath3 quark mass difference . it leads to the suppression of these decays in comparison with the main decay modes . in the process @xmath0 the intermediate scalar @xmath4 state is taken into account . for the @xmath5 decays the intermediate states with @xmath4 , @xmath6 and @xmath7 mesons are used . our estimates are compared with the results obtained in other works . + * keywords * : tau decays , chiral symmetry , nambu - jona - lasinio model , radial excited mesons + * pacs numbers * : 13.35.dx decays of taus 12.39.fe chiral lagrangians
in discussions of quantum information technology @xcite , it has been pointed out that the ability to coherently transfer information between `` flying '' and stationary qubits will be essential . atomic samples are good candidates for quantum storage and processing due to their long coherence times and large , controllable interactions , while photons are the fastest and most robust way to transfer information . this implies that methods to transfer information between atoms and photons will be important to the development of this technology . recently the observation of ultra - slow light ( usl ) @xcite , propagating at group velocities more than seven orders of magnitude below its vacuum speed ( @xmath4 ) , and the subsequent stopping and storing of light pulses in atomic media @xcite has demonstrated a tool to possibly accomplish this @xcite . the technique relies on the concept of _ electromagnetically induced transparency _ ( eit)@xcite in three level @xmath0-configuration atoms ( see fig . [ fig : diagram](a ) ) . a coupling light field @xmath5 is used to control the propagation of a pulse of probe light @xmath6 . the probe propagates at a slow group velocity and , as it is doing so , coherently imprints it s amplitude and phase on the coherence between two stable internal states of the atoms , labelled @xmath7 and @xmath8 ( which are generally particular hyperfine and zeeman sublevels ) . switching the coupling field off stops the probe pulse and ramps it s intensity to zero , freezing the probe s coherent ( that is , intensity _ and _ phase ) information into the atomic media , where it can be stored for a controllable time . switching the coupling field back on at a later time writes the information back onto a revived probe pulse , which then propagates out of the atom cloud and can be detected , for example , with a photo - multiplier tube ( pmt ) ( see fig . [ fig : diagram](b ) ) . in the original experiment @xcite , the revived output pulses were indistinguishable in width and amplitude from non - stored usl pulses , indicating the switching process preserved the information in the atomic medium with a high fidelity . * schematic of ultra - slow light ( usl ) and stopped experiments . * * ( a ) * the @xmath0 energy level structure which we consider has two stable states , @xmath9 and @xmath8 , and an excited state @xmath10 which decays at a rate @xmath11 ( which is @xmath12 mhz for na and @xmath13 mhz for rb-87 ) . atoms which spontaneously decay from @xmath14 are assumed to exit the levels under consideration . * ( b ) * we consider the two light fields to be _ co - propagating _ and input in the @xmath15 direction ( the long axis of the bose - einstein condensate ( bec ) ) at @xmath16 . the output intensity at @xmath17 can be detected experimentally with a photo - multiplier tube ( pmt ) . the probe field is pulsed with half - width @xmath18 . the coupling beam is c.w . but can be switched off quickly to stop the probe pulse while it is in the bec . switching it back on later regenerates the probe light pulse . ] in addition to the ability to store coherent information , compressing ultra - slow probe pulses to lengths shorter than the atomic cloud puts the atoms in spatially dependent superpositions of @xmath7 and @xmath8 , offering novel possibilities to study two component bec dynamics @xcite . in @xcite slow light pulses propagating in bose - einstein condensates ( becs ) @xcite of sodium were used to create density defects with length scales @xmath19 , near the bec healing length , leading to quantum shock waves and the nucleation of vortices . motivated by both observations of coherent optical information storage in atom clouds and the interesting dynamics which slow and stopped light pulses can induce in becs , the present paper theoretically examines new possibilities offered by stopping light in becs . we first present a novel treatment of the switching process and establish that the switch - off and switch - on of the coupling field can occur without dissipating the coherent information , provided the length scale of variations in the atomic wavefunctions are sufficiently large . we find that the switching of the coupling field can be done _ arbitrarily _ fast in the ultra - slow group velocity limit @xmath20 . this is consistent with previous work @xcite , but goes further as it applies to pulses in both the weak and strong probe regimes @xcite . we also present the first explicit calculation of how the behavior of the system differs when one switches the coupling field faster than the natural lifetime of the excited state @xmath21 ( see fig . [ fig : diagram](a ) ) and find the information is successfully transferred even in this regime . furthermore we will see the analysis presented here , being phrased in terms of the spatial characteristics of the atomic wavefunctions , is well suited to addressing the issue of storage for times long compared to the time scale for atomic dynamics . investigation of this very issue of longer storage times constitute the central results of the paper . in this regime , there are important differences between optical information storage in becs versus atom clouds above the condensation temperature ( thermal clouds ) , which have been used in experiments to date . during the storage time , the light fields are off and the atomic wavefunctions evolve due to kinetic energy , the external trapping potential , and atom - atom interactions . ( we label all these _ external _ dynamics to distinguish them from couplings between the internal levels provided by the light fields . ) to successfully regenerate the probe pulse by then switching on the coupling field , there must be a non - zero value of the _ ensemble average _ ( over all atoms ) of the coherence between @xmath9 and @xmath8 . when revivals were attempted after longer storage times in thermal clouds ( several milliseconds in @xcite ) the external dynamics had washed out the phase coherence between @xmath7 and @xmath8 ( due to the atoms occupying a distribution of single particle energy levels ) and no output pulses were observed . in zero temperature becs the situation is completely different . there each atom evolves in an identical manner , described by a non - linear schrodinger equation ( gross - pitaevskii equation @xcite ) , preserving the ensemble average of the coherence during the external dynamics . that is , even as the amplitude and phase of the wavefunctions representing the bec evolve , the relative phase of the two components continues to be well defined at all points in space . thus , if we switch the coupling field back on , we expect that the evolved bec wavefunctions will be written onto a revived probe field . in this way , the probe pulses can be _ processed _ by the bec dynamics . we show several examples , relevant to current rb-87 becs , of the interesting two component dynamics which can occur during the storage time . this atom possesses internal states with very small inelastic loss rates @xcite , and thus , long lifetimes @xcite . we find that , depending upon the relative scattering lengths of the internal states involved and the relative intensity of the probe and coupling fields , one could observe the formation of interference fringes , gentle breathing motion , or the formation and motion of two - component ( vector ) solitons @xcite . in particular we find that using the levels @xmath22 and @xmath23 in rb-87 could allow for long , robust storage of information and controllable processing . in each case , we observe the amplitude and phase due to the dynamical evolution is written onto revived probe pulses . these pulses then propagate out of the bec as slow light pulses , at which point they can be detected , leaving behind a bec purely in its original state @xmath7 . for practical applications of this technique to storage and processing , one must understand in detail how , and with what fidelity , the information contained in the atomic coherence is transferred and output on the light fields . thus , the last part of the paper is devoted to finding the exact relationship between the bec wavefunctions before the switch - on and the observed output probe pulses . we find equations linking the two in the ideal limit ( without absorption or distortion ) . then , using our earlier treatment of the switching process , we identify several sources of imperfections and calculate the fidelity of the information transfer for gaussian shaped wavefunctions of various amplitudes and lengths . in the weak probe limit , we find a simple relationship between the wavefunction in @xmath8 and the output probe field . we find that optimizing the fidelity involves balancing considerations related to , on one hand , absorption of small length scale features in the bec wavefunctions and , on the other hand , imperfect writing of wavefunctions which are too near the condensate edge . for stronger probes , we find a more complicated relationship , though we see one can still reconstruct the amplitude and phase of the wavefunction in @xmath24 using only the output probe . in this regime , we find that the fidelity of the writing and output is nearly independent of the probe strength when the oscillator strengths for the two transitions involved ( @xmath25 and @xmath26 ) are equal . however , unequal oscillator strengths lead to additional distortions and phase shifts , and therefore lower fidelity of information transfer , for stronger probes . we first , in section [ sec : formalism ] , introduce a formalism combining maxwell - schrodinger and gross - pitaevskii equations @xcite , which self - consistently describe both the atomic ( internal and external ) dynamics as well as light field propagation . we have written a code which implements this formalism numerically . section [ sec : switching ] presents our novel analysis of the switching process , whereby the coupling field is rapidly turned off or on , and extends previous treatments of the fast switching regime , as described above . section [ sec : processing ] shows examples of the very rich variety of two - component dynamics which occur when one stops light pulses in becs and waits for much times longer than the characteristic time scale for atomic dynamics . section [ sec : systematic ] contains our quantitative analysis of the fidelity with which the bec wavefunctions are transferred onto the probe field . we conclude in section [ sec : conclusion ] and anticipate how the method studied here could eventually be applied to transfer of information between distant becs , and the generation of light with squeezed statistics . we first introduce our formalism to describe the system and review usl propagation within this formalism . assume a bose - condensed sample of alkali atoms , with each atom in the bec containing three internal ( electronic ) states in a @xmath0-configuration ( fig . [ fig : diagram](a ) ) . the states @xmath7 and @xmath27 are stable and the excited level @xmath14 , radiatively decays at @xmath11 . in the alkalis which we consider , these internal levels correspond to particular hyperfine and zeeman sublevels for the valence electron . when the atoms are prepared in a particular zeeman sub - level and the proper light polarizations and frequencies are used @xcite , this three level analysis is a good description in practice . all atoms are initially condensed in @xmath9 and the entire bec is illuminated with a _ coupling field _ , resonant with @xmath28 transition and propagating in the @xmath15 direction ( see fig . [ fig : diagram](b ) ) . a pulse of _ probe field _ , with temporal half - width @xmath18 , resonant with the @xmath29 transition , and also propagating in the @xmath15 direction , is then injected into the medium . the presence of the coupling field completely alters the optical properties of the atoms as seen by the probe . what would otherwise be an opaque medium ( typical optical densities of becs are @xmath30 ) is rendered transparent via _ electromagnetically induced transparency _ ( eit ) @xcite , and the light pulse propagates at ultra - slow group velocities ( @xmath1 m / s)@xcite . as this occurs , all but a tiny fraction of the probe energy is temporarily put into the coupling field , leading to a compression of the probe pulse to a length smaller than the atomic medium itself @xcite . to describe the system theoretically , we represent the probe and coupling electric fields with their rabi - frequencies @xmath31 , where @xmath32 are slowly varying envelopes of the electric fields ( both of which can be time and space - dependent ) and @xmath33 are the electric dipole moments of the transitions . the bec is described with a two - component spinor wavefunction @xmath34 representing the mean field of the atomic field operator for states @xmath7 and @xmath8 . we ignore quantum fluctuations of these quantities , which is valid when the temperature is substantially below the bec transition temperature @xcite . the excited level @xmath14 can be adiabatically eliminated @xcite when the variations of the light fields envelopes are slow compared to the excited state lifetime @xmath21 ( which is 16 ns in sodium ) . the procedure is outlined in the appendix . the functions @xmath35 evolve via two coupled gross - pitaevskii ( gp ) equations @xcite . for the present paper , we will only consider dynamics in the @xmath36 dimension , giving the bec some cross - sectional area @xmath37 in the transverse dimensions over which all dynamical quantities are assumed to be homogeneous . this model is sufficient to demonstrate the essential effects here . we have considered effects due to the transverse dimensions , but a full exploration of these issue is beyond our present scope . the gp equations are @xcite : @xmath38 \psi_1 \nonumber \\ & & - i \hbar \left ( \frac{\|\omega_p\|^2}{2 \gamma } \psi_1 + \frac{\omega_p^\ast \omega_c}{2 \gamma } \psi_2 \right ) , \nonumber \\ i \hbar \frac{\partial \psi_2}{\partial t } & = & \bigg[-\frac{\hbar^2}{2m}\frac{\partial^2}{\partial z^2 } + v_2(z ) + u_{22 } & & - i \hbar \left ( \frac{\|\omega_c\|^2}{2 \gamma } \psi_2 + \frac{\omega_p \omega_c^\ast}{2 \gamma } \psi_1 \right),\end{aligned}\ ] ] where @xmath39 is the mass of the atoms and we will consider a harmonic external trapping potential @xmath40 . the potential @xmath41 can in general differ from @xmath42 . for example , in a magnetic trap the states @xmath7 and @xmath8 can have different magnetic dipole moments . however , in the examples we consider the potentials are equal to a good approximation , and so @xmath43 is assumed in the calculations . atom - atom interactions are characterized by the @xmath44 , where @xmath45 is the total number of condensate atoms and @xmath46 are the s - wave scattering lengths for binary collisions between atoms in internal states @xmath47 and @xmath48 . the last pair of terms in each equation represent the coupling via the light fields and give rise to both coherent exchange between @xmath49 as well as absorption into @xmath14 . in our model , atoms which populate @xmath14 and then spontaneously emit are assumed to be lost from the condensate ( fig . [ fig : diagram](a ) ) , which is why the light coupling terms are non - hermitian . the light fields propagation can be described by maxwell s equations . assuming slowly varying envelopes in time and space ( compared to optical frequencies and wavelengths , respectively ) and with the polarization densities written in terms of the bec wavefunctions these equations are @xcite : @xmath50 where @xmath51 are the dimensionless oscillator strengths of the transitions and @xmath52 is the resonant cross section ( @xmath53 and @xmath54 for the @xmath55 lines of sodium and rb-87 , respectively ) . these equations ignore quantum fluctuations of the light fields , just as the gp equations ( [ eq : formalism12 ] ) ignore quantum fluctuations of the atomic fields . note that in the absence of any coherence @xmath56 these equations predict absorption of each field with the usual two - level atom resonant absorption coefficient ( the atom density times the single atom cross - section ) . for much of our analysis , we will solve eqs . ( [ eq : formalism12])-([eq : formalismpc ] ) self consistently with a numerical code , described in @xcite . in @xcite , this was successfully applied to predict and analyze the experimental observation of the nucleation of solitons and vortices via the _ light roadblock_. when the light fields are off , the two states do not exchange population amplitude and eqs . ( [ eq : formalism12 ] ) reduce to coupled gross - pitaevskii equations for a two - component condensate . by contrast , when they are on and the probe pulse length @xmath18 ( and inverse rabi frequencies @xmath57 ) are much faster than the time scale for atomic dynamics ( typically @xmath58 ms ) , the internal couplings induced by the light field couplings will dominate the external dynamics . for our our initial conditions , we consider @xmath59 atoms are initially in @xmath7 in the condensed ground state . we determine the wavefunction of this state @xmath60 numerically by propagating ( [ eq : formalism12 ] ) in imaginary time @xcite , though the thomas - fermi approximation @xcite provides a good analytic approximation to @xmath60 . in the cases presented in this paper , we choose the transverse area @xmath37 so that the central density and chemical potential @xmath61 , as in previous experiments @xcite . in fig . [ fig : switching ] we consider an example with @xmath62 sodium atoms ( with @xmath63 nm @xcite ) , @xmath64 and @xmath65 . the ground state density profile @xmath66 is indicated with the dotted curve in fig . [ fig : switching](b ) . in this case the chemical potential @xmath67khz . * coherent storage of a light pulse via fast switching . * results of numerical integration of eqs . ( [ eq : formalism12])-([eq : formalismpc ] ) showing both a slow and stopped light experiment . * ( a ) * the dashed curve shows the normalized _ input _ probe pulse intensity @xmath68 ( a @xmath69s @xmath70 half - width , @xmath71mhz pulse ) , while the dotted curve shows @xmath72 for a constant input coupling field @xmath73mhz . the thick solid curve shows the delayed output probe pulse ( at @xmath17 , see fig . [ fig : diagram](b ) ) . the time that the tail of the input has vanished and the rise of the output has not begun ( marked by the arrow ) corresponds to the time the pulse is completely compressed inside the bec . in the output coupling intensity ( thin solid curve ) we see adiabatons . the values @xmath74 and @xmath75 and @xmath76 have been used . * ( b ) * the atomic density in the two states , @xmath77 ( solid curve ) and @xmath78 ( dashed curve ) , at @xmath79s ( indicated by the arrow in ( a ) ) . the dotted curve shows the original density @xmath80 before the probe is input . the arrow here and in ( c)-(d ) marks the position @xmath81 discussed later in the text . * ( c ) * spatial profiles of the probe @xmath82 ( solid curves ) and coupling @xmath83 ( dashed curves ) field intensities at various times while the coupling field input is switched off . the switch - off is an error function profile with a width @xmath84s , centered at @xmath85s . successively thinner curves refer to @xmath86 , and @xmath87s . the dotted curve indicates the original condensate density ( arb . units ) . * ( d ) * spatial profiles of the normalized dark @xmath88 ( dashed curves , scale on the left ) and absorbed @xmath89 ( solid curves , scale on right ) field intensities , at the same times as in ( c ) . * ( e ) * the input and output probe and coupling intensities , with the same conventions as ( a ) , in a stopped light simulation . in this case , the coupling field is switched off at @xmath90 and then back on at @xmath91 . ] with our initial ground state @xmath60 determined , consider that we initially ( @xmath92 ) input a constant coupling field with a rabi frequency @xmath93 and then inject a gaussian shaped probe pulse at @xmath16 ( see fig . [ fig : diagram](b ) ) with a temporal half - width @xmath18 and a peak rabi frequency @xmath94 . we define our times such that @xmath95 corresponds to the time the peak of the pulse is input . the dotted and dashed curves in fig . [ fig : switching](a ) show , respectively , a constant coupling input @xmath96mhz and a weaker input probe pulse with peak amplitude @xmath97mhz and width @xmath98s . solving eqs . ( [ eq : formalism12])-([eq : formalismpc ] ) reveals that , when the conditions necessary for eit hold , the pulse will compress upon entering the bec and propagate with a slow group velocity . as it does this , it transfers the atoms into superpositions of @xmath7 and @xmath8 such that @xcite : @xmath99 which is a generalization to becs of the single atom _ dark state thus , the probe field @xmath6 imprints its ( time and space dependent ) phase and intensity pattern in the bec wavefunctions as it propagates . when ( [ eq : darkstate ] ) is exactly satisfied , it is easily seen that the two light field coupling terms in each of eqs . ( [ eq : formalism12 ] ) cancel , meaning that @xmath7 and @xmath8 are completely decoupled from the excited state @xmath100 . however , time dependence of @xmath6 causes small deviations from ( [ eq : darkstate ] ) to occur , giving rise to some light - atom interaction which is , in fact , the origin of the slow light propagation . if one assumes the weak probe limit ( @xmath101 ) and disregards terms of order @xmath102 the group velocity of the probe pulse is @xcite : @xmath103 and so is proportional the coupling intensity and inversely proportional to the atomic density @xmath104 , which is a reduction from it s free space value by a factor @xmath105 , while its peak amplitude does not change . thus only a tiny fraction of the input probe pulse energy is in the probe field while it is propagating in the bec . most of the remaining energy coherently flows into the coupling field and exits the bec , as can be seen in the small hump in the output coupling intensity ( thin solid curve in fig . [ fig : switching](a ) ) during the probe input . this has been dubbed an adiabaton @xcite . the magnitude of the adiabaton is determined by the intensity of the probe pulse . in the weak probe limit @xmath106 never significantly deviates from the ground - state wavefunction @xmath107 and the coupling field is nearly unaffected by the propagation @xmath108 ( both of these results hold to @xmath109 ) . in this case ( [ eq : darkstate ] ) shows @xmath110 follows the probe field @xmath6 as the pulse propagates . the arrow in fig . [ fig : switching](a ) indicates a time where the probe has been completely input and has not yet begun to output . during this time , the probe is fully compressed in the bec and fig . [ fig : switching](b ) shows the atomic densities in @xmath9 and @xmath8 at this time . the spatial region with a non - zero density in @xmath8 corresponds to the region occupied by the probe pulse ( in accordance with ( [ eq : darkstate ] ) ) . equation ( [ eq : darkstate ] ) applies to the phases as well as the amplitudes , however the phases in this example are homogenous and not plotted . once the pulse has propagated through the bec , it begins to exit the @xmath15 side . the energy coherently flows back from the coupling to probe field , and we see the output probe pulse ( thick solid curve ) . correspondingly , we see a dip in the coupling output at this time . in the experiments the delay between the input and output probe pulses seen in fig . [ fig : switching](a ) is measured with a pmt ( fig . [ fig : diagram](b ) ) . this delay and the length of the atomic cloud is used in to calculate the group velocity . the group velocity at the center of the bec in the case plotted is @xmath111 m/s . note that the output pulse plotted in fig . [ fig : switching](a ) at @xmath17 is slightly attenuated . this reduction in transmission is due to the eit bandwidth . the degree to which the adiabatic requirement @xmath112 is not satisfied , will determine the deviation the wavefunctions from the dark state ( [ eq : darkstate ] ) , which leads to absorption into @xmath14 and subsequent spontaneous emission . quantitatively , this reduces the probe transmission ( the time integrated output energy relative to the input energy ) to @xcite : @xmath113 where @xmath114 is the optical density . in the fig . [ fig : switching ] example , @xmath115 . the peak intensity of the pulse is reduced by a factor @xmath116 while the temporal width is increased by @xmath117 ( this spreading can be seen in fig . [ fig : switching](a ) ) . the appearance of the large optical density @xmath118 in ( [ eq : tfreqwidth ] ) represents the cumulative effect of the pulse seeing a large number of atoms as it passes through the bec . to prevent severe attenuation and spreading , we see we must use @xmath119 , which is @xmath120s in our example . we now turn our attention to the question of stopping , storing and reviving probe pulses . we will show here that once the probe is contained in the bec , the coupling field can be switched off and on _ faster _ than the eit adiabatic time scale without causing absorptions or dissipation of the information @xcite . while we find no requirement on the time scale for the switching , we _ will _ obtain criteria on the length scales of @xmath35 which must be maintained to avoid absorption events . while previous work @xcite had addressed the fast switching case in the weak - probe limit , here we obtain results that are valid even when @xmath121 @xcite . as we will see , for shorter storage times where the external atomic dynamics play no role , these length scale requirements can be related to the adiabatic requirements on the input probe pulse width @xmath18 . the advantage of the current analysis is that it is easily applied to analyzing probe revivals after the atomic dynamics have completely altered the wavefunctions @xmath35 . in such a case it is these altered wavefunctions , rather than the input pulse width @xmath18 , which is relevant . consider that we switch the coupling field input off at some time @xmath122 with a fast time scale @xmath123 . [ fig : switching](c ) plots the probe and coupling intensities as a function of @xmath36 at various times during a switch - off with @xmath124 . we see the probe intensity smoothly ramps down with the coupling field such that their ratio remains everywhere constant in time . remarkably , for reasons we discuss below , the wavefunctions @xmath35 are _ completely unaffected _ by this switching process . motivated by this , we will , in the following , assume that @xmath35 do not vary in time during this fast switching period , and later check this assumption . to understand this behavior , it is useful to go into a dark / absorbed basis for the light fields , similar to that used in @xcite , by defining @xcite : @xmath125 where @xmath126 . from ( [ eq : dalight ] ) one sees that when the condensate is in the dark state ( [ eq : darkstate ] ) , @xmath127 . using the notation @xmath128 and transforming the propagation equations ( [ eq : formalismpc ] ) according to ( [ eq : dalight ] ) , one gets : @xmath129 where @xmath130 , \nonumber \\ \alpha_l & \equiv & \frac{1}{\psi_0 ^ 2}\bigg(\|\psi_1\|^2\frac{d \phi_1}{dz}+\|\psi_2\|^2\frac{d \phi_2}{dz}\bigg),\end{aligned}\ ] ] and we have ignored the vacuum propagation terms @xmath131 in eq . ( [ eq : formalismpc ] ) . they are unimportant so long as the fastest time scale in the problem is slow compared to time it takes a photon travelling at @xmath132 to cross the condensate ( about @xmath133ps ) . note that @xmath134 represents the usual absorption coefficient weighted according to the atomic density in each of @xmath7 and @xmath8 . the terms @xmath135 arise from spatial variations in the wavefunctions , which make the transformation ( [ eq : dalight ] ) space - dependent . the term @xmath136 represents an additional effect present when the light - atom coupling coefficient differs on the two transitions ( @xmath137 and is discussed in detail in section [ subsec : f23 ] . consider for the moment a case with @xmath138 ( implying @xmath139 ) and assume a region in which @xmath35 are homogenous ( so @xmath140 ) . equation ( [ eq : aprop0 ] ) shows that the _ absorbed _ field @xmath141 attenuates with a length scale @xmath142 , the same as that for resonant light in a two - level atomic medium , and less than @xmath143 m at the cloud center for the parameters here . one would then get @xmath144 after propagating several of these lengths . conversely , ( [ eq : dprop0 ] ) shows the _ dark _ light field @xmath145 experiences no interaction with the bec and propagates without attenuation or delay . however , spatial dependence in @xmath35 gives rise to @xmath146 in ( [ eq : dprop0])-([eq : aprop0 ] ) , introducing some coupling between @xmath145 and @xmath141 , with the degree of coupling governed by the spatial derivatives @xmath147 . a simple and relevant example to consider is the case of a weak ultra - slow probe pulse input and contained in a bec , as discussed above ( see fig . [ fig : switching](b ) ) . the wavefunction @xmath110 has a homogenous phase and an amplitude which follows the pulse shape according to ( [ eq : darkstate ] ) , meaning that @xmath148 scales as the inverse of the pulse s spatial length @xmath149 . it is important to note that this coupling is determined by spatial variations in the _ relative _ amplitude @xmath150 and does not get any contribution from variations in the _ total _ atomic density @xmath151 . to see this we note that if we can write the wavefunctions as @xmath152 , where @xmath153 are constants independent of @xmath36 , then @xmath148 and @xmath154 as defined in ( [ eq : alphadef ] ) vanish . when variations in the relative amplitude are present , examination of ( [ eq : aprop0 ] ) reveals that when the damping is much stronger than the coupling ( @xmath155 ) , one can ignore the spatial derivative term in analogy to an adiabatic elimination procedure . when we do this , ( [ eq : aprop0 ] ) can be approximated by : @xmath156 strictly speaking , one can only apply this procedure in the region where @xmath141 has propagated more than one absorption length ( that is for @xmath157 where @xmath81 is defined by @xmath158 ) . however , @xmath81 is in practice only a short distance into the bec ( marked with arrows in figs . [ fig : switching](b)-(d ) ) . when the probe has already been input , as in fig.2(b ) , @xmath110 and therefore @xmath141 ( see ( [ eq : dalight ] ) ) are already trivially zero for @xmath159 so ( [ eq : omegaa ] ) holds everywhere . we next plug eq . ( [ eq : omegaa ] ) into eq . ( [ eq : dprop0 ] ) , giving : @xmath160 \omega_d.\ ] ] these last two equations rely only on assumptions about the spatial derivatives of @xmath35 and _ not _ on the time scale of the switch - off @xmath161 ( though we have assumed @xmath162 in adiabatically eliminating @xmath14 from our original eqs . ( [ eq : formalism12])-([eq : formalismpc ] ) ) . these results allow us to conclude two important things which hold whenever @xmath163 and the probe has been completely input . first , the coefficients governing the propagation of @xmath145 in ( [ eq : dprop ] ) are extremely small . the length scales @xmath164 are already generally comparable to the total bec size and the length scales for changes in @xmath145 , given by ( [ eq : dprop ] ) , scale as these terms multiplied by the large ratio @xmath165 . ( the @xmath154 term in ( [ eq : dprop ] ) can lead to additional phase shifts , which we discuss in section [ subsec : f23 ] ) . therefore the dark field @xmath145 propagates with very little attenuation . as we have noted ( and later justify and discuss in more detail ) the wavefunctions ( and therefore the propagation constant in brackets in ( [ eq : dprop ] ) ) are virtually unchanged during the switch - off . under this condition , changes initiated in @xmath145 at the entering edge @xmath16 quickly propagate across the entire bec . to apply this observation to the switch - off , we note when the pulse is contained so @xmath166 at the entering edge , ( [ eq : dalight ] ) shows @xmath167 there . switching off the coupling field at @xmath16 then amounts to switching off @xmath145 at @xmath16 and ( [ eq : dprop ] ) shows that this switch - off propagates through the entire bec with little attenuation or delay . second , from ( [ eq : omegaa ] ) we see that as @xmath145 is reduced to zero @xmath141 is reduced such that the ratio @xmath168 remains constant in time . a numerical simulation corroborating this behavior is plotted in fig . [ fig : switching](d ) . the dark field intensity @xmath169 is seen to switch - off everywhere as the coupling field input is reduced to zero over a @xmath170s timescale , confirming that the changes in the input propagate across the bec quickly and with little attenutation . the ratio @xmath171 is everywhere much smaller than unity and constant in time . the only exception to this is in a small region @xmath159 at the cloud entrance , where @xmath141 has not yet been fully damped . the plot of @xmath172 during the switch - off indeed demonstrates how ( [ eq : omegaa ] ) is a generally good approximation , but breaks down in this region . in the case plotted ( and any case where the probe is fully contained ) the wavefunction @xmath110 is so negligible in this region that @xmath141 is rather small and unimportant ( note the scale on the right hand side of the plot ) . translating this back into the @xmath173 basis , we note from ( [ eq : dalight ] ) that keeping @xmath174 means the probe must ( at all @xmath36 ) constantly adjust to the coupling field via : @xmath175 thus we see that @xmath6 smoothly ramps down with @xmath5 even if @xmath5 is ramped down quickly , as seen in fig . [ fig : switching](c ) . using our results for the light fields , we can now see why the wavefunctions @xmath35 do not change during the switching . physically , the probe is in fact adjusting to maintain the dark state ( [ eq : omegapadjust ] ) , and in doing so induces some transitions between @xmath7 and @xmath8 . however , only a fraction @xmath176 of the input energy is in the probe while it is contained in the medium , and so the probe is completely depleted before any significant change occurs in @xmath35 @xcite . in fact , the energy content of the probe field right before the switch - off is less than 1/100th of a free - space photon in the case here . thus , @xmath6 is completed depleted after only a fraction of one @xmath177 transition . note that ( [ eq : omegapadjust ] ) is equivalent to ( [ eq : darkstate ] ) . however , writing it in this way emphasizes that , during the switching , the probe is being driven by a reservoir consisting of the coupling field and atoms and adjusts to establish the dark state . this is contrast to the situation during the probe input , when many photons from both fields are being input at a specific amplitude ratio , forcing the _ atomic fields _ to adjust to establish the appropriate dark state . plugging in our results for the light fields ( [ eq : omegaa])-([eq : dprop ] ) into our eqs . ( [ eq : formalism12 ] ) we can calculate the changes that occur in @xmath178 and @xmath110 during the switch - off . doing this , we find relative changes in @xmath35 are both smaller than @xmath179 , which can be made arbitrarily small for fast switching @xmath123 . the little change which does occur is due not to any process associated with the switching itself but is due to the small amount of propagation during the switch - off . it is therefore safe to assume ( as we saw numerically ) that @xmath35 are constant in time during the switch - off . one can also show with this analysis that the ratio of _ coherent _ exchange events ( from @xmath7 to @xmath8 or vice - versa ) to _ absorptive _ events ( transitions to @xmath180 followed by spontaneous emission ) is @xmath181 implying that the switch - off occurs primarily via coherent exchanges . in the original stopped light experiment @xcite , the coupling field was then switched back on to @xmath93 , after some controllable storage time @xmath182 , at a time @xmath183 . in that experiment , revivals were observed for storage times @xmath182 too short for significant external atomic dynamics to occur , so the state of the atoms was virtually identical at @xmath184 and @xmath122 . then the analysis of the switch - on is identical to the switch - off as it is just the same coherent process in reverse . the probe is then restored to the same intensity and phase profile as before the switch - off . an example of such a case is plotted in fig . [ fig : switching](e ) . the output revived pulse then looks exactly like the normal usl pulse ( compare the output in figs . [ fig : switching](a),(e ) ) , as was the case in the experiment . we have established the requirement @xmath185 is necessary and sufficient for coherent switching to occur . under what conditions is this satisfied ? when the switch - off occurs while the probe is compressed , it is _ always _ satisfied because of bandwidth considerations mentioned above ( see eq . ( [ eq : tfreqwidth ] ) ) . specifically , the input pulse must satisfy @xmath186 . however , this leads to a pulse width in the medium of @xmath187 , implying that @xmath185 is satisfied ( as @xmath188 ) . therefore , any pulse which can successfully propagate to the cloud center , can be abruptly stopped and coherently depleted by a rapid switch - off of the coupling field . of course , if the switch - on is then done before significant atomic dynamics , the same reasoning applies then . in this case , our requirements on the spatial derivatives are already encapsulated by the adiabatic requirements on the probe pulse . it is when the external atomic dynamics during the storage significantly change @xmath35 , that the analysis of the switch - on becomes more complicated . this is the central purpose of the sections [ sec : processing]-[sec : systematic ] below . as hinted above , when the pulse is not yet completely input , the switching can cause absorptions and dissipation of the information . in this case @xmath110 is significant in the region @xmath159 and so @xmath141 is significant for several absorption lengths into the bec . physically , the coupling field sees atoms in a superposition of @xmath7 and @xmath8 immediately upon entering the condensate rather than only @xmath7 atoms . numerical simulations of this situation confirm this . during both the switch - on and switch - off , a significant number of atoms are lost from both condensate components , primarily concentrated in the @xmath159 region , and the revived probe pulse is significantly attenuated relative to the pulse before the storage . incidentally , this explains the apparent asymmetry between the probe and coupling in a stopped light experiment . whenever _ both _ fields are being input , the temporal variations in both fields must be slow compared with the eit adiabatic time scale , as the atomic fields must adjust their amplitudes to prevent absorptions . however , when one of the fields is no longer being input ( like a contained probe pulse ) , the other field s input can be quickly varied in time . when the switching is done quickly compared to the natural lifetime of @xmath14 ( @xmath189 ) , the assumptions needed for adiabatic elimination of @xmath190 are no longer valid . the analysis of the switching then becomes more complicated but , remarkably , we find the quality of the storage is not reduced in this regime . to see this , we performed numerical simulations of switching in this regime by solving eqs . ( [ eq : fullgp1])-([eq : maxwell ] ) ( which are the analogues of ( [ eq : formalism12])-([eq : formalismpc ] ) without the adiabatic elimination of @xmath190 ) for the same parameters as fig . [ fig : switching ] but varying @xmath161 from @xmath191 to @xmath192 @xcite . an example with @xmath193 is shown in fig . [ fig : fastswitching ] . we plot the probe and coupling amplitudes versus time at a point near the center of the bec , and zoom in on the regions near the switch - on and -off . note that that while the coupling field smoothly varies to it s new value in each case , the probe field slightly overshoots the value given by ( [ eq : omegapadjust ] ) before returning to it . in all cases with @xmath189 the probe amplitude @xmath194 experienced damped oscillations before reaching its final value . the frequency of the oscillations was determined by @xmath195 and the damping rate was @xmath11 . * switching faster than the natural linewidth . * solid and dashed lines show , respectively , the normalized probe and coupling amplitudes at a point in the center of the bec @xmath196 when the switching is faster than the natural linewidth @xmath197 ( but otherwise the same parameters as fig . [ fig : switching](e ) ) . the two insets magnify the regions near the @xmath122 and @xmath184 . ] during these oscillations , the absorbing field @xmath141 is in fact quite significant . however , the maximum value of @xmath198 and the total atomic loss due to spontaneous emissions events was _ completely independent _ of @xmath161 over the range we explored . to understand this , we note that even though @xmath141 is larger , the frequency components of the quickly varying coupling field are spread over a range much wider than @xmath11 . thus a very small fraction of the energy in the field @xmath5 is near resonant with the atomic transition and it passes through the bec unattenuated . analyzing the problem in fourier space one can see that the increase in @xmath141 for smaller @xmath161 is precisely offset by this frequency width effect , leading us to the conclusion that spontaneous emission loss is independent of @xmath161 . unlike the case of switching slowly compared with @xmath21 , the adjustment of the probe is not a coherent process . rather , spontaneous emission damps the system until the dark state ( [ eq : omegapadjust ] ) is reached . however , because we are in the regime @xmath20 the amount of energy in the probe is much less than one photon , and thus still has virtually no impact on @xmath35 . this will not necessarily hold when @xmath199 and in @xcite it was predicted that there are , in fact , adiabatic requirements for @xmath161 in this regime . however , in all cases of interest here this inequality is well satisfied and there is no restriction on @xmath161 . we now turn to the question of the atomic dynamics during the storage time . these dynamics depend strongly on the relative scattering lengths of the states used , and the probe to coupling intensity ratio . we will see that the these dynamics are written onto revived probe pulses by switching the coupling field back on . the length scale requirement on variations in the relative amplitude @xmath150 derived above ( @xmath200 ) plays a central role in the fidelity with which these dynamics are written . we will demonstrate with several examples , relevant to current rb-87 experimental parameters . in @xcite the revived pulses were seen to be attenuated for longer storage times with a time constant of @xmath201 ms . the relative phase between the @xmath7 and @xmath8 ( or coherence ) , averaged over all the atoms , washed out because the atoms , although at a cold temperature of @xmath202 , were above the bose - condensation temperature and so occupied a distribution of energy levels . by contrast , a zero temperature two - component bec will maintain a well defined phase ( the phase between @xmath178 and @xmath110 ) at all @xmath36 , even as @xmath35 evolve . they evolve according to the two - component coupled gp equations ( [ eq : formalism12 ] ) with the light fields set to zero ( @xmath203 ) , and the initial conditions are determined by the superposition created by the input pulse at @xmath122 ( see fig . [ fig : switching](b ) ) . after such an evolution , we can then switch the coupling field back on and the probe will be revived according to ( [ eq : omegapadjust ] ) but with the _ new _ wavefunctions @xmath35 , evaluated at @xmath184 . here we examine a wide variety of different two - component bec dynamics which can occur after a pulse has been stopped . because of the initial spatial structure of @xmath35 created with the usl pulses , we see some novel effects in the ensuing dynamics . upon the switch - on , in many cases the probe will be revived via ( [ eq : omegapadjust ] ) . we label the spatial profile of the revived probe by @xmath204 . when we are in the weak probe limit ( @xmath205 or , stated in terms of the wavefunctions , @xmath206 ) , ( [ eq : omegapadjust ] ) becomes @xmath207 allowing us to make a correspondence between the revived probe and @xmath110 if @xmath93 and @xmath208 are known . the revived pulse will then propagate out of the bec to the pmt at @xmath17 . in the absence of any attenuation or distortion during the propagation out , the spatial features of the revived probe get translated into temporal features . thus if we observe the output @xmath209 then we would deduce that revived pulse was @xmath210 where @xmath211 is the time it takes the probe pulse to travel from some point @xmath36 to @xmath212 . note that we are relying here on the fact that we can switch the coupling field onto its full value of @xmath93 with a time scale fast compared to the pulse delay times @xmath213 . a slow ramp up would lead to a more complicated relationship between @xmath214 and @xmath215 . combining ( [ eq : proberevive ] ) and ( [ eq : probeout ] ) shows how the phase and amplitude information that was contained in @xmath110 at the time of the switch - on is transferred to the output probe @xmath215 . this transfer will be imperfect for three reasons . first , when sufficiently small spatial features are in @xmath110 , then @xmath148 is comparable to @xmath134 , giving rise to a significant @xmath141 . the resulting absorptions will cause deviations of @xmath216 from our expectation ( [ eq : proberevive ] ) . second , we mentioned how spatial features are translated into temporal features on the output probe . during the output , fast time features on the output probe will be attenuated via the bandwidth effect discussed in ( [ eq : tfreqwidth ] ) , affecting the accuracy of the correspondence ( [ eq : probeout ] ) . finally , stronger output probes , which will occur when @xmath217 at the time of the switch - on , make both the writing at the switch - on and the subsequent propagation out more complicated and less reliable . our following examples will demonstrate these considerations . interesting dynamics occur in when the two internal states are trapped equally ( @xmath218 ) and the scattering lengths @xmath219 are slightly different . we consider a case with @xmath220 rb-87 atoms and choose @xmath221 , and @xmath222 . the two lower states @xmath7 and @xmath8 are magnetically trapped with nearly identical magnetic moments , and we consider a trap with @xmath223 and assume a tranvserse area @xmath224 . these two states have an anomalously small inelastic collisional loss rate @xcite and so have been successfully used to study interacting two - component condensates for hundreds of milliseconds @xcite . the elastic scattering lengths are @xmath225 @xcite . * formation of interference fringes in rb-87 and writing fringes to probe light field . * * ( a ) * the initial wavefunctions created by a @xmath226 probe pulse . the coupling field is initially on at @xmath227 then switched off at @xmath228 . the cloud parameters and atomic levels are described in the text . the thick solid curve shows the density @xmath229 ( multiplied by 10 for clarity ) while the thick dashed curve shows the phase @xmath230 . the thin curve shows the density @xmath231 . the phase @xmath232 is approximately homogenous and not shown . * ( b ) * the effective potential @xmath233 ( [ eq : v2eff ] ) seen by the condensate in @xmath8 . * ( c ) * the resulting evolution of the density @xmath234 ( white represents higher density ) as the wave packet is pushed down the both sides of potential hill and interferes with itself upon reaching the hard wall in @xmath235 at the condensate edge . the vertical grey lines show the condensate boundary , and the horizontal grey lines indicate the times plotted in ( d ) . * ( d ) * plots of the density and phase of the wavefunctions at the indicated times ( with the same conventions as ( a ) ) at @xmath236 and @xmath237 . * ( e ) * spatial profiles of the probe and coupling field directly after a fast switch - on after a storage time @xmath238 corresponding to the times in ( d ) . the thick ( thin ) solid curves show the intensities @xmath239 and the dashed curve shows the phase @xmath240 ( @xmath241 is nearly homogenous and not plotted ) . * ( f ) * the time profile of the probe output intensity ( solid curve ) and phase ( dots ) at the output @xmath17 . ] figure [ fig : fringeex](a ) shows the initial wavefunctions @xmath35 after the input and stopping of a probe pulse . we see a gaussian shaped density profile in @xmath8 reflecting the input probe s gaussian shape . the subsequent dynamics are governed by eqs . ( [ eq : formalism12 ] ) with @xmath203 . in this case , a weak probe ( @xmath242 ) was input . as a result , @xmath243 is nearly constant in time and the evolution of @xmath110 is governed by essentially linear dynamics , with a potential determined by the magnetic trap _ and _ interactions with @xmath244 atoms : @xmath245 figure [ fig : fringeex](b ) shows this potential in this case . the hill in the middle has a height @xmath246 $ ] and arises because atoms feel a stronger repulsion from the condensate in @xmath7 when they are in @xmath27 . figures [ fig : fringeex](c - d ) show the subsequent dynamics . one sees that the @xmath8 condensate is pushed down both sides of the potential hill and spreads . however , once it reaches the border of the bec , it sees sharp walls from the trap potential . even for the fairly moderate scattering length difference here , there is sufficient momentum acquired in the descent down the hill to cause a reflection and formation of interference fringes near the walls . the wavelength of the fringes is determined by this momentum . what happens if one switches the coupling field back on after these dynamics ? figures [ fig : fringeex](e ) shows the revived probe pulses @xmath214 upon switch - ons at the times corresponding to fig . [ fig : fringeex](d ) . one sees a remarkable transfer of the sharp density and phase features of the @xmath8 condensate onto the probe field , according to ( [ eq : proberevive ] ) . because we are in the weak probe regime , the coupling field intensity only very slightly deviates from it s input value @xmath93 . figures [ fig : fringeex](f ) then shows the output probe @xmath215 . the sharp interference fringes are able to propagate out , though there is some attenuation and washing out of the features . note that the features at more positive @xmath36 propagate out first , leading to a mirror image - like relationship between the spatial fig . [ fig : fringeex](e ) and temporal fig . [ fig : fringeex](f ) patterns as predicted by ( [ eq : probeout ] ) . in this particular example , we successfully output many small features from the @xmath110 to the output probe field . here the fringes are @xmath247 , which is still larger than the absorption length @xmath248 , but not substantially so , leading to some small amount of dissipation during the switch - on and output . this gives us a sense of the `` information capacity '' . the number of absorption lengths , or optical density , which in this case is @xmath249 , ultimately limits the number of features ( for a given desired fidelity ) which could be successfully written and output . note also that in the @xmath250 case , some of the @xmath110 amplitude occupies the entering region @xmath251 leading to additional imperfections in the writing process in this region . * breathing behavior and long storage * in this case the states chosen for @xmath7 and @xmath252 are reversed , changing the sign of the curvature in the effective potential ( [ eq : v2eff ] ) in ( a ) . all plots are the plotted with the same conventions as fig . [ fig : fringeex ] except we additionally plot thin dotted curves in ( c ) showing the phase @xmath253 and in ( d ) showing the phase @xmath254 . ] the dynamics in the previous example are quite dramatic , but are not particularly conducive to preserving or controllably processing the information in the bec . for this , it would be preferable to switch the roles of @xmath7 and @xmath8 . such a case is shown in fig . [ fig : breatheex2 ] . because @xmath255 in this case , the potential hill is turned into a trough ( see fig . [ fig : breatheex2](a ) ) . the effective potential in the region of the @xmath7 condensate , in the thomas - fermi limit , becomes @xmath256 $ ] and so it harmonic , with a much smaller oscillator frequency than the magnetic trap . the evolution can be easily calculated by decomposing the wavefunction @xmath110 into a basis of the harmonic oscillator states of this potential . in the example here , there is significant occupation of the first several oscillator levels , and so one sees an overall relative phase shift in time ( from the ground state energy of the zeroeth state ) , a slight dipole oscillation ( from occupation of the first excited state ) and breathing ( from the second ) . after one oscillator period ( 310 ms in the case shown ) , @xmath110 replicates its original value at the switch - off . in this case , the dynamics are quite gentle , and so the spatial scales of the phase and density features are always quite large compared with the absorption length @xmath142 . the fidelity of the writing and output of the information on the probe field ( shown in figs . [ fig : breatheex2](d - e ) ) is correspondingly better than in the previous example . additionally , because the @xmath27 is trapped near the bec center , this case avoids problems associated with @xmath110 occupying the region @xmath159 . because of both the ease of analyzing the evolution and the high fidelity of outputting the information , we expect this case to be well - suited to controlled processing of optical information . for example , if the input pulse created a wavefunction @xmath110 corresponding to the ground state of the oscillator potential , the evolution would result only in a homogenous phase shift , proportional to the storage time , allowing long storage of the information or introduction of controllable phase shifts . by choosing the pulse lengths differently so several oscillator states are occupied , one could achieve linear processing or pulse reshaping . furthermore , note that in this example there is a small but discernable dip in the @xmath257 density at 51 ms , indicating some nonlinearity in the evolution of @xmath110 . one can tune this nonlinearity by varying the probe to coupling ratio , leading to nonlinear processing of the information . * strong probe case : formation of vector solitons . * dynamics resulting from using a relatively strong probe input @xmath258 . in this case we have @xmath259 rb-87 atoms in a @xmath223 trap and use @xmath260 , and @xmath261 corresponding to the system @xmath262 , and @xmath263 . we artificially set @xmath264 to exaggerate the phase separation dynamics . * ( a ) * evolution of density in @xmath8 @xmath265 remains almost constant in time . * ( c ) * the densities @xmath266 are plotted as dotted and dashed curves . ] the two - component dynamics are even richer when one uses strong probe pulses so nonlinear effects become very evident in the evolution . in such a case , the qualitative features of the dynamics will be strongly effected by whether or not the relative scattering lengths are in a phase separating regime @xmath267 @xcite . experiments have confirmed that the @xmath268 studied in our previous examples in rb-87 are very slightly in the phase separating regime @xcite . relative scattering lengths can also be tuned via feshbach resonances @xcite . figure [ fig : solitondynex ] shows the evolution of the two components following the input and stopping of a stronger probe ( @xmath269 ) . in this case we chose our levels to be @xmath270 , and @xmath263 , which would require an optical trap @xcite in order to trap the @xmath7 and @xmath8 equally . this level scheme is advantageous in the strong probe case since @xmath138 which , as we will discuss in more detail in section [ subsec : f23 ] , improves the fidelity of the writing and output processes . we chose the scattering length @xmath264 , higher than the actual background scattering length , to exaggerate the phase separation dynamics . one sees in fig . [ fig : solitondynex](a),(c ) that over a 30 ms timescale , the phase separation causes the density in @xmath27 to become highly localized and dense . this occurs because the scattering length @xmath271 causes @xmath7 atoms to be repelled from the region occupied by the @xmath8 atoms , and in turn the @xmath8 atoms find it favorable to occupy the resulting `` well '' in the @xmath7 density . these two processes enhance each other until they are balanced by the cost of the kinetic energy associated with the increasingly large spatial derivatives and we see the formation of two - component ( vector ) solitons @xcite . in the case here , two solitons form and propagate around the bec , even interacting with each other . the alternating grey and white regions along each strip in fig . [ fig : solitondynex](a ) indicate that the solitons are undergoing breathing motion on top of motion of their centers of mass . [ fig : solitondynex](b ) shows that the total density profile @xmath151 varies very little in time . it is the _ relative _ densities of the two components that accounts for nearly all the dynamics . we found solitons formed even in only slightly phase separating regimes ( @xmath272 ) . the number of solitons formed , the speed of their formation , and their width were highly dependent on this ratio as well as the number of atoms in @xmath8 . the ability of stopped light pulses to create very localized two - component structures seems to be a very effective method for inducing the formation of vector solitons , which has hitherto been unobserved in atomic becs . a full exploration of these dynamics is beyond our scope here . * writing solitons onto probe field . * * ( a ) * writing of vector solitons onto probe and coupling fields directly after a switch - on after a storage time of @xmath273 and * ( b ) * probe output . the plots use the same conventions as fig . [ fig : breatheex2](d - e ) . ] these non - trivial features can also be written onto probe pulses as shown in fig . [ fig : solitonoutput](a ) . note that the condition for coherent revivals @xmath274 does _ not _ depend on the weak probe limit . therefore , the writing process is still primarily a coherent process , and ( [ eq : omegapadjust ] ) is still well satisfied . however , we see in fig . [ fig : solitonoutput](a ) , the coupling field is strongly affected by the writing and is far from homogenous , meaning the weak probe result ( [ eq : proberevive ] ) can not be used . nonetheless , we see that the qualitative features of both the density and phase of @xmath110 have still been transferred onto @xmath6 . the writing in the strong probe regime will be studied in more detail in section [ sec : systematic ] . in fig . [ fig : solitonoutput](b ) we plot the resulting output probe pulse in this case . as in the weak probe case there is some degradation in the subsequent propagation due to attenuation of high frequency components . in addition , unlike the weak probe case , there are nonlinearities in the pulse propagation itself which causes some additional distortion . section [ subsec : f23 ] also addresses this issue . even so , we again see a very clear signature of the two solitons in the both the intensity and phase of the output probe . we have performed some preliminary simulations in traps with weaker transverse confinement and which calculate the evolution in the transverse degrees of freedom . they show that the solitons can break up into two - component vortex patterns via the snake instability @xcite . we chose our parameters to be in the phase separating regime here . just as in the weak probe case , other regimes will lead instead to much gentler dynamics . for example , it has been shown @xcite that when @xmath275 then there exist `` breathe - together '' solutions of the bec whereby complete overlap of the wavefunctions @xmath35 persists . in this case the nonlinear atomic interaction can lead to spin - squeezing @xcite . it would be extremely interesting to investigate a probe - revival experiment in such a case , to see to what extent the squeezed statistics are written into probe pulse , producing squeezed light @xcite . this would require taking both the light propagation and atomic dynamics in our formalism beyond the mean field . steps in this direction have been taken in @xcite , but these analyses are restricted to the weak probe case . the above examples demonstrate that a rich variety of two component bec dynamics can occur , depending on the relative interaction strengths and densities of the states involved . they also show that remarkably complicated spatial features in both the density and phase can be written into temporal features of a probe light pulse and output . the purpose of this section is to quantify the fidelity with which wavefunctions can be written onto the probe field . to do this we consider an example of a two - component bec with a gaussian shaped feature in @xmath110 , with parameters characterizing length scales and amplitudes of density and phase variations . note these gaussian pulse shapes are relevant to cases in which one might perform controlled processing , _ e.g. _ fig . [ fig : breatheex2 ] . switching on a coupling field then generates and outputs a probe pulse with these density and phase features . we will calculate this output , varying the parameters over a wide range . in each case , we will compare the output to what one expect from an `` ideal '' output ( without dissipation or distortion ) , and calculate an error which characterizes how much they differ . using the analysis of the switching process in section [ sec : switching ] and the usl propagation in section [ sec : formalism ] we also obtain analytic estimates of this error . in the weak probe case , we find a simple relationship between @xmath276 and the output @xmath215 . we find that to optimize the fidelity in this case one should choose the gaussian pulse length between two important length scales . the shorter length scale is determined by eit bandwidth considerations , which primarily contribute error during the output . on the other hand , for large pulses comparable to the total condensate size , the error during the writing process dominates due to @xmath110 partially occupying the condensate entering edge region ( @xmath159 ) . in the strong probe case , we find that the relationship between @xmath276 and @xmath215 is more complicated , however , when @xmath138 , one can still make an accurate correspondence . we find the fidelity in this case depends only weakly on the probe strength . conversely , when @xmath277 , even a small nonlinearity causes additional phase shifts and distortions , making this correspondence much more difficult , leading to higher errors for stronger probes . for this reason , a system that uses , for example @xmath278 ( as in fig . [ fig : solitonoutput ] ) would be preferable when one is interested in applications where there are comparable densities in the two states . by considering the switch - on and output processes , starting with an arbitrary initial two - component bec , we emphasize that this method of outputting the atomic field information onto light fields works regardless of how the bec state was generated . one could prepare a bec in a coherent superposition of @xmath7 and @xmath8 by any available method including , but not limited to , inputting a slow light pulse . assume we have a two component bec with wavefunctions @xmath35 ( and again define @xmath279 ) . for now we assume @xmath138 ( so @xmath280 , see eq . ( [ eq : alphadef ] ) ) , as we treat the @xmath281 case in section [ subsec : f23 ] . upon a rapid switch - on of a coupling field with amplitude @xmath93 , inverting ( [ eq : dalight ] ) shows the pattern written onto the probe field will be : @xmath282 in the ideal limit , where the spatial variations of the wavefunction are sufficiently small , @xmath283 and the second term vanishes . furthermore , ( [ eq : dprop ] ) shows in this case that the dark field intensity is constant @xmath284 . so in the ideal limit we have for the amplitude : @xmath285 in practice this will be a good approximation so long as the inequality @xmath286 is satisfied . the extent to which the second term in ( [ eq : probereviveex ] ) can not be neglected will determine the error between the ideal @xmath287 and actual @xmath288 output . note that in all the cases studied in section [ sec : processing ] , @xmath289 was nearly constant in time , being always well approximated by the original ground state @xmath290 . thus @xmath291 can be considered a known function of @xmath36 . turning now to the phase , the simple relationship @xmath292 is always satisfied in the dark state ( see ( [ eq : omegapadjust ] ) ) . if @xmath241 was just a constant then the probe phase would simply reflect the relative phase of the two wavefunctions @xmath293 ( choosing our phase conventions so @xmath294 ) . this is indeed the case in the weak probe limit , as then the coupling field phase @xmath241 is unaffected by the atomic fields upon the switch - on . however , this is not necessarily so in the strong probe regime . to calculate the true phase shift , we calculate phase of @xmath145 from ( [ eq : dprop ] ) . in the ideal limit @xmath295 ( choosing our phase conventions such that @xmath296 ) . defining @xmath297 and using the definition of @xmath154 ( [ eq : alphadef ] ) and ( [ eq : proberevivemag ] ) , we find the relationship : @xmath298 the equation is written in this way ( the atomic phase in terms of the probe phase and intensity ) because our purpose is to derive the phase pattern @xmath299 based on the observed probe field , rather than vice versa . the nonlinear correction @xmath300 is fact the phase imprinted on the coupling field @xmath241 during the switch - on . however , in an experiment , we do not have direct access to the revived probe @xmath301 . rather , we observe the temporal output @xmath302 . in practice , the intensity of this quantity can be measured with a pmt , while the phase pattern could be measured by beating it with a reference probe field which did not propagate though the bec . in the ideal limit ( the absence of attenuation or distortion ) the relationship ( [ eq : probeout ] ) will link the observed output to the revived probe field @xmath303 . thus we define the `` ideal '' output via @xmath304 in practice , additional absorption events and distortion can occur during the output . if variations of the relative amplitude @xmath150 have some characteristic scale @xmath305 then time features with a scale @xmath306 will be introduced into the probe @xmath307 . using our result for the bandwidth induced attenuation ( [ eq : tfreqwidth ] ) yields an estimate for transmission energy ( that is , the energy of the actual output pulse @xmath308 relative to an unattenuated output of @xmath216 ) : @xmath309 for this simple estimate we evaluate @xmath134 and @xmath310 at the location of the center of the pulse @xmath311 , though one could also construct more sophisticated estimates by integrating over the spatial distribution of the pulse . just as in our discussion below eq . ( [ eq : tfreqwidth ] ) , the temporal width is increased by a factor @xmath117 and the peak intensity reduced by @xmath116 during the propagation out . we now apply these findings to a gaussian pulse , with @xmath110 assumed to be of the form , @xmath312\ ] ] note @xmath110 has a density feature of amplitude @xmath313 and length @xmath314 and a phase feature of amplitude @xmath315 and length @xmath316 . for @xmath317 we choose the amplitude so the total density matches the ground state for the trap @xmath318 and in the phase we put in a shift with some amplitude @xmath319 and length @xmath320 . an example is shown in fig . [ fig : test1](a ) . these parameters will be varied throughout this section to learn how they effect the writing and output . we choose a @xmath321hz trap ( @xmath322 ) with @xmath323 rb-87 atoms . we use @xmath260 , corresponding to the system @xmath262 , and @xmath263 in rb-87 . * writing and outputting atomic field information onto the probe fields . * * ( a ) * a test case with a density feature of length @xmath324 of amplitude @xmath325 and phase features of @xmath326 ( see eq . ( [ eq : psi2test ] ) ) and @xmath327 . the conventions for the curves are the same as fig . [ fig : solitondynex](c ) . * ( b ) * upon switch - on of the coupling field with @xmath328mhz , we show the dark intensity @xmath329 ( thick solid curve ) and coupling intensity @xmath330 ( thin solid ) . the dotted , dashed , and dot - dashed curves show , respectively , the phase profiles @xmath241 , @xmath331 and @xmath299 . * ( c ) * the normalized probe intensity profile @xmath332 ( solid curve ) is indistinguishable from the expected profile @xmath333 ( thick dashed ) ( see eq . ( [ eq : proberevivemag ] ) ) . the thin solid curve shows @xmath240 and the dot - dashed curve shows @xmath334 , while the thin dashed curve shows the expected probe phase profile with the nonlinear correction @xmath335 ( [ eq : proberevivephase ] ) . for reference , the dotted curve shows the total density @xmath151 profile . * ( d ) * the temporal pattern of the output probe intensity @xmath336 ( thick solid curve ) and phase @xmath337 ( thin solid ) , versus an estimate based on perfect propagation of the revived pulse @xmath303 according to ( [ eq : probeout ] ) ( dashed and dotted curves ) . the dot - dashed curve shows the ideal output attenuated by @xmath116 ( see ( [ eq : tspace ] ) and the discussion afterward ) using the estimate @xmath338 . ] in the example of fig . [ fig : test1](a ) , @xmath325 is not particularly small , so the weak probe limit can not be assumed . figure [ fig : test1](b ) shows the spatial profiles of the dark field @xmath145 and coupling field @xmath5 immediately following a fast switch - on ( @xmath124 ) . we see very little attenuation of @xmath145 occurs across the bec , as @xmath339 is quite small compared with @xmath340 . translating this back into the @xmath341 basis , ( [ eq : dalight ] ) predicts that @xmath5 acquires a dip in intensity with a height proportional to the density @xmath342 . this behavior is indeed seen in the figure . the phase difference @xmath299 is plotted as the dot - dashed curve in figure [ fig : test1](b ) . one sees that in the region where @xmath299 is inhomogeneous , a small phase shift , equal to @xmath300 ( [ eq : proberevivephase ] ) , is introduced in the coupling field ( dotted curve ) . again this shift only arises in the strong probe regime . the phase shift in the dark field @xmath331 is plotted as the dashed curve . figure [ fig : test1](c ) then compares written probe field to our ideal limit predictions ( [ eq : proberevivemag])-([eq : proberevivephase ] ) in this example . the numerically calculated intensity @xmath343 ( solid curve ) is almost indistinguishable from our prediction @xmath344 ( [ eq : proberevivemag ] ) ( dashed curve ) . the phase written onto the probe field @xmath345 is also very close to the ideal limit prediction ( [ eq : proberevivephase ] ) . one sees that including the nonlinear correction @xmath346 to the simpler estimate @xmath334 ( dot - dashed curve ) is important in making the comparison good . figure [ fig : test1](d ) then shows intensity ( thick solid curve ) and phase ( thin solid ) of the output pulse @xmath347 . for comparison the dashed and dotted curves show , respectively , the output expected from an unattenuated and undistorted transmission of the revived pulse @xmath348 via ( [ eq : probeoutideal ] ) . one sees a very good agreement in the phase pattern , while there is a visible reduction in the intensity , due to bandwidth considerations . the dot - dashed curve shows the our estimate with the estimated reduction in intensity @xmath116 ( see ( [ eq : tspace ] ) ) using a characteristic length scale @xmath305 calculated from a quadrature sum of contributions from the amplitude and phase features in ( [ eq : psi2test ] ) ( the expression is given in the caption ) . we have thus calculated a method by which the output probe pulse can be solely used to calculate the relative density and phase of the wavefunctions which generated it , and demonstrated the method with a generic example . furthermore , we have identified the leading order terms which will cause errors in these predictions . in particular we have seen in our example in fig . [ fig : test1](d ) that including the expected bandwidth attenuation accounts for most of the deviation between our ideal predictions and the actual output pulse . we now quantify the deviations from our predictions ( [ eq : proberevivemag]-[eq : probeoutideal ] ) for our example ( [ eq : psi2test ] ) , varying the length and amplitude parameters over a wide range . these results can be directly applied to pulses which are approximately gaussian ( as in fig . [ fig : breatheex2 ] ) . the results here should also provide a good guide to the expected fidelity in more complicated cases so long as the length scale and amplitude of features can be reasonably estimated . to quantify the deviation from our ideal case prediction ( [ eq : proberevivemag])-([eq : proberevivephase ] ) we define the write error : latexmath:[\[\label{eq : ewrite } e_w \equiv \frac{\int_{z_{\mathrm{in}}}^{z_{\mathrm{out } } } \ , dz ' \bigg\|\,\omega_p^{(\mathrm{rev } ) } - e^{i(\phi_{21}+\phi_{p}^{\mathrm{(nl)}}+\pi)}\bigg\|^2 } { \int_{z_{\mathrm{in}}}^{z_{\mathrm{out } } } \ , dz ' we plot this quantity in a series of cases with different @xmath314 in fig . [ fig : serieslength](a ) ( circles ) in a case with amplitude @xmath325 and no phase profiles . this is compared with a calculated prediction for @xmath350 ( solid curve ) based on ( [ eq : probereviveex ] ) where we calculate @xmath141 with ( [ eq : omegaa ] ) and calculate the small attenuation of @xmath145 with ( [ eq : dprop ] ) . these errors grow as @xmath314 becomes comparable to @xmath351 , according to the discussion in section [ subsec : daanalysis ] . we see the agreement between the analytic and numerical estimates is quite good for small to moderate @xmath314 , confirming that this is the leading source of error . however , when @xmath314 becomes comparable to the total bec size ( the thomas - fermi radius is @xmath352 here ) , we begin to see additional errors because @xmath110 becomes non - zero at the bec edge . thus , we see that for a given bec length and density , one must choose @xmath353 to minimize the write error @xmath350 . because this edge effect depends on the density structure near the bec edge , it is difficult to estimate analytically , but it is related to the fraction of the pulse in or near the region @xmath159 . for pulses near the bec center , this will depend on @xmath354 , whereas for pulses far off center , it will also depend on the location of the center of the pulse . we also plot a series of cases keeping @xmath355 and @xmath356 but varying the length scale @xmath316 of a phase profile of amplitude @xmath357 . one sees similar but slightly smaller errors @xmath350 in this case . * writing and outputting atomic field information onto light fields . * * ( a ) * circles show the write error @xmath350 ( [ eq : ewrite ] ) versus the length @xmath314 for becs with no phase jumps ( @xmath358 ) and with @xmath325 ( see eq . ( [ eq : psi2test ] ) ) . the solid curve shows an analytic result based on the analytic expressions ( [ eq : probereviveex ] ) , ( [ eq : omegaa ] ) and ( [ eq : dprop ] ) . the squares show a case holding @xmath359 and introducing a phase feature @xmath357 with varying length scale @xmath316 . * ( b ) * the output error @xmath360 ( [ eq : eout ] ) for the same cases as ( a ) . the solid curves showing the prediction @xmath361 discussed in the text ( choosing @xmath362 ) for the case with no phase jumps ( circles ) . the insets show the output in the cases indicated with arrows , with the same plotting conventions as fig . [ fig : test1](d ) ( in the lower inset on the right @xmath363 is homogenous and not plotted ) , we again use @xmath321hz , @xmath322 , @xmath364 and @xmath260 , corresponding to the system @xmath262 , and @xmath263 in rb-87 . ] we define the error accumulated as @xmath303 propagates out using ( [ eq : probeout ] ) : @xmath365(b ) we plot this quantity for the cases corresponding to fig . [ fig : serieslength](a ) . for comparison , for the series with no phase shift , we calculated an estimate based on the expected attenuation @xmath116 and spreading @xmath117 of a gaussian pulse , due bandwidth considerations ( see eq . ( [ eq : tspace ] ) and subsequent discussion ) , choosing @xmath362 . in the limit of small @xmath366 this calculation yields @xmath367 . this is plotted as a solid curve and we see good agreement with the numerical data . the error @xmath368 is seen to dominate @xmath350 for small @xmath314 , due to the fact that the large optical density effects the former ( @xmath369 in the case plotted ) . thus our analytic estimate @xmath370 is a good estimate of the total error . but for larger @xmath314 the edge effect in @xmath350 becomes important . to minimize the total error @xmath371 we should choose @xmath314 so the error @xmath372 is comparable to the edge effect error . in fig . [ fig : serieslength ] the optimal length is @xmath373 and the total error is @xmath374 . the scaling of @xmath375 with the condensate size @xmath376 is difficult to estimate . assuming it roughly increases as @xmath377 , then ( [ eq : tspace ] ) shows @xmath378 , giving us a guide as to the improvement in fidelity we can expect by increasing the total optical density of the condensate . the insets show the plots of the ideal output of @xmath301 ( solid curves ) versus actual outputs @xmath302 ( dotted curves ) in two of the cases . in one case , with no phase jumps and a small @xmath314 , we see an overall reduction in amplitude and slight spreading . in the other , with a larger @xmath314 but a phase shift with a small length scale @xmath316 , we see that the attenuation is localized in the middle of the pulse . this is because it is the components which contain the sharpest phase profile , near the middle , which are most severely attenuated . generally , in cases with complicated spatial features , one must be aware of this potential for local attenuation of the sharpest features . note , however , that in many cases of interest , such as the interference fringes ( fig . [ fig : fringeex ] ) and solitons ( fig . [ fig : solitondynex ] ) , large phase shifts occur primarily in regions of low density , meaning these features often survive during the output propagation . * writing and outputting atomic field information onto light fields . * * ( a ) * the write error @xmath350 ( [ eq : ewrite ] ) versus @xmath379 for the case @xmath380 . * ( b ) * the output error @xmath360 ( [ eq : eout ] ) in the same cases . the solid line shows the estimate @xmath361 ( see eq . ( [ eq : tspace ] ) ) using @xmath381 . the condensate parameters are the same as in fig . [ fig : serieslength ] . ] our analysis of switching process in section [ sec : switching ] applies even with a large amplitude @xmath379 , as was seen in the @xmath325 case in fig . [ fig : test1 ] . in fig . [ fig : seriesamp](a ) we plot the results of a series of simulations with @xmath379 varying all the way up to @xmath382 and see only a very small impact on @xmath350 for @xmath383 . most of our results for the propagation of usl pulses , _ e.g. _ ( [ eq : vg ] ) and ( [ eq : tfreqwidth ] ) , rely on the weak probe limit . surprisingly though fig . [ fig : seriesamp](b ) shows that the output error @xmath360 is independent of @xmath379 through @xmath384 . the lack of distortion in the strong probe cases discussed so far in this section is a result of us choosing a system with equal oscillator strengths @xmath138 . the importance of the relative size of the oscillator strengths can be understood by considering the adiabatons ( the changes in the output coupling intensity seen in fig . [ fig : switching](a , e ) ) . the adiabatons arise because of the coherent flow of photons between the two light fields and their amplitude is determined by the requirement that the sum of the number of photons in both fields is constant upon such an exchange . in a case where @xmath385 one must take into account that it is the total field strength @xmath386 , rather than simply the input coupling strength @xmath387 which enters the numerator in our equation for the group velocity @xmath388 ( [ eq : vg ] ) . however , the expressions for the rabi frequencies involve the oscillator strengths of the relevant transitions . thus , preserving the total number of _ photons _ at each point in space will result in a homogenous @xmath389 if and only if @xmath138 . during the time that the probe is completely contained in the bec , @xmath390 will differ from @xmath387 in the vicinity of the pulse according to : @xmath391 , where we compare a delayed weak probe regular usl output pulse to cases with strong probes @xmath392 and vary @xmath393 ( @xmath75 in all cases ) . each time we chose the coupling intensity so that that the rabi frequency is @xmath227 . one sees that when @xmath138 there is no asymmetrical distortion , though we get a slight spreading and a shorter delay time . however , in the other two cases , there is asymmetry in the output pulses , consistent with the picture that the center of the pulse will travel slower or faster than the edges according to ( [ eq : adiabaton ] ) and ( [ eq : vg ] ) ( with @xmath387 replaced by @xmath390 in the numerator ) . the result is a breakdown of our estimate ( [ eq : probeoutideal ] ) . as this is a propagation effect , the absolute magnitude of this effect should roughly be linear with optical density and the relative distortion will go as the optical density divided by the pulse length . this is the primary reason it would be advantageous to use the levels we chose in our simulation in fig . [ fig : solitonoutput ] if there is a strong probe . * usl with a strong probe . * the dotted curve shows a strong probe pulse amplitude with peak value @xmath394 input into a sodium cloud with an optical density @xmath395 . we use an oscillator strength @xmath75 but vary the coupling oscillator strength @xmath396 . the output @xmath397 is shown in the cases @xmath398 ( thick solid curve ) , @xmath399 ( dot - dashed curve ) , and @xmath76 ( dashed curve ) . for reference , the thin solid curve shows output for a weak input probe @xmath400 . ] we just outlined the reason that the equal oscillator strengths can be important in reducing distortion during propagation and thus @xmath360 . it turns out that the writing process is also more robust when @xmath138 . the reason is the presence of the additional term @xmath136 ( [ eq : alphadef ] ) . this term is quite small in the weak probe limit , but in the strong probe case leads to an additional phase shift , not accounted for in ( [ eq : proberevivephase ] ) , which depends in detail on the pulse amplitude and structure . figs . [ fig : test2](a - d ) shows a case , similar to fig . [ fig : test1 ] , but with @xmath401 ( as in the cases in figs . [ fig : fringeex]-[fig : breatheex2 ] ) , and with a fairly small nonlinearity @xmath402 . one sees in fig . [ fig : test2](b ) that there is a dip in @xmath145 in the region of the probe due to the @xmath136 term . fig . [ fig : test2](c ) shows the written probe pulse . there is a small but discernable difference between the predicted @xmath403 ( [ eq : proberevivephase ] ) and actual @xmath404 phase . [ fig : test2](d ) then shows the asymmetric distortion which develops as the strong probe propagates out . one sees the phase jump is distorted in addition to the amplitude . [ fig : test2](e - f ) demonstrate how these effects lead to substantially higher errors @xmath350 and @xmath360 as @xmath379 becomes larger . from our estimate from the previous section we expect an output error of @xmath405 and we see in fig . [ fig : test2](f ) that the distortion effect leads to errors higher than this when @xmath406 . * writing information with unequal oscillator strengths . * * ( a - d ) * plots are for the same parameters as fig . [ fig : test1 ] except with @xmath407 and @xmath402 . * ( e - f ) * write @xmath350 and output @xmath360 errors versus @xmath379 for the parameters @xmath408 . ] we thus conclude that fidelity of both the writing and output is compromised in a system with unequal oscillator strengths . however , this only comes into effect in cases where the nonlinearity is important . in weak probe cases neither of these effects is important and the fidelity can still be estimated with the analysis of the @xmath138 case in sections [ subsec : fidelity]-[subsec : error ] . in conclusion , we have established first several new results regarding the fast switching of the coupling field in light storage experiments . we found the switching could be done arbitrarily fast without inducing absorptions so long as the probe is completely contained in the atomic medium ( and @xmath20 which is usually the case in practice ) . we have also seen that when the switching is slow compared to the excited state lifetime ( @xmath409 ns ) , the probe smoothly follows the temporal switching of the coupling field , while in the other limit the probe amplitude undergoes oscillations which damp out with this time scale ( see fig . [ fig : fastswitching ] ) . next , we saw that these stopped pulses can induce novel and rich two - component becs dynamics during the storage time . both the relative scattering lengths of the states used and the probe to coupling intensity ratio have a strong effect on the qualitative features of the dynamics . in the weak probe case , the @xmath27 condensate sees en either an effective repulsive hill ( when @xmath410 , see fig.[fig : fringeex](b ) ) or harmonic oscillator potential ( when @xmath255 , see fig.[fig : breatheex2](a ) ) . the characteristic timescale for the dynamics is given my the chemical potential of the bec ( in our cases @xmath411khz ) times the relative scattering length difference @xmath412 . in the latter case , the resulting evolution can be easily calculated by decomposing the pulse into the various harmonic oscillator eigenstates . thus it is possible to choose input pulses which preserve their density over time or undergo predictable reshaping such as dipole sloshing or breathing , allowing controlled storage and linear processing of optical information . inputting stronger probes will add nonlinearity to the evolution , making nonlinear processing possible . very strong probes and phase separating scattering lengths @xmath413 lead to the formation and motion of vector solitons , which have not been observed in becs to date . we then showed that switching the coupling field on after the dynamics writes the various density and phase features of the wavefunctions onto revived probe pulses . this was seen qualitatively for various examples ( figs . [ fig : fringeex]-[fig : solitonoutput ] ) . a precise relationship between the wavefunctions and output pulses was found ( [ eq : proberevivemag])-([eq : probeoutideal ] ) , meaning the output probe pulse can be used as a diagnostic of the relative density and phase in the bec wavefunctions . we have also identified sources of attenuation and distortion in the writing and output processes and quantitative errors were calculated for a wide range of length and amplitudes of gaussian pulses ( figs . [ fig : serieslength]-[fig : seriesamp ] ) . as we saw there , the error during the output dominates for shorter pulses and is @xmath361 ( see eq . ( [ eq : tspace ] ) ) . for longer pulses effects due to the condensate edge introduce errors into the writing process . balancing these two considerations one can optimize the fidelity , which improves with optical density . for the strong probe case , we found that for equal oscillator strengths ( @xmath414 ) , one could still relate the wavefunctions to the output pulses by taking into account an additional nonlinear phase shift ( [ eq : proberevivephase ] ) . the fidelity of transfer of information was virtually independent of the probe strength @xmath415 even when @xmath416 ( see fig . [ fig : seriesamp ] ) . for unequal oscillator strengths ( @xmath417 ) we found strong probes lead to additional features in the phase pattern during the writing process and in distortion during the propagation during the output , leading to much higher errors ( see fig . [ fig : test2 ] ) . looking towards future work , we note that it is trivial to extend the analysis here to cases where there are two or more spatially distinct becs present . so long as they are optically connected , information contained in the form of excitations of one bec could be output with this method and then input to another nearby bec , leading to a network . in the near future , there is the exciting possibility of extending these results beyond the mean field to learn how using this method of writing onto light pulses could be used as a diagnostic of quantum evolution in becs . a particular example of interest would be to investigate the spin squeezing due to atom - atom interactions in a two - component system and the subsequent writing of the squeezed statistics onto the output probe pulses . furthermore , we expect that performing revival experiments after long times could be used as a sensitive probe of decoherence in bec dynamics , similar to the proposal in @xcite . lastly , there is the prospect of inputting two or more pulses in a bec and using controllable nonlinear processing from atom - atom interactions to design of multiple bit gates , such as conditional phase gates . we anticipate the results presented here can be applied to these problems as well as to applications in quantum information storage , which require the ability to transfer coherent information between light and atom fields . we arrived at eqs . ( [ eq : formalism12])-([eq : formalismpc ] ) by first considering a system of equations with all three levels considered and then adiabatically eliminating the wavefunction for @xmath418 . for completeness we write the original equations here . @xmath419 \psi_1 \nonumber \\ & & \hspace{1 cm } + \frac{1}{2}\hbar \ , \omega_p^ * \psi_3 , \\ \label{eq : fullgp2 } i \hbar \frac{\partial \psi_2}{\partial t } & = & \bigg[-\frac{\hbar^2}{2m}\frac{\partial^2}{\partial z^2 } + v_2(z ) + u_{22 } & & \hspace{1 cm } + \frac{1}{2}\hbar \ , \omega_c^ * \psi_3 , \\ \label{eq : fullgp3 } i \hbar \frac { \partial \psi_3}{\partial t } & = & \frac{1}{2 } \omega_p \hbar \ , \psi_1 + \frac{1}{2 } \omega_c \hbar \psi_2 - i \frac{\gamma}{2 } \psi_3\end{aligned}\ ] ] in adiabatically eliminating @xmath190 , we assume all quantities vary slowly compared to excited state lifetime @xmath21 ( @xmath422 ns in sodium ) . this allows us to set @xmath423 in ( [ eq : fullgp3 ] ) and arrive at : this work was supported by the u.s . air force office of scientific research , the u.s . army research office osd multidisciplinary university research initiative program , the national science foundation , and the national aeronautics and space administration . the authors would like to thank janne ruostekoski for helpful discussions . phillips , a. fleischhauer , a. mair , r.l . walsworth , and m.d . lukin , phys . * 86 * , 783 ( 2001 ) ; a. mair , j. hager , d.f . phillips , r.l . walsworth , and m.d . lukin , phys . a * 65 * , 031802 ( 2002 ) . l.v . hau , `` bec and light speeds of 38 miles / hour '' , feb . 10 , 1999 , workshop on bose - einstein condensation and degenerate fermi gases , center for theoretical atomic , molecular , and optical physics , ( boulder , co ) , http://fermion.colorado.edu/@xmath58chg/talks/hau ; l.v . hau , `` bose - einstein condensation and light speeds of 38 miles per hour '' , harvard university physics dept . colloquium , feb . 22 , 1999 , videocasette ( harvard university , cambridge , ma , 1999 ) . m. inguscio , s. stringari , and c. wieman , eds . , _ bose - einstein condensates in atomic gases , proceedings of the international school of physics enrico fermi , course cxl _ , ( international organisations services b.v . , amsterdam , 1999 ) . myatt , e.a . burt , r.w . ghrist , e.a . cornell , and c.e . weiman , phys . lett . * 78 * , 586 ( 1997 ) ; j.p . burke , c.h . greene , j.l bohn , phys . lett . * 81 * , 3355 ( 1998 ) . a mathematica notebook is available at http://fermion.colorado.edu/@xmath58chg / collisions/. the code uses a crank - nicolson technique for the external propagation of the atomic fields and a symmmetric second order propagation for the light field internal coupling terms . the light fields are propagated in space with a second - order runge - kutta algorithm . the time steps varied between @xmath425 ns during the light propogation to as high as @xmath426 during the storage time with the light fields off . a spatial grid with typical spacing @xmath427 was used . the intervals @xmath428 were varied to assure they did not effect the results . in this code , described in @xcite , we keep track of the amplitudes @xmath429 and propagate them and @xmath173 according to the analogues of eqs . ( [ eq : fullgp1])-([eq : maxwell ] ) . we do not include the atomic dynamics in this code , which are on a much slower time scale .	we theoretically explore coherent information transfer between ultra - slow light pulses and bose - einstein condensates ( becs ) and find that storing light pulses in becs allows the coherent condensate dynamics to process optical information . we consider becs of alkali atoms with a @xmath0 energy level configuration . in this configuration , one laser ( the coupling field ) can cause a pulse of a second pulsed laser ( the probe field ) to propagate with little attenuation ( electromagnetically induced transparency ) at a very slow group velocity ( @xmath1 m / s ) , and be spatially compressed to lengths smaller than the bec . these pulses can be fully stopped and later revived by switching the coupling field off and on . here we develop a formalism , applicable in both the weak and strong probe regimes , to analyze such experiments and establish several new results : ( 1 ) we show that the switching can be performed on time scales much faster than the adiabatic time scale for eit , even in the strong probe regime . we also study the behavior of the system changes when this time scale is faster than the excited state lifetime . ( 2 ) stopped light pulses write their phase and amplitude information onto spatially dependent atomic wavefunctions , resulting in coherent two - component bec dynamics during long storage times . we investigate examples relevant to rb-87 experimental parameters and see a variety of novel dynamics occur , including interference fringes , gentle breathing excitations , and two - component solitons , depending on the relative scattering lengths of the atomic states used and the probe to coupling intensity ratio . we find the dynamics when the levels @xmath2 and @xmath3 are used could be well suited to designing controlled processing of the information input on the probe . ( 3 ) switching the coupling field on after the dynamics writes the evolved bec wavefunctions density and phase features onto a revived probe pulse , which then propagates out . we establish equations linking the bec wavefunction to the resulting output probe pulses in both the strong and weak probe regimes . we then identify sources of deviations from these equations due to absorption and distortion of the pulses . these deviations result in imperfect fidelity of the information transfer from the atoms to the light fields and we calculate this fidelity for gaussian shaped features in the bec wavefunctions . in the weak probe case , we find the fidelity is affected both by absorption of very small length scale features and absorption of features occupying regions near the condensate edge . we discuss how to optimize the fidelity using these considerations . in the strong probe case , we find that when the oscillator strengths for the two transitions are equal the fidelity is _ not _ strongly sensitive to the probe strength , while when they are unequal the fidelity is worse for stronger probes . applications to distant communication between becs , squeezed light generation and quantum information are anticipated .
since the introduction of the burger s equation by bateman@xcite , many authors have used variety of numerical methods in attempting to solve the burger s equation . various forms of the finite element methods are constructed to compute the burger s equation numericallyad , ad1,nu2,sku , ad5,ofem3,bspgal , qq , zi . the spline collocation procedures are also presented for getting solutions of the burger s equationrt , er , ras , cbspcol , irk , ram , qbspcol2,qbspcol , dag , quincol . the spline functions are wished to be accompanied to the numerical method to solve the differential equations since the resulting matrix system is always diagonal and can be solved easily and approximate solutions having the accuracy of the degree less than the degree of the spline functions can be set up . high order continuous differentiable approximate solutions can be produced for the differential equations of higher order . the numerical procedure for nonlinear evolution equations based on the the b - spline collocation method have been increasingly applied to various fields of science . however application of the ctb collocation method to non linear evolution problems are a few in comparison with the method the collocation based on the b - spline functions . the numerical methods for solving a type of ordinary differential equations with quadratic and cubic ctb are given by a. nikolis in the papersg1,g2 . the linear two - point boundary value problems of order two are solved using cubic ctb interpolation method @xcite . the another numerical method employed the cubic ctb are set up to solve a class of linear two - point singular boundary value problems in the study@xcite . very recently a collocation finite difference scheme based on new cubic ctb is developed for the numerical solution of a one - dimensional hyperbolic equation ( wave equation ) with non - local conservation condition@xcite . a new two - time level implicit technique based on the cubic ctb is proposed for the approximate solution of the nonclassical diffusion problem with nonlocal boundary condition in the study@xcite . some researches have established types of the b - spline finite element approaches for solving the burger s equation but not with ctb as far as we know the literature . in this paper , ctb are used to establish a collocation method and then the suggested numerical method is applied to find the numerical solutions of burger s equation . it is also well - known that this problem arise in many branch of the science and the development of the numerical methods for the burger s equation have been attracted for finding the steep front solutions . the use of the lower viscosity in the burger s equation cause the appearance of the steep front and shock wave solutions . that makes difficulty in modelling solutions of the burger s equation when solved numerically . so many authors have developed various kinds of numerical scheme in computing the equation effectively for small values of the viscosity . we consider the burger s equation @xmath0 with appropriate initial conditions and the boundary conditions:@xmath1 @xmath2 , @xmath3 where subscripts @xmath4 and @xmath5 denote differentiation , @xmath6 and @xmath7 is the reynolds number characterizing the strength of viscosity . @xmath8 @xmath9 are the constants and @xmath10 is a sufficiently differentiable unknown function and @xmath11 is a bounded function . the implementation of the proposed scheme is given in the second section . two classical text problems are dealt with to show the robustness of the scheme . consider a uniform partition of the problem domain @xmath12 $ ] at the knots @xmath13 with mesh spacing @xmath14 on this partition together with additional knots @xmath15 outside the problem domain , @xmath16@xmath17 can be defined as @xmath18 $ \\ $ \omega ( x_{i-2})(\omega ( x_{i-2})\phi ( x_{i})+\phi ( x_{i+1})\omega ( x_{i-1}))+\phi ( x_{i+2})\omega ^{2}(x_{i-1}),$ & $ x\in \left [ x_{i-1},x_{i}% \right ] $ \\ $ \omega ( x_{i-2})\phi ^{2}(x_{i+1})+\phi ( x_{i+2})(\omega ( x_{i-1})\phi ( x_{i+1})+\phi ( x_{i+2})\omega ( x_{i})),$ & $ x\in \left [ x_{i},x_{i+1}\right ] $ \\ $ \phi ^{3}(x_{i+2}),$ & $ x\in \left [ x_{i+1},x_{i+2}\right ] $ \\ \multicolumn{1}{c } { } & \multicolumn{1}{c}{$0,\text{otherwise}$}% \end{tabular}% \right . \label{r2}\ ] ] where @xmath19 @xmath20 are twice continuously differentiable piecewise trigonometric cubic b - spline on the interval@xmath21 $ ] . the iterative formula @xmath22 gives the cubic b - spline trigonometric functions starting with the ctb - splines of order @xmath23 @xmath24 each @xmath20 is twice continuously differentiable and the values of @xmath25 and @xmath26 at the knots @xmath27 s can be computed from eq.([r3 ] ) as @xmath28 } $ \\ \hline $ x_{i}$ & $ \frac{2}{1 + 2\cos ( h)}$ & 0 & $ \frac{-3\cot ^{2}(\frac{3h}{2})}{% 2 + 4\cos ( h)}$ \\ \hline $ x_{i+1}$ & $ \sin ^{2}(\frac{h}{2})\csc \left ( h\right ) \csc ( \frac{3h}{2})$ & -$\frac{3}{4}\csc ( \frac{3h}{2})$ & $ \frac{3(1 + 3\cos ( h))\csc ^{2}(\frac{h% } { 2})}{16\left [ 2\cos ( \frac{h}{2})+\cos ( \frac{3h}{2})\right ] } $ \\ \hline $ x_{i+2}$ & 0 & 0 & 0 \\ \hline \end{tabular}% \end{tabular}%\ ] ] @xmath20 , @xmath29 are a basis for the trigonometric spline space . an approximate solution @xmath30 to the unknown @xmath31 is written in terms of the expansion of the ctb as @xmath32 where @xmath33 are time dependent parameters to be determined from the collocation points @xmath13 and the boundary and initial conditions . the nodal values @xmath31 and its first and second derivatives at the knots can be found from the ( [ r4 ] ) as @xmath34@xmath35 the time derivative and space derivatives can be approximated by using the standard finite difference formula and the crank nicolson scheme respectively to have the time - integrated burger s equation : @xmath36where @xmath37 is the solution of the equation at the @xmath38th time level . here @xmath39 @xmath40 , and @xmath41 is the time step , superscripts denote @xmath42 th time level , @xmath43 the nonlinear term @xmath44 in eq . ( [ r6 ] ) is linearized by using the following form @xcite:@xmath45so linearized time - integrated burger equation have the following form:@xmath46substitution [ r4 ] into [ r8 ] and evaluation resulting equation at knots leads to the fully - discretized equation : @xmath47 where@xmath48 the system consist of @xmath49 linear equation in @xmath50 unknown parameters @xmath51 . the above system can be made solvable by elimination the time parameters @xmath52with help of the boundary conditions @xmath53 when written as @xmath54a variant of thomas algorithm is used to solve the system . initial parameters @xmath55 must be found to start the iteration process . to do so , initial condition and boundary values of derivative of initial conditions gives the following equation 1 . @xmath56 @xmath57 @xmath58 2 . @xmath59 3 . @xmath60 the above system yields an ( @xmath50)@xmath61(@xmath50 ) matrix system , which can be solved by use of the thomas algorithm . once the initial parameters @xmath62 has been obtained from the initial and boundary conditions , the recurrence relation gives time evolution of vector @xmath63from the time evolution of the approximate solution @xmath64 can be computed via the equation ( [ r6 ] ) . * ( a ) * analytical solution of the burger s equation with the problem sine wave initial condition @xmath65 and boundary conditions @xmath66 can be expressed as an infinite series @xcite where @xmath68 are the modified bessel functions . this problem gives the decay of sinusoidal disturbance . the convergence of the solution r11 is slow for small values of @xmath69so that the numerical solutions of the burger s equation are looked for . using the parameters @xmath70 @xmath71 graphical solutions at different times are depicted in the figs 1 - 4 the amplitude of the solution decays as time pass , seen in fig 1 - 2 clearly and the sharpness through the right boundary develops when the smaller viscosities are used . he same incidents also exist for studies given in the paper @xcite the results of proposed numerical methods are compared with the cubic b - spline collocation , cubic b - spline galerkin . galerkin procedure are seen to produce slightly same results with the ctb collocation method . our advantage is that the cost of the ctb procedure is less than the galerkin methods given in the tables 1 - 3 . @xmath74@xmath75@xmath76 * ( b ) * well - known other solution of the burger s equation is@xmath77where @xmath78 @xmath79 and @xmath80 are arbitrary constants . initial conditions are@xmath81 @xmath82 or @xmath83 @xmath84 for @xmath85 this form of the solution is known as the travelling wave equation and respresent the propogation of the wave front through the right . parameter @xmath69 determine the sharpness of the solution the initila solutions are taken from the analytical solution when @xmath86 . the program is run for the parameters @xmath87 @xmath88 @xmath89 and @xmath90 . solutions at some space values @xmath4 are presented in table 5 and compared with those obtained in the studies @xcite using cubic b - spline collocation , quadratic / cubic b - spline galerkin methods . solution behaviours are illustrated in fig 5 - 6 for the coefficient @xmath91 and @xmath92 at times @xmath93 with smaller @xmath94the sharp front is formed and propogates to right during run of the program . graphical presentation of the absolute errors at time @xmath95 is drawn i m figs 7 - 8@xmath96 @xmath98.}% \end{array}% \begin{tabular}{l } \frame{itbpf}{2.7224in}{2.3065in}{0in}{}{}{fig6.bmp}{\special{language " scientific word";type " graphic";maintain - aspect - ratio true;display " usedef";valid_file " f";width 2.7224in;height 2.3065in;depth 0in;original - width 2.6801in;original - height 2.2667in;cropleft " 0";croptop " 1";cropright " 1";cropbottom " 0";filename ' fig6.bmp';file-properties " xnpeu " ; } } \\ figure 6 : solutions at different times for \\ $ \lambda = 0.005,$ $ h=1/36,$ $ \delta t=0.001,\text { } x\in \lbrack 0,1].$% \end{tabular}%\ ] ] the collocation methods with trigonometric b - spline functions is made up to find solutions the burger s equation . we have hown that methods is capable of producing solutions of the burgers equation fairly . the method can be used as an alternative to the methods accompanied b - spline functions . nur nadiah abd hamid , ahmad abd . majid , and ahmad izani md . ismail , cubic trigonometric b - spline applied to linear two - point boundary value problems of order , world academy of science , engineering and technology 70 798 - 803 , 2010 . i. da , a. canivar and a. ahin , taylor galerkin and taylor - collocation methods for the numerical solutions of burgers equation using b - splines , commun nonlinear sci numer simulat vol.16 , pp . 26962708,2011 . m. abbas , a. a. majid , a. i. m ismail and a. rashid , the application of the cubic trigonometric b - spline to the numerical solution of the hyperbolic problems , applied mathematica and computation , vol 239 , 74 - 88 , 2014 .	the cubic trigonometric b - spline(ctb ) functions are used to set up the collocation method for finding solutions of the burgers equation . the effect of the ctb in the collocation method is sought by studying two text problems . the burgers equation is fully - discretized using the crank - nicholson method for the time discretizion and ctb function for discretizion of spatial variable . numerical examples are performed to show the convenience of the method for solutions of burgers equation * keywords : * collocation methods , cubic trigonometric b - spline , burgers equation
the dirac equation , together with minimal coupling to an external electromagnetic field , predicts , for an elementary charged lepton @xmath1 of spin @xmath2 , mass @xmath3 , and charge @xmath4 , a pauli - type coupling @xmath5 of the spin to the magnetic field , where the magnetic moment , proportional to the particle s spin , is given in units of the bohr magneton , by @xmath6 with @xmath7 . quantum corrections arising from loops will bring in new contributions and shift the gyromagnetic factor @xmath8 away from this value . given a sufficient level of precision , all degrees of freedom , known or unknown ( i.e. light or heavy ) , will eventually contribute in a visible way . the existence of an _ anomalous _ magnetic moment @xmath9 is thus an indirect probe of the existence and properties of all degrees of freedom , within , but also beyond , the standard model . it actually turns out that , on the experimental level , the anomalous magnetic moments of the electron ( @xmath10 ) and of the muon ( @xmath0 ) are among the most precisely measured low - energy observables in particle physics . in the case of the muon , the very precisely known experimental value @xcite @xmath11 , \ ] ] is dominated by the results obtained by the bnl - e821 experiment @xcite . the situation is even more impressive in the case of the electron . the value @xmath12\ ] ] follows from the measurement , at a relative precision of @xmath13ppt , of the gyromagnetic factor @xmath14 @xcite . the high level of precision achieved by these experiments is certainly challenging for theory and theoreticians . the question naturally arises , whether theoretical evaluations are able to reach the same level of accuracy , and whether the standard model accounts for these values . most exciting , of course , is the possibility that eventually a significant discrepancy ( with reliable theoretical and experimental uncertainties ) remains , thus signaling the existence of degrees of freedom beyond the standard model . this review summarizes the present status of the theoretical evaluations of @xmath0 and of @xmath10 . in order to reach an accuracy comparable to the experimental results , it is necessary to consider contributions from electromagnetic interactions ( section 2 ) , from weak interactions ( section 3 ) , and from strong interactions ( section 4 ) . the latter , although larger than the corrections from the standard model weak interactions , are particularly challenging , since they heavily involve the non - perturbative regime . this is why they have been kept for the end . this review will end with section 5 , which includes a summary and some conclusions , and presents perspectives for the future . the interactions involving only photons and charged leptons can be treated perturbatively , .the coefficients @xmath15 , of the perturbative qed expansion for the anomalous magnetic moment of the electron ( left ) and of the muon ( right ) . the values are taken from @xcite and @xcite . [ cols="^,^,^",options="header " , ] [ table_1 ] @xmath16 the challenge comes , however , from the high orders in the perturbative expansion that one needs to consider in order to reach a level of precision comparable to the experimental one . in the case of the electron , the experimenal uncertainty @xmath17 , together with @xmath18 , indicates that one needs to compute at least five loop contributions . the one - loop coefficient @xmath19 was obtained by j. schwinger @xcite long ago . the two - loop and three - loop coefficients are also known analytically . for surveys of these calculations , and references to the original works see @xcite . the higher order coefficients involve 891 diagrams at four loops , and 12672 diagrams at five loops . they have been computed through systematic numerical evaluations of the multidimensional integrals over the corresponding feynman parameters . complete results have now been published @xcite . the results of these calculations are displayed in table [ table_1 ] , and we merely make a few comments . + * i ) * the accuracy of the coefficients @xmath20 and @xmath21 , which are known analytically , is actually limited by the uncertainties on the experimental values @xcite of the mass ratios @xmath22 and , to a lesser extent , @xmath23 and @xmath24 . in the electron case , these uncertainties would only affect the digits beyond those shown , and are not relevant given the present size of @xmath25 . the uncertainties on the four and five - loop coefficients come , in both cases , from the numerical integration procedure . + * ii ) * in the muon case , the experimental accuracy is somewhat lower , @xmath26 , and one might think that even the fourth order is not needed . but this is without reckoning with the structure of the coefficients , which , in the muon case , are logarithmically enhanced by the presence of the much lighter electron in the loops . starting at three loops , one encounters terms involving @xmath27 ! this explains why the coefficients @xmath28 in table [ table_1 ] are typically larger than in the electron case . despite the size of @xmath29 , @xmath30 , so that this contribution remains marginal in view of the present experimental error . but it will also have to be considered as the latter improves ( see section 5 ) . + * iii ) * some subsets of diagrams contributing to @xmath31 and @xmath32 are also known analytically @xcite , in terms of expansions in powers of the mass ratios . these analytical results provide useful and welcome checks of the numerical evaluations . + * iv ) * finally , notice that @xmath33 , where @xmath34 denotes the uncertainty on @xmath28 , is below the experimental precision @xmath26 for @xmath35 , so that in practice @xmath36 is , for the time being , free of any theoretical error . at this stage , using the latest high - precision measurement of the fine - structure constant @xcite @xmath37 , \ ] ] we have @xmath38 and @xmath39 the first two errors come from the contributions at orders @xmath40 and @xmath41 , respectively , and the third one comes from the experimental uncertainty on @xmath42 . in the case of @xmath36 there is an additional error from the uncertainties on the mass ratios . it then follows that @xmath43 in the case of the electron , @xmath10 is well described by qed , within the uncertainties , which are dominated by the present uncertainty on the determination of @xmath42 . in the case of the muon , the discrepancy with the experimental value is substantial , so that the difference is to be ascribed to the contributions of the weak and of the strong interactions . + notice that one can use the measurement of @xmath10 in order to reduce ( by a factor 2.5 ) the error on @xmath42 and obtain a determination of the fine - structure constant at the @xmath44ppb level @xcite . but this means that @xmath10 then no longer provides a possible test of the standard model ( more on this at the end of section 5 ) . the one - loop contributions due to the weak interaction sector of the standard model have been computed @xcite more than forty years ago . they read @xmath45 , \ ] ] and correspond numerically to @xmath46 . two - loop electroweak corrections are also available @xcite , and reduce the above value by about 20% . a recent reevaluation @xcite gives , for the sum of the one- and two - loop weak contributions , @xmath47 after taking these contributions into account , we now have @xmath48 and only the strong interactions remain in order to close this gap . the strong - interaction contributions mainly come from the low - energy , light - quark , sector , so that a perturbative approach is no longer adapted . it is both useful and customary to distinguish three types of corrections involving the hadronic sector , as shown in fig . [ fig-1 ] . the diagram on the left corresponds to hadronic contributions to the photon vacuum polarization function . the diagram in the middle is the so - called hadronic light - by - light contribution . finally , there is also a hadronic contribution to the two - loop weak corrections discussed in the previous section , arising from the exchange of a virtual photon and a virtual neutral weak gauge boson between the external lepton line and the hadronic blob ( right diagram ) . in comparison to the other two - loop weak corrections , this last correction is small . it was already included in the values given in the preceding section , and will not be discussed further here . for details , i refer the reader to the literature quoted in the preceding section . let us start by examining the case of hadronic vacuum polarization . this contribution can be expressed as @xcite @xmath49 with @xmath50 and @xmath51 represents the @xmath52-ratio of the cross section of @xmath53 . the advantage of this representation is threefold . first it tells us that the contribution is positive ( @xmath54 ) . second , that it is dominated by the low - energy domain , since @xmath55 for @xmath56 large . and finally , it is directly related to an experimental quantity , which allows @xmath57 to be evaluated using available data on @xmath53 . the results from the two latest published evaluations @xcite give comparable values , @xmath58 , \nonumber\\ a_{\mu}^{\rm hvp - lo } & = & + 694.9(4.3 ) \cdot 10^{-10 } \quad [ 30 ] .\end{aligned}\ ] ] the corresponding value for the electron reads @xcite @xmath59 and is larger than the present uncertainty on @xmath60 . there are also contributions of the same type at order @xmath61 , obtained upon inserting a second ( hadronic or leptonic ) vacuum polarization in one of the two virtual photons of the leftmost diagram in fig . [ fig-1 ] , or upon inserting a hadronic vacuum polarization in one of the photon lines of the two - loop qed contributions . these contributions can be written in a similar form as in eq . ( [ hvp - lo ] ) , but involve now a ( known ) kernel different from @xmath62 @xcite . these next - to - leading hadronic vacuum polarization corrections , evaluated with the same @xmath53 data , amount to @xcite @xmath63 for the muon , and @xcite @xmath64 for the electron . recently , the next - next - to - leading hadronic vacuum polarization corrections were also evaluated @xcite , with the result @xmath65 summing these three contributions gives @xmath66 with the input from @xcite , or @xmath67 with the input from @xcite , and the difference @xmath68 equals @xmath69 or @xmath70 , respectively . we have kept for the end the hadronic contribution to virtual light - by - light scattering , that remains the most challenging one at present . here , we have no direct link to an experimental observable , and even the overall sign can not be fixed from the outset . one has to resort to a certain amount of model dependence in order to describe , in the non - perturbative regime , the hadronic four - point function ( see fig . [ fig-1 ] ) that is involved . only two calculations have tried to give a complete description of this four - point function from the point of view of its contribution to @xmath0 . after correcting for the sign @xcite of the dominant contribution coming from the reducible exchange of a single pseudoscalar meson between pairs of ( virtual ) photons , their results @xcite read @xmath71 , \nonumber\\ a_\mu^{\rm hlxl } & = & + 8.96(1.54 ) \cdot 10^{-10 } \quad [ 37 ] , \end{aligned}\ ] ] and show good agreement . but if one looks into the details @xcite of these calculations , there are sometimes sizeable differences , either between contributions common to the two models , or in the contributions that are included or not . the limit of a large number of colours @xmath72 @xcite can be useful in order to organize the various contributions @xcite . at leading order , one has to consider contributions from reducible single meson exchanges ( pseudoscalars , axial vectors , tensors ) . of these , those of higher masses will have their contributions suppressed , so that the contribution from single @xmath73 exchanges dominates . this is indeed the case in the models that have been considered in @xcite . in the framework of the effective low - energy theory , the leading - order contribution can be worked out exactly @xcite , @xmath74 . \label{large - nc}\ ] ] in this expression , @xmath75 is the pion - decay constant ( @xmath76 in the large-@xmath72 limit ) , and @xmath77 , the mass of the @xmath78 meson , is a typical hadronic mass scale where the effective theory ceases to be applicable . the constant @xmath79 in front of the subleading contribution is related @xcite to a low - energy constant @xmath80 @xcite that also appears in the decay @xmath81 , and can thus be extracted @xcite from data . the last constant , @xmath82 , is not fixed by any requirement . in ( [ large - nc ] ) , the sign of the leading contribution comes out as positive . since it results from a model - independent approach , this result has a general validity . in model calculations , the scale @xmath77 enters through the form factors that describe the @xmath83 vertex , and which are necessary in order to make the contribution from @xmath73 exchange finite ( taking the constant vertex that follows from the wess - zumino term would lead to a divergent contribution ) . as the uv - regulator @xmath77 is sent to infinity , one should recover the behaviour ( [ large - nc ] ) . this has been checked in a variety of models @xcite . although the pion - exchange practically gives the final result in @xcite , this partly also results from the cancellation among other contributions , which , taken individually , can be sizeable . some of these contributions are regularly updated or reevaluated ( for a recent example see @xcite , and the contributions in @xcite ) . there are also contributions coming from short distances , and an important issue is the matching between contributions arising from different scales and regimes . this is certainly a feature that one should improve . for instance , although some short - distance constraints are implemented @xcite , an extensive study of the four - point function ( cf . fig . [ fig-1 ] ) from this point of view has not been done so far . eventually , the total error on @xmath84 will not only reflect our ability to reduce the errors on individual contributions , but must also reflect our confidence in the way we are able to put them together , while respecting the known properties of qcd at short and at long distances . in order to summarize the present situation , one may quote the best estimate " from @xcite @xmath85 whereas a more conservative estimate @xcite gives @xmath86 in the case of the electron , this contribution is much smaller , and one finds @xcite @xmath87 the anomalous magnetic moments of the electron and of the muon are among the most precisely measured low - energy observables of the standard model . the standard - model value obtained for @xmath10 agrees with the experimental measurement @xmath88 however , at present the value obtained for @xmath0 wihin the standard model misses the experimental one . the difference @xmath89 ranges from @xmath90 , taking the inputs from @xcite for hvp and from @xcite for hlxl , @xmath91 , \ ] ] to @xmath92 , @xmath93 , \ ] ] when the inputs for hvp and hlxl are taken from @xcite and @xcite , respectively . this discrepancy has been with us for quite some time , and it is not obvious to find a straightforward explanation for it . it is almost twice as large as the correction from the weak interactions ( section 3 ) , and the evaluation of the hadronic light - by - light correction would have to be off by a factor of 2 to 3 to explain it . could it come from higher order qed effects ? after all , the coefficients @xmath28 in table [ table_1 ] display a dramatic increase with @xmath94 . an estimate of the contribution at twelfth order @xmath95 that shows the enhancement mechanism mentioned previously , based on the electron ligh - by - light loop @xmath96 , corrected by three electron loops , inserted in all possible ways in the three photon lines that are internal to the diagram , gives @xcite @xmath97 ^ 3 \cdot 10 \nonumber\\ \!\!\!\ ! \!\!\!\!\!\ ! \!\ ! & \sim & \!\ ! 0.6 \cdot 10 ^ 4 .\end{aligned}\ ] ] the corresponding contribution to @xmath0 , @xmath98 is however way too small to explain even part of the discrepancy . could higher order qcd effects be at work ? besides @xmath99 @xcite already included , higher - order corrections to @xmath100 have also been considered in @xcite . the estimated value @xmath101 again fails to match the size of the discrepancy . finally , are we already seeing manifestations of bsm degrees of freedom ? there are many proposals , see e.g. @xcite , but the present situation is clearly inconclusive . + clarification will hopefully come from two new experiments , that are being prepared at fnal @xcite and at j - parc @xcite , with the aim of reducing the experimental uncertainty by a factor of 4 . in order to lead to a conclusive situation , these experimental efforts need to be met by a comparable improvement of the theoretical uncertainty . forthcoming high - precision data from e.g. vepp-2000 , babar , besiii or kloe-2 , should allow for a further reduction of the uncertainty on the hadronic vacuum polarization part , probably below the 0.5% level . hadronic light - by - light will then become the dominant source of theoretical errors . there are proposals to address this issue @xcite , but much remains to be done in order to lead these programs to the desired goal . finally , increasing efforts are being made in order to obtain reliable evaluations of the hadronic contributions to @xmath0 from qcd simulations on the lattice . the success of this enterprise will hinge on the ability to develop appropriate strategies to overcome the limitations set by the lattice . for interesting proposals and recent advances in this direction , see @xcite and references therein , or the contributions in @xcite , as well as @xcite and @xcite . this , admittedly incomplete , list of items and of references should merely convey the feeling that the subject remains very active also at the theoretical level . although the challenge is high , it is reasonable to expect that significant improvements will become available around the time the data from the new experiments will be released . + i would like to close this review with a short remark concerning the anomalous magnetic moment of the electron . although @xmath10 is expected to be less sensitive to new physics than @xmath0 by a about a factor @xmath102 , there are exceptions to this naive scaling , and moreover @xmath10 is measured with a precision that is @xmath103 times better than the experimental precision on @xmath0 . this leads one to consider the possibility @xcite that degrees of freedom beyond the standard model could be first detected in @xmath10 . indeed , improvements on both @xmath25 and on the experimental determination of @xmath42 are within reach @xcite on a timescale comparable to the new generation of @xmath104 experiments at fnal and at j - parc . we are therefore looking forward to very exciting times in high - precision physics with the forthcoming experiments and , hopefully , with the accompanying improvements on the theoretical aspects . i would like to thank s. narison and the organizing committee of the qcd-2014 conference for the invitation to give this presentation . i have benefited from many informative and stimulating discussions with l. lellouch , a. nyffeler , and e. de rafael , and i thank l. lellouch and e. de rafael for their comments on the manuscript . i would also like to express my gratitude to the mainz institute for theoretical physics ( mitp ) for its hospitality and support . this work was supported in part by the ocevu labex ( anr-11-labx-0060 ) and the amidex project ( anr-11-idex-0001 - 02 ) funded by the `` investissements davenir '' french government program managed by the anr . j. schwinger , phys . rev . 73 * , 416l ( 1948 ) . t. kinoshita ( chapter 3 ) , s. laporta and e. remiddi ( chapter 4 ) in _ lepton dipole moments _ , advanced series on directions in high energy physics vol . 20 , b. lee roberts and william j. marciano eds , world scientific co. pte . ltd . ( 2010 ) . t. kinoshita , m. nio , phys . d * 73 * , 053007 ( 2006 ) . t. aoyama et al . d * 78 * , 053005 ( 2008 ) ; phys . d * 78 * , 113006 ( 2008 ) ; phys . d * 81 * , 053009 ( 2010 ) ; phys . d * 82 * , 113004 ( 2010 ) ; phys . d * 83 * , 053002 ( 2011 ) ; phys . rev . d * 83 * , 053003 ( 2011 ) ; phys . d * 84 * , 053003 ( 2011 ) ; phys . d * 85 * , 033007 ( 2012 ) ;	the calculations entering the prediction of the standard model value for the anomalous magnetic moment of the muon @xmath0 are reviewed , and compared to the very accurate experimental measurement . the situation for the electron is discussed in parallel .
type ia supernova ( sn ia ) explosions play a critical role in regulating chemical evolution through the cycling of matter in galaxies . as supernovae ( sne ) are the primary contributors of heavy elements in the universe , observed variations in their rates with redshift provide a diagnostic of metal enrichment over a cosmological timeline . the frequency of these events and the processes involved provide important constraints on theories of stellar evolution . sne ia are thought to originate from the thermonuclear explosion of carbon - oxygen white dwarfs that approach the chandrasekhar mass via accretion of material from a binary companion ( for reviews , see * ? ? ? * ; * ? ? ? this process can result in a significant `` delay time '' between star formation and sn explosion , depending on the nature of the progenitor system @xcite . the sn ia volumetric rate ( ) evolution therefore represents a convolution of the cosmic star - formation history with a delay - time distribution ( dtd ) . as such , measuring the global rate of sn ia events as a function of redshift may be useful for constraining possible dtds and , ultimately , progenitor models the detailed physics of sne ia remains poorly understood , with several possible evolutionary paths ( e.g. * ? ? ? * ; * ? ? ? one complication for rates studies is that many sn surveys at low redshifts are galaxy - targeted , counting discoveries in a select sample of galaxies and converting to a volumetric rate by assuming a galaxy luminosity function . this method can be susceptible to systematic errors if it preferentially samples the bright end of the galaxy luminosity function , biasing toward sne in more massive , or brighter , galaxies ( see , e.g. , * ? ? ? since many sn ia properties are correlated with their hosts , the recovered rates may then not be representative of all types of sne ia . a second type of sn survey involves making repeat observations of pre - defined fields in a `` rolling search '' , to find and follow sne in specific volumes of sky over a period of time . such surveys minimize the influence of host bias , but still suffer from malmquist bias and other selection effects . it is reasonably straight forward although often computationally expensive to compensate for the observational biases within rolling searches . the advent of these wide - field rolling surveys has significantly enhanced sn statistics at cosmological distances . the supernova legacy survey ( snls ) in particular has contributed a large sample of type ia sne out to redshifts of @xmath16 @xcite . although its primary goal is to assemble a sample of sne ia to constrain the cosmological parameters ( e.g. * ? ? ? * ; * ? ? ? * ) , the snls is also ideal for studies of sn rates @xcite . the snls is a rolling high - redshift search , with repeat multi - color imaging in four target fields over five years and as such has consistent and well - defined survey characteristics , along with significant follow - up spectroscopy . however , due to the selection effects ( including incomplete spectroscopic follow - up ) and other systematic errors , such as contamination and photometric redshift errors , present in any sn survey , a detailed understanding of internal biases is necessary for accurate rate calculations . in the past decade , volumetric sn ia rates have been measured to varying degrees of accuracy out to redshifts of @xmath17 ( fig . [ fig : rates_lit ] ) . @xcite compute the sn ia rate in the local universe ( @xmath18 ) from a combined visual and photographic sample of @xmath19 galaxies , yet their ability to distinguish core - collapse sne from type ia sne was severely limited . more recent work by @xcite using @xmath20 sne ia from the lick observatory supernova search ( loss ; * ? ? ? * ) has made significant improvements in the statistics over previous studies on local sne ia . the rates published by @xcite include data from 516 sne ia at redshifts @xmath21 from the sdss - ii supernova survey ( sdss - sn ) , with roughly half of these confirmed through spectroscopy . at intermediate redshifts , rate measurements are provided by ( * ? ? ? * sne from the supernova cosmology project in the range @xmath22 ) , ( * ? ? ? * sne within @xmath23 ) , and ( * ? ? ? * @xmath24 sne from the ifa deep survey , 23 of which have spectra ) . @xcite used a spectroscopic sample of 58 sne ia from the first two years of snls to measure a cumulative volumetric rate in the redshift range @xmath25 . sn ia rates out to @xmath26 are presented by @xcite using 25 sne ia ( 19 with spectra ) from _ hubble space telescope ( hst ) _ observations of the great observatories origins deep survey ( goods ) fields . these data were reanalyzed by @xcite using a bayesian identification algorithm , and the _ hst _ sample updated by @xcite extending the 2004 sample to 56 sne . ground - based measurements from the subaru deep field have also been made by @xcite using 22 sne ia , updated by @xcite with 150 events . the general trend of fig . [ fig : rates_lit ] reveals that the rates typically increase from @xmath27 to @xmath28 . there is a wide spread in the existing rate measurements , particularly in the range @xmath29 . at higher redshifts , data from the goods collaboration provide some apparent evidence for a turnover in the sn ia rates . in particular , @xcite report a decline in sn ia rates beyond @xmath30 . if present , this decline might point to a larger characteristic delay time between star formation and sn explosion ( see also * ? ? ? however , another independent analysis of the _ hst _ goods data finds rates that are offset , with measurements by @xcite consistently lower than those of @xcite . @xcite argue that their results do not distinguish between a flat or peaked rate evolution . ground - based data in this range @xcite , while consistent with the _ hst_-based results , show no obvious evidence for a decline above @xmath31 . in this paper we use four years of data from the snls sample to investigate the evolution of sn ia rates with redshift out to @xmath32 . the sample presented comprises @xmath33 photometrically identified sne ia from snls detected with the real - time analysis pipeline @xcite . one third of these have been typed spectroscopically , and one half of the @xmath34 have a spectroscopic redshift ( sometimes from ancillary redshift surveys in the snls fields ) . no other data set currently provides such a well - observed and homogeneous sample over this range in redshift . additionally , rigorous computation of the survey detection efficiencies and enhancements in photometric classification techniques are incorporated into the new snls rate measurements . monte carlo simulations of artificial sne ia with a range of intrinsic parameters are performed on all of the detection images used in the snls real - time discovery @xcite ; these provide an exhaustive collection of recovery statistics , thereby helping to minimize the effects of systematic errors in the rate measurements . the snls sne ia can be used to examine the relationship between the and redshift , given some model of the sn ia dtd . the size of the snls sample also permits a division of the sne ia by light - curve width ( in particular the stretch ; see * ? ? ? * ) , allowing a search for differences in the volumetric rate evolution expected by any changing demographic in the sn ia population . brighter , more slowly - declining ( i.e. , higher stretch ) sne ia are more frequently found in star - forming spirals , whereas fainter , faster - declining sne ia tend to occur in older stellar populations with little or no star formation @xcite . if the delay time for the formation of the lowest - stretch sne ia is sufficiently long ( i.e. , their progenitors are low - mass stars @xmath35 gyr old ) , these sne ia will not occur at high redshifts @xcite . the behavior of the high-@xmath4 rates can reveal the properties of the progenitor systems . the organization of this paper is as follows : an overview of the rate calculation is provided in [ sec : ratecalc ] . the snls data set , along with the light - curve fitting and selection cuts used to define the photometric sample , is introduced in [ sec : snls ] . sn ia detection efficiencies and the rate measurements are presented in [ sec : effs ] and [ sec : snlsrates ] , respectively . several models of the sn ia dtd are then fit to the rate evolution in [ sec : dtds ] , and the results discussed . finally , the stretch dependence of the rate evolution is investigated in [ sec : stretchdep ] . we adopt a flat cosmology with ( @xmath36,@xmath37)=(0.27,0.73 ) and a hubble constant of @xmath38 . the volumetric sn ia rate in a redshift ( @xmath4 ) bin @xmath39 is calculated by summing the inverse of the detection efficiencies , @xmath40 , for each of the @xmath41 sne ia in that bin , and dividing by the co - moving volume ( @xmath42 ) appropriate for that bin @xmath43 the factor @xmath44 corrects for time dilation ( i.e. , it converts to a rest - frame rate ) , @xmath45 is the effective search duration in years , and the volume @xmath42 is given by @xmath46 ^ 3 \mathrm{mpc}^3\ ] ] where @xmath47 is the area of a search field in deg@xmath48 and in this equation @xmath49 is the speed of light , and @xmath50 , @xmath36 and @xmath37 are the cosmological parameters , and we assume a flat universe . @xmath40 is a recovery statistic which describes how each sn ia event should be weighted relative to the whole population ; @xmath51 gives the fraction of similar sne ia that exploded during the search interval but that were not detected , for example due to sampling or search inefficiencies . @xmath40 is a function of the sn stretch @xmath52 , a dimensionless number that expands or contracts a template light curve defined as @xmath53 to match a given sn event , the sn color @xmath49 , defined as the rest - frame @xmath54 color at maximum light in the rest - frame @xmath55-band , and the sn @xmath4 . the @xmath40 are evaluated separately for each year and field of the survey , and are further multiplied by the sampling time available for finding each object ( @xmath45 ) to convert to a `` per year '' rate . typically these are 5 months for the snls , though this is dependent on the field and year of the survey . thus , in practice , eqn . ( [ eq : rate ] ) is evaluated for each search field and year that the survey operates . this `` efficiency '' method is particularly suited for use with monte - carlo simulations of a large , well - controlled survey such as snls . its disadvantage is that it is not straight forward to correct for the likely presence of sne that are not represented ( in @xmath56 parameter space ) among the @xmath41 in eqn . ( [ eq : rate ] ) ( for example , very faint or very red sne ia ) without resorting to assuming a luminosity function to give the relative fractions of sne with different properties . in particular , we are not sensitive to , and nor do we correct for , spectroscopically peculiar sne ia in the sn2002cx class ( e.g. , * ? ? ? * ) , and similar events such as sn2008 ha ( e.g. , * ? ? ? * ) , super - chandrasekhar events ( e.g. , * ? ? ? * ) , and other extremely rare oddballs ( e.g. , * ? ? ? we also exclude sub - luminous sne ia ( here defined as @xmath57 , a definition that would include sn1991bg - like events ) but note that these are studied in considerable detail for the snls sample in our companion paper , @xcite . thus , we are presenting a measurement of the rates of `` normal '' , low to moderate extinction sne ia ( explicitly , @xmath58 ) , restricting ourselves to the bulk of the sn ia population that we can accurately model . we allow for these incompletenesses when comparing to other measurements of the sn ia rate in [ sec : dtds ] , which do include some of these classes of sne ia . the photometric sample begins with the set of all possible detections , to which we apply a series of conservative cuts to remove interlopers . the snls sample and the culling process are described next in [ sec : snls ] . to each resulting sn ia must then be applied the corresponding @xmath40 ; these are calculated using a detailed set of monte carlo simulations on the snls images , a procedure described in [ sec : effs ] . the rate results and the measurement of their associated errors are presented afterwards in [ sec : snlsrates ] . in this section , we describe the snls search and the sn ia sample that we will subsequently use for our rate analysis . the snls is a rolling sn search that repeatedly targeted four @xmath59 fields ( named d14 ) in four filters ( @xmath60 ) using the megacam camera @xcite on the 3.6 m canada france hawaii telescope ( cfht ) . the snls benefited from a multi - year commitment of observing time as part of the cfht legacy survey . queued - service observations were typically spaced @xmath61 days apart during dark / grey time , yielding @xmath62 epochs on the sky per lunation . key elements of the snls are its consistent and well - defined survey characteristics , and the high - quality follow - up spectroscopy from 8m - class telescopes such as gemini @xcite , the eso very large telescope ( vlt ; * ? ? ? * ) , and keck @xcite . due to the finite amount of follow - up observing time available , not all of the sn ia candidates found by snls were allocated for spectroscopic follow - up ( for a description of follow - up prioritization , see * ? ? ? * ; * ? ? ? the availability of well - sampled light curves and color information from the snls nonetheless allow us to perform photometric identification and redshift measurements , even in the absence of spectroscopic data . crrcc d1 & 02:26:00.00 & -04:30:00.0 & 0.8822 & 4 + d2 & 10:00:28.60 & + 02:12:21.0 & 0.9005 & 4 + d3 & 14:19:29.01 & + 52:40:41.0 & 0.8946 & 4 + d4 & 22:15:31.67 & -17:44:05.7 & 0.8802 & 4 + to identify the photometric sn ia sample , we begin with all variable object detections in the snls real - time pipeline @xcite . other articles will describe a complementary effort to measure the rates with a re - analysis of all of the snls imaging data ( e.g. , * ? ? ? we use snls data up to and including the fifth year of d3 observing in june , 2007 filter on megacam was damaged during a malfunction of the filter jukebox . candidates discovered after this period were observed with a new @xmath63 filter , requiring new calibrations for subsequent images , and were thus not included in the present study . ] . the first ( 2003 ) season of d3 is omitted in this analysis ; this was a pre - survey phase when the completeness of the sn data differed significantly from the rest of the survey . the remaining detections made during four observing seasons for each of the four deep fields are considered in this analysis . each period of observation on a given field is called a `` field - season '' , with 16 field - seasons in total ( 4 fields observed for 4 seasons ) . the coordinates of the field centers and other information are provided in table [ tab : deepfields ] . we remove all candidates falling within masked regions in the deep stacks . these regions include areas in and around saturated bright stars or galaxies , as well as in the lower signal - to - noise edge regions of the dithered mosaic . the remaining unmasked areas in each field are listed in table [ tab : deepfields ] , and add up to a total of @xmath64 square degrees . galaxy catalogs from these image stacks are used to determine the placement of test objects in the simulations described later in [ sec : effs ] . this cut therefore ensures that the areas being considered in the rates calculation match those used in the detection efficiency measurements . we next fit each event with a light curve fitter to determine its redshift ( where no spectroscopic redshift is available ) and photometric parameters ( @xmath65[sec : light - curve - fitting ] ) . we then remove sn ia candidates with insufficient light curve coverage ( [ sec : culls ] ) . finally , we use the light curve fits to identify and remove core - collapse sne as well as other transients , such as agn and variable stars ( [ sec : removing - non - sne ] ) . each of the remaining sne ia will then correspond to some fraction of the true number of events having similar photometric properties but which were undetected by our survey . this detection efficiency will be determined from the monte carlo simulations presented in [ sec : effs ] . we fit template light curves to the sn ia candidates to identify those that do nt match typical sne ia . flux measurements are made on all of the final `` elixir - preprocessed '' images @xcite . the canadian snls photometric pipeline @xcite was used to measure the sn fluxes , using images processed with the accumulated flat - fields and fringe - maps from each queue run , and aligning photometrically to the tertiary standard stars of @xcite . two light - curve fitting tools were used to help identify the sne ia : estimate_sn @xcite for preliminary identification and for measuring sn ia photometric redshifts , and sifto @xcite for final light - curve fitting to measure the stretch and color of each candidate . the estimate_sn routine is not designed for exact measurement of sn ia parameters , and sifto does not fit for redshift , so we require this two - step process to fully characterize the photometric sample of events . in estimate_sn , the measured fluxes in @xmath60 are fit using sn ia spectral templates from @xcite . the current version of the code includes the addition of priors in stretch , color , and @xmath66mag . these are determined from the distributions measured for the spectroscopic sample . the photometric redshifts ( @xmath67 ) output from this routine are used for candidates with no available spectroscopic redshifts from either the sn or its host . sifto is an empirical light - curve analysis method that works by manipulating a spectral energy distribution ( sed ) model rather than operating in integrated filter space @xcite . sifto does not impose a color model to relate the observed filters during the light - curve fits . the implication of this is that sifto can not easily be modified to fit for redshift , and thus requires a known input @xmath4 . output sn ia fits are parameterized by stretch , date of maximum light , and peak flux in each filter . the stretch measurement provided by sifto is largely invariant to changes in input redshift , as demonstrated in fig . [ fig : sinvar ] . even when the input redshift is off by @xmath68 , the output stretch remains within @xmath69 of its actual value . opacity effects in the sn ejecta are more pronounced in the bluer bands , causing a more rapid decline @xcite ; light curves measured at shorter wavelengths are therefore intrinsically narrower . as a result , if sifto is ( for example ) given an incorrectly small input redshift , it measures the `` wrong '' time dilation but simultaneously samples further towards the blue end of the spectrum where the template is intrinsically narrower . the latter effect partially negates the first , resulting in the same stretch measurement regardless of marginal deviations in input redshift . while this is not similarly true of the derived color or measured fit quality , this stretch invariance is extremely useful for establishing an initial constraint in fitting photometric redshifts with estimate_sn . to @xmath70 . the gray shaded area represents the standard deviation of the measured points about the median @xmath71 . the bottom plot shows the mean output stretch for each sn ia as a function of its known stretch ( at zero redshift offset ) . the error bars for each sn ia in the lower plot represent the full range in stretch values output from sifto as the input redshift is changed.,scaledwidth=80.0% ] spectroscopic redshifts are available for 525 ( 43% ) of the detections remaining after the observational cuts : 420 from snls spectroscopy and the rest from host - galaxy measurements ( including data from deep / deep2 @xcite , vvds @xcite , zcosmos @xcite , and additional snls vlt mos observations ) . the external redshifts are assigned based on a simple ra / dec matching between the snls and the redshift catalogues , with a maximum allowed separation of 1.5 . for the snls mos work , the host was identified following the techinques of @xcite . the known redshifts are then held fixed in the light - curve fits . we also considered the use of galaxy photometric redshifts for the snls fields ( e.g. * ? ? ? * ) . however , though these catalogs have an impressive precision , they tend to be incomplete and untested below a certain galaxy magnitude . sn ia photometric redshifts do not suffer these problems . sn photometric redshifts ( photo-@xmath4 s ) are calculated for the remaining objects using a multi - step procedure . preliminary redshift estimates are obtained using a first round of estimate_sn fits without any constraints on the input parameters . the resulting fit redshifts are then used as input to sifto to measure the stretch for each object . these stretch values are then fixed in a subsequent round of estimate_sn fits to obtain a more robust measurement of the sn redshift constraining at least one input parameter to estimate_sn improves the quality of the light - curve fits . [ fig : zcomp ] shows that the @xmath67 are in good agreement with the spectroscopic redshifts ( @xmath72 ) out to @xmath73 , with a small systematic offset above that . the median precision in @xmath67 for the confirmed sne ia is @xmath74 with @xmath75 . for comparison , @xcite find @xmath76 with a smaller sample and real - time data ( and a previous version of the estimate_sn code ) . in [ sec : errsims ] , we describe how these @xmath67 errors and the systematic offset are incorporated into the rates analysis . each candidate must pass a set of light curve quality checks to be included in the photometric sample of sne ia for the rate calculation . requiring that the sn light curves are well measured ensures that the photometric typing technique is more reliable , and that it is straight forward to correct for the effects of the selection cuts on the rates themselves . therefore , candidates with insufficient light - curve coverage to measure accurate redshift , stretch , and color values from template fits are removed from the detected sample . we define observational criteria in terms of the phase , @xmath77 , of the sn in effective days ( d ) relative to maximum light in the rest - frame @xmath55-band , where @xmath78 and @xmath79 is the observer - frame phase of the sn . the time of maximum light is determined using the light curve fitter sifto @xcite , described in the previous section . each object is required to have a minimum of each of the following : * one observation in each of @xmath63 and @xmath80 between @xmath81d and @xmath82d for early light - curve coverage and color information ; * one observation in @xmath83 between @xmath81d and @xmath84d for additional color information ; * one observation in each of @xmath63 and @xmath80 between @xmath85d and @xmath86d for coverage near peak ; * one observation in either @xmath63 or @xmath80 between @xmath84d and @xmath87d to constrain the later stages of the light curve . these conditions differ slightly from those used by @xcite in their analysis of the first year of snls data . note that no cuts are made on the signal - to - noise ( s / n ) on a particular epoch ; that is , a detection of a candidate on each of the observation epochs is not a requirement . we also neglect the redshift offset seen in fig . [ fig : zcomp ] in calculating the above rest - frame epochs . we estimate that this would shift the effective epochs by only one day in the worst case ( @xmath87d ; a @xmath28 sn ) , and in most cases would be far smaller than this . table [ tab : culls ] provides the numbers of candidates that survive each of the applied cuts . in total , @xmath88 snls detections pass the light curve coverage cuts , @xmath89 of which are spectroscopically confirmed sne ia . ( for consistency , these same objective requirements are also applied to the artificial sne ia used in the monte carlo simulations ( [ sec : effs ] ) , thereby directly incorporating the effects of this cut into the detection efficiencies . ) with these objective criteria satisfied , we can then use light - curve fitting to define a photometric sn ia sample . lcc masking cut & 1538 & 325 + observational cuts & 1210 & 305 + fit quality and @xmath52 cuts & 691 & 286 a set of @xmath90 goodness - of - fit cuts ( here , @xmath90 is the @xmath91 per degree of freedom , @xmath92 ) are applied to all of the sn ia light - curve fits from estimate_sn to help eliminate non - ias from the current sample ( see also * ? ? ? an overall @xmath90 cut along with individual @xmath63 and @xmath80 filter @xmath90 constraints are applied separately for cases both with and without known redshifts . light - curve fit quality limits for the sample with known input redshifts are set to @xmath93 , @xmath94 , and @xmath95 ( here the @xmath92 is omitted for clarity ) . those with fit redshifts are given stricter limits of @xmath96 , @xmath97 , and @xmath98 . the tighter @xmath90 limits for candidates without known redshifts are necessary since core - collapse sne specifically sne ib / c can sometimes achieve better fits to sn ia templates when @xmath4 is permitted to float from the true value . the limits are determined empirically by maximizing the fraction of sne ia remaining in the sample , while also maximizing the number of known non - ias that are removed . note that estimate_sn , unlike sifto , enforces a color relation between the fluxes in different filters , which leads to larger @xmath90 than in sifto fits . in the case of fixed [ floating ] redshifts input to the fit , @xmath99 [ @xmath100 of spectroscopically identified sne ia survive the @xmath90 cuts ( we correct for this slight inefficiency when calculating our final rate numbers ) , while @xmath101 [ @xmath102 of confirmed non - ias remain in the sample . a final round of sifto fits is then performed to determine the output values of stretch and color . the input redshifts to sifto are set to @xmath67 wherever no @xmath72 values are available . one final light - curve fitting cut is then applied on the sample , requiring that the output sifto template fits have @xmath103 . this step removes all but one of the remaining confirmed non - ias but the host has a spectroscopic redshift of @xmath104 . ] when all redshifts are allowed to float , while at the same time maximizing the number of confirmed sne ia passing the cut . no known contaminants remain when all available @xmath72 values are fixed in the fits . the final photometric sn ia ( phot - ia ) sample is restricted to @xmath105 . above this redshift , the rates are found to be too uncertain to include in subsequent analyses . this is a result of low s / n , poor detection efficiency , 100% spectroscopic incompleteness , and the potential for increased contamination from non - ias . only candidates having stretch values within @xmath106 are considered in the present study . this range is characteristic of the snls spectroscopic sample shown by the red histogram in the central plot of fig . [ fig : objpars ] but excludes sub - luminous events such as sn1991bg . these sub - luminous , low - stretch sne ia in the snls sample have been studied in detail by @xcite our stretch limit removes 22 such objects from our sample . extremely red ( @xmath107 ) , and presumably highly extincted , candidates are also removed . this cut eliminates only one event : snls-04d2fm , a faint sn of unknown type at @xmath108 . the final redshift , stretch , and color distributions resulting from the various cuts are shown in fig . [ fig : objpars ] . the phot - ia sample consists of @xmath109 objects , @xmath110 of which have known redshifts . a total of @xmath0 objects in this sample have been spectroscopically confirmed as type ia sne ( table [ tab : culls ] ) . the redshift histogram reveals that the incompleteness of the spectroscopic sample ( in red ) begins to increase beyond @xmath111 , where the rise in the required exposure time makes taking spectra of every candidate too expensive . the effects of incompleteness in the observed snls sample become severe above @xmath112 . the full phot - ia sample has median stretch and color values of @xmath113 and @xmath114 , respectively . the color distribution peaks at a slightly redder value than the estimated typical color of a sne ia of @xmath115 , based on the distribution observed for the spectroscopic snls sample . with the final sn ia sample in hand , we now need to estimate the weight that each of these events contributes in our final rates calculation . these `` detection efficiencies '' depend on many observational factors and will obviously vary with sn ia characteristics . for example , at higher - redshift , the higher stretch sne ia are more likely to be recovered not only because they are brighter , but also because they spend a longer amount of time near maximum light , and are therefore more likely to pass the culls of @xmath65[sec : culls ] . in a rolling search like snls , such effects can be directly accounted for by measuring recovery statistics for a range of simulated input sn ia properties using the actual images ( and their epochs ) observed . this is a brute - force approach , but is a practical way to accurately model a survey such as snls , helping to control potential systematic errors by avoiding assumptions about image quality limitations and data coverage that may bias the rate calculation . uncertainties on search time and detection area are avoided since the actual values are well defined . an exhaustive set of monte carlo simulations were performed for each field - season to determine the recovery fraction as a function of redshift , stretch , and color . full details about these simulations are presented in @xcite . a total of @xmath116 artificial sne ia with a flat redshift distribution were added to galaxies present in the snls fields . each host galaxy was chosen to have a photometric redshift within 0.02 of the artificial sn redshift , with the probability of selecting a particular galaxy weighted by the `` a+b '' sn rate model with coefficients from ( * ? ? ? * hereafter ) . within their host galaxies , the artificial sne were assigned galactocentric positions drawn from the two - dimensional gaussian distribution about the host centroid returned by sextractor , i.e. , the artificial sne are placed with a probability that follows the light of the host galaxy . the simulated objects were assigned random values of stretch from a uniform distribution in the range @xmath117 , with colors calculated from the stretch color relationships presented in @xcite ( the use of a uniform distribution in stretch ensures that the parameter space of sn ia events is equally sampled ) . peak apparent rest - frame @xmath55 magnitudes ( @xmath118 ) at each selected redshift were calculated for our cosmology and a sn ia absolute magnitude , and adjusted for the color luminosity and stretch luminosity relations . we use an empirical piece - wise stretch - luminosity relationship with different slopes above and below @xmath119 ( e.g. * ? ? ? * ; * ? ? ? * ) , and sn ia photometric parameters from the snls3 analysis @xcite . these peak apparent magnitudes were then further adjusted by an amount @xmath66mag according to the observed intrinsic dispersion ( @xmath120 ) in sn ia magnitudes following @xmath52 and @xmath49 corrections . here , @xmath120 parameterizes a gaussian distribution from which a @xmath66mag can be assigned for each artificial event . the sn color luminosity relation includes both effects intrinsic to the sn , and extrinsic effects such as dust . we use coefficients consistent with the snls3 analysis , which favor a slope between @xmath118 and @xmath49 of @xmath121 , the value expected based on milky way dust . as there is no evidence that this slope evolves with redshift @xcite , we keep it fixed for all the artificial sne . for the detection efficiency grids , our @xmath49 values range up to 0.6 , corresponding to a sn that is @xmath122mag fainter in @xmath55-band than a normal sn ia . each artificial sn was assigned a random date of peak magnitude . for the field - season under study , this ranged from 20 observer - frame days before the first observation , to 10 days after the last observation . this ensures that the artificial events sample the entire phase range allowed by the culls in @xmath65 [ sec : culls ] at all redshifts . the light curve of each event in @xmath63 was then calculated using the @xmath123-correction appropriate for each epoch of observation , and each artificial object was added at the appropriate magnitude into every @xmath63 image . the real - time search pipeline @xcite , the same one that was used to discover the real sne , was run on each epoch of data to determine the overall recovery fraction as a function of the various sn ia parameters . the variation in candidate recovery over magnitude , redshift , stretch , and color are shown by the solid lines in fig . [ fig : compl ] . the 50% detection incompleteness limit lies at @xmath124 mag in the ab system . as expected , sne ia that are high stretch , blue , or at lower redshift are all generally easier to recover . note that at lower redshifts , the faster ( less time - dilated ) nature of the sn ia light curves means that the observational criteria of [ sec : culls ] are slightly more likely to remove events ( as there are fewer opportunities to observe a faster sn ) , hence the observed decrease in the recovered fraction towards lower - redshifts . that is , a low-@xmath4 sn that peaks during bright time is less likely to be recovered than a higher-@xmath4 sn peaking at the same epoch , even if they had the same observed peak magnitude . this is also partially reflected in the fraction recovered as a function of magnitude , with a curvature in the recovered fraction towards brighter magnitudes . the recovery results are discussed in detail in @xcite . a grid of detection efficiencies was constructed independently for each field - season using the recovery statistics in bins of measured redshift ( @xmath125 ) , stretch ( @xmath126 ) , and color ( @xmath127 ) . these bin sizes were found to provide adequate resolution in each parameter . we investigated the use of a higher resolution in stretch and color , and found no significant on impact our results . every observed phot - ia in the snls sample is thereby assigned a detection efficiency by linearly interpolating in @xmath56 space that corresponds to the field - season during which it was detected , along with its other measured parameters : @xmath128 . these detection efficiencies are plotted in fig . [ fig : objeffs_raw ] prior to any adjustments for sampling time and the availability of observations . redder , lower - stretch sne ia tend to have smaller detection efficiencies , as shown by the open circles in fig . [ fig : objeffs_raw ] . for clarity , detection efficiency errors are not shown in fig . [ fig : objeffs_raw ] . statistical uncertainties on @xmath128 for well - sampled data are governed by the number of monte carlo simulations performed , and are small in comparison to the systematic error resulting from assumptions made about the underlying intrinsic sn ia magnitude dispersion ( the @xmath66mag distribution , parameterized by @xmath120 ) . to estimate these latter errors , the detection efficiencies are recalculated using a range of @xmath120 values from @xmath129 . bins with @xmath130 will effectively have zero uncertainty , since the likelihood of recovery will not depend on the details of the population distribution ; by contrast , `` low - efficiency '' bins are more seriously affected . these detection efficiency `` errors '' are included into the overall rate uncertainties in [ sec : errsims ] . to remain consistent in the selection criteria used for both the observed snls sample and the fake objects , we also apply the same observational cuts described in [ sec : culls ] to the artificial sne ia . using the peak date of each simulated light curve , we determine whether the minimum observing requirements are met in each filter by comparing with the snls image logs . this directly incorporates the observational cuts into the detection efficiency calculations , while factoring in losses due to adverse weather and the gaps between epochs . the recovery fractions that include these observational requirements are shown by the lower dashed lines in fig . [ fig : compl ] . each candidate s detection efficiency is multiplied by a factor to account for its corresponding sampling time window for detection , yielding a `` time corrected '' rest - frame efficiency @xmath131 : @xmath132 the sampling period @xmath133 ( in years ) for a given field - season is @xmath134 , \label{eq : searchtime}\ ] ] where mjd is the modified julian date of the available detection images . the extra 30 days account for the range in peak dates allowed for the artificial sn ia light curves , from 20 days prior to the first observation in a given field - season to 10 days past the final epoch . the resulting time - corrected rest - frame detection efficiencies for the phot - ia sample are plotted as a function of redshift in fig . [ fig : objeffs ] . since each field is observable for at most @xmath135 months of the year , @xmath131 peaks at @xmath136 even for bright , nearby objects . figs . [ fig : objeffs_raw ] and [ fig : objeffs ] show that there is a drop - off in the efficiencies above @xmath137 , in particular for the redder @xmath49 bins , making it more difficult to calculate accurate rates at these redshifts due to color stretch bins which are not sampled . at @xmath138 , it is not possible to measure sn ia rates using this method due to poor survey sensitivity and inadequate statistical sampling of spectral templates . therefore , we restrict our volumetric rate calculations to the range @xmath139 . volumetric sn ia rates are calculated from eqn . [ eq : rate ] by summing the observed sne ia weighted by the inverse of their time - corrected rest - frame efficiencies . the total sampling volumes for the deep fields in redshift bins of @xmath125 ( eqn . [ eq : volume ] ) are provided in column 2 of table [ tab : rates ] . columns 3 and 4 show the numbers of observed candidates in each bin for the entire sample ( @xmath140 ) and for the spectroscopically confirmed sne ia ( @xmath141 ) in each redshift bin . the `` raw '' measured rates ( @xmath142 ) with their weighted statistical errors are given in column 5 , in units of @xmath143 ; see later sections for the meaning of the remaining columns . lccccccc @xmath144 & 17.3 & 4 & 3 & @xmath145 & & 0.16 & @xmath146@xmath147 + @xmath148 & 42.8 & 16 & 16 & @xmath149 & & 0.26 & @xmath150@xmath151 + @xmath152 & 75.7 & 31 & 24 & @xmath153 & & 0.35 & @xmath154@xmath155 + @xmath156 & 112.7 & 42 & 29 & @xmath157 & & 0.45 & @xmath154@xmath158 + @xmath159 & 151.5 & 72 & 47 & @xmath160 & & 0.55 & @xmath161@xmath158 + @xmath162 & 190.1 & 91 & 36 & @xmath163 & @xmath164 & 0.65 & @xmath165@xmath166 + @xmath167 & 227.2 & 110 & 56 & @xmath168 & @xmath164 & 0.75 & @xmath169@xmath170 + @xmath171 & 262.1 & 128 & 44 & @xmath172 & @xmath173 & 0.85 & @xmath174@xmath151 + @xmath175 & 294.1 & 141 & 25 & @xmath176 & @xmath177 & 0.95 & @xmath178@xmath179 + @xmath180 & 323.0 & 50 & 6 & @xmath181 & @xmath182 & 1.05 & @xmath183@xmath184 + contamination by non - ias that survive the culling criteria is estimated to contribute under @xmath185 to the total measured rates to @xmath186 . the contribution is found to be negligible up to @xmath187 , at which point it increases to around @xmath188 at @xmath186 . this is determined by summing @xmath189 for the known non - ias , and dividing by the corresponding value for objects in each redshift bin with available spectroscopy . the @xmath131 values used here are based on the results obtained when allowing redshift to vary in the fits , not when holding @xmath4 fixed at the spectroscopic redshifts . we now correct the raw measured rates for potential systematic offsets in the photometric redshifts and other parameters . this is done using the technique described next in [ sec : errsims ] , which also computes a combined statistical and systematic uncertainty on the final rates . the sn ia rates are potentially also sensitive to the inclusion of very low detection efficiency candidates at @xmath190 , and we must consider the effects of undetected sne ia in @xmath56 bins with very poor detection recovery rates . these low - efficiency issues are discussed later in [ sec : distcorr ] . in addition to the simple `` root-@xmath41 '' statistical errors , a number of additional uncertainties also affect our measured rates . these can include errors in the measured sn ( photometric ) redshift , stretch , and color , which together determine the detection efficiency ( and hence weight ) assigned to each event . as shown in fig . [ fig : zcomp ] , there is a redshift uncertainty in each measure of @xmath67 . the discrepancy between @xmath72 and @xmath67 for the confirmed sn ia sample is presented in fig . [ fig : zerrs ] for two bins in stretch . there is a small offset above @xmath191 , increasing to @xmath192 ( @xmath193 ) at @xmath194 in the @xmath195 sample , and @xmath196 ( @xmath197 ) in the @xmath198 sample . on average , the @xmath67 measurements are underestimated , with an increasing offset to higher redshift . an offset is expected based on a malmquist bias , such that brighter objects are more likely to have a spectroscopic type at a fixed redshift @xcite . however , we estimate this effect to be smaller : the solid lines in fig . [ fig : zerrs ] show this predicted offset as a function of redshift . this is calculated by estimating the rest - frame @xmath55-band apparent magnitude with @xmath4 for the adopted cosmology , and applying the @xmath66mag offsets contributed by spectroscopic selection as measured in @xcite . to study these various uncertainties , and to handle this redshift migration effectively , we perform a set of monte carlo simulations on the measured rates . we begin with the basic rate measurements , @xmath199 , from table [ tab : rates ] , and calculate how many measured sne ia that rate represents in each redshift bin by multiplying by the volume in that bin : @xmath200 . many realizations ( 5000 ) are performed by drawing @xmath201 objects from typical snls - like distributions of artificial sne ia . these are the same artificial objects as used in the detection efficiency calculations ( [ sec : effs ] ) , although the stretch , color , and @xmath66mag distributions are matched to those of the spectroscopically confirmed sn ia sample ( see * ? ? ? the distributions for input to the error calculations are shown in fig . [ fig : simdist ] . only a fraction of the objects in each redshift bin have a spectroscopic redshift , with the remainder having a sn ia photometric redshift and accompanying uncertainty ( fig . [ fig : zerrs ] ) . this `` spectroscopic fraction '' , @xmath202 is calculated in each redshift bin from fig . [ fig : objpars ] . the procedure for each realization is then as follows : @xmath200 is randomized according to the poisson distribution , using the poisson error based on @xmath203 but scaled to @xmath200 , to give @xmath204 simulated objects . each simulated object @xmath205 in each redshift bin is assigned a random redshift appropriate for that bin ( @xmath206 ) . within each bin , the probability follows a scaled number - density profile according to the expected increase in volume with redshift . 3 . the @xmath206 are then randomly matched to a sn ia from the artificial distribution with the same redshift ( fig . [ fig : simdist ] ) , and that event s stretch ( @xmath207 ) and color ( @xmath208 ) assigned to the simulated event . 4 . using the fraction of spectroscopic redshifts in each bin @xmath202 ( fig . [ fig : objpars ] ) , we assign this fraction of the @xmath209 objects to correspond to a spectroscopic redshift measurement . the remaining redshifts are assumed to come from a photometric fit , and are shifted and randomized using the median offsets and standard deviations shown in fig . [ fig : zerrs ] to give @xmath210 . any event with a @xmath72 is not adjusted . 5 . correlated stretch and color errors are then incorporated for all objects , using typical covariances produced by sifto for the snls sample , and the @xmath207 and @xmath208 randomized to @xmath211 and @xmath212 . @xmath210 , @xmath211 , and @xmath212 are used to match each simulated object to a detection efficiency . efficiency errors are included by applying a random shift drawn from an two - sided gaussian representing the asymmetric uncertainties on each value ( [ sec : mc ] ) . 7 . random numbers between zero and one are generated to evaluate whether each simulated object gets `` found '' : if the selected number is lower than the detection efficiency associated with the simulated object , that event is added to the rate calculated for that iteration . 8 . the rate for that iteration is then calculated using the appropriate detection efficiencies from step 6 . the final volumetric rates are presented in columns 7 of table [ tab : rates ] . these are calculated as the mean of the 5000 simulated rates in each redshift bin . we also calculate the standard deviation in each bin , subtract from this in quadrature the statistical poisson uncertainty based on @xmath203 , with the remainder our estimate of the systematic uncertainty in each redshift bin . the effects of the simulations described above on the input redshift , stretch , and color distributions are shown in fig . [ fig : parshift ] . the redshift histogram shows that the offsets applied in step 4 of the simulations produce a net increase in redshift to compensate for the small bias in the photometric fitting , with the effect increasing towards higher @xmath4 . this causes a flattening of the output rates calculated by the simulations as compared with the measured ( and uncorrected ) rates ( table [ tab : rates ] ) . in addition to the offset towards higher redshifts at @xmath213 , there is also a very small spread in the stretch and color distributions ( fig . [ fig : parshift ] ) . the detection efficiencies in some of the reddest @xmath49 bins begin to rapidly decrease at @xmath190 ( fig . [ fig : objeffs ] ) . as the contribution to the rate from each observed sn ia goes as @xmath189 , the measured volumetric rates are particularly sensitive to any objects with very low detection efficiencies ( @xmath131 ) . there is also the potential for a complete omission of sne ia in some @xmath56 bins . for example , in @xmath56 bins with less than @xmath214 detection efficiency , on average at least 10 sne ia must be observable in a given field - season for just one to be detected . if that one sn is not detected by the real - time pipeline , the 10 sne ia that it truly represents in that bin will never be counted in the final rates tally ( and the rate measurement will be biased ) . to examine the sensitivity of our rates on the `` low-@xmath131 regions '' of @xmath56 parameter space , we use the @xmath215 detection - efficiency - corrected snls sample as a model for the true ( @xmath52 , @xmath49 ) sn ia distribution at @xmath216 . this population is assumed observationally complete ( fig . [ fig : compl ] ) and is taken to be representative of the underlying sample of sne ia in the universe . the two - dimensional ( @xmath52 , @xmath49 ) distribution at @xmath215 is fit to the five @xmath216 bins , and the best - fit scaling determined . the total rates @xmath217 are then calculated from those scaled numbers ( tabulated in table [ tab : rates ] ) these tests indicate that , while the results remain consistent within their errors up to @xmath218 ( @xmath219 compared with @xmath220 ) , there is a significant amount of uncertainty in the snls rates at higher redshifts due to sample incompleteness . at @xmath221 , the scaled rate is @xmath222 , whereas our the calculated value is @xmath223 . this finding is consistent with the results shown in figs . [ fig : objpars ] and [ fig : objeffs_raw ] : the phot - ia sample numbers drop significantly at @xmath112 , and those that are found in the sample can have very low @xmath131 . for these reasons , we limit the formal analysis of the snls rates to @xmath224 . having measured volumetric sn ia rates and associated errors , we now compare our measurements with those of other studies . we also examine the sn ia rate evolution as a function of redshift , and compare with predictions based on various simple delay - time distribution models from the literature . for comparison of the snls rate measurements to various sn ia models , additional data in redshift ranges not sampled by snls is required . in the rest of this section , we will make use of an extended sn ia rate sample comprising the @xcite loss measurement at @xmath225 , the @xcite sample from sdss - sn at @xmath226 , and the recent @xcite subaru deep field ( sdf ) sample at higher redshifts , together with our snls results . clearly other samples could have been chosen however , these three are the largest sn ia samples in their respective redshift ranges , and have the greatest statistical power . in the case of the sdss and sdf samples , they are also built on rolling sn searches similar to snls . we make some small corrections to these published rates in order to ensure a fair comparison across samples . the @xcite sample includes all sub - classes of sne ia , including the peculiar events in the sn2002cx - like class and sub - luminous events in the sn1991bg - like class . sn2002cx - like events make up 5% of the loss volume - limited sn ia sample @xcite . these are not present ( or accounted for ) in the snls sample , and are excluded from the sdss analysis @xcite , so we therefore exclude these from the @xcite sample , reducing their published rate value by 5% . both the @xcite and @xcite samples include sne ia in the sub - luminous sn1991bg category ( see also * ? ? ? * ) , which we exclude here in the snls analysis ( these are studied in * ? ? ? while we could correct our own rates for this population using the @xcite results , it is unclear how to treat the sdf sample in the same way ( do sn1991bg - like events even occur at @xmath194 ? ) and the snls sub - luminous measurement is quite noisy . instead we use the very well - measured fraction of sn1991bg - like sne in the volume - limited loss sample ( 15% ) , and reduce both the loss and sdss published rates by this amount . this 15% is based on the classifications given in @xcite . we confirm that this is appropriate for our stretch selection ( i.e. , we require @xmath227 ) by fitting the available @xcite light curves with sifto . 16% of the available @xcite sample has a fitted stretch @xmath228 , consistent with the 15% reported as 91bg - like by @xcite . [ fig : rates_comp ] shows the snls volumetric sn ia rates for comparison with recent published results . the snls volumetric rate at @xmath111 published by @xcite is @xmath229 $ ] , consistent with our binned rates in the same redshift range . note that the effects of non - ia contamination ( [ sec : snlsrates ] ) have not been incorporated into the snls errors shown in fig . [ fig : rates_comp ] . our results are also consistent with @xcite at @xmath230 , and with @xcite at higher redshifts , although those latter measurements have significantly greater uncertainties . the snls rates show a rise out to @xmath186 , with no evidence of a rollover at @xmath111 . we can parameterize the snls rate evolution as a simple power - law : @xmath231 with the best - fit shown as the solid line in fig . [ fig : rates_powerlaw ] . we find @xmath232 and @xmath233 ( all errors in this section are statistical only ) . this evolution is shallower than a typical fit to the cosmic star - formation history , with @xmath234 ( e.g. * ? ? ? * ) , though the constraining power of the snls data alone at @xmath21 is not great . by comparison , @xcite find @xmath235 and @xmath236 using the lower - redshift sdss - sn data , completely consistent with our results . including all the external data gives @xmath237 and @xmath238 . lccccccccccc & & & @xmath239 ( gyr ) & & & @xmath239 ( gyr ) & & & @xmath239 ( gyr ) & & + extended & gaussian & 18 & 3.4 & & 2.62 & 3.4 & & 1.60 & 3.4 & & 3.29 + extended & gaussian & 17 & @xmath240 & & 2.64 & @xmath241 & & 1.60 & @xmath242 & & 3.00 + ext.+d08 & gaussian & 19 & 3.4 & & 2.72 & 3.4 & & 1.68 & 3.4 & & 3.43 + + & & & @xmath243 & & & @xmath243 & & & @xmath243 & & + snls & power law & 8 & 1 & & 0.76 & 1 & & 1.36 & 1 & & 0.81 + extended & power law & 18 & 1 & & 0.72 & 1 & & 1.06 & 1 & & 0.78 + extended & power law & 17 & @xmath244 & & 0.76 & @xmath245 & & 0.92 & @xmath244 & & 0.81 + + & & & @xmath239 ( gyr ) & & & @xmath239 ( gyr ) & & & @xmath239 ( gyr ) & & + snls & exponential & 7 & @xmath246 & & 0.85 & @xmath247 & & 0.58 & @xmath248 & & 0.91 + extended & exponential & 17 & @xmath249 & & 1.33 & @xmath250 & & 1.15 & @xmath251 & & 1.44 + + & & & @xmath243 & & & @xmath243 & & & @xmath243 & & + snls & & 8 & 0.5 & & 4.14 & 0.5 & & 4.81 & 0.5 & & 4.31 + extended & & 18 & 0.5 & & 4.08 & 0.5 & & 4.77 & 0.5 & & 4.19 + + & & & @xmath252 & @xmath55 & & @xmath252 & @xmath55 & & @xmath252 & @xmath55 & + snls & @xmath253 & 7 & @xmath254 & @xmath255 & 0.74 & @xmath256 & @xmath257 & 0.59 & @xmath258 & @xmath259 & 0.78 + extended & @xmath253 & 17 & @xmath260 & @xmath261 & 0.60 & @xmath262 & @xmath263 & 0.77 & @xmath264 & @xmath265 & 0.63 + snls & @xmath266 & 8 & 0 & @xmath267 & 1.81 & 0 & @xmath268 & 0.54 & 0 & @xmath269 & 1.92 + extended & @xmath266 & 18 & 0 & @xmath270 & 6.22 & 0 & @xmath271 & 2.95 & 0 & @xmath272 & 6.70 + snls & @xmath273 & 8 & @xmath274 & 0 & 9.72 & @xmath275 & 0 & 9.74 & @xmath276 & 0 & 10.09 + extended & @xmath273 & 18 & @xmath277 & 0 & 9.42 & @xmath278 & 0 & 9.56 & @xmath277 & 0 & 9.66 + + & & & @xmath279 & @xmath280 & & @xmath279 & @xmath280 & & @xmath279 & @xmath280 & + extended & 2-bin & 17 & @xmath281 & @xmath282 & 0.64 & @xmath283 & @xmath284 & 0.79 & @xmath285 & @xmath286 & 0.67 + we now compare our rate evolution with simple parameterizations of the delay - time distribution ( dtd ) from the literature relating the cosmic star - formation history ( sfh ) to sn ia rates . the dtd , @xmath287 , gives the sn ia rate as a function of time for a simple stellar population ( ssp ) , i.e. , following a @xmath288-function burst of star formation . the sn ia rate ( ) at time @xmath77 is then @xmath289 where sfr(@xmath77 ) is the star - formation rate as a function of time . thus , different functional forms of the dtd can be tested against observations of volumetric rates if the sfr(@xmath77 ) , or the cosmic sfh , is known ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? an implicit assumption in this test is that the dtd is invariant with redshift and environment . as our default sfh model , we choose the @xcite update to the @xcite fit to a compilation of recent star - formation density measures . for simplicity we use a @xcite initial mass function ( imf ) with mass cut - offs at 0.1 and 100 , and assume that stars began to form at @xmath290 . the default sfh is parameterized in a piece - wise fashion , and is over - plotted on the data in fig . [ fig : rates_comp ] . as shown by @xcite and @xcite , the choice of sfh can add an additional significant systematic uncertainty in any comparisons of dtds to sn ia volumetric data . we therefore compare with results obtained using an alternative parameterization of the sfh by @xcite following @xcite , as well as a sfh fit to slightly different data by @xcite ( see also * ? ? ? * ) . in @xmath65 [ sec : comparison - other - dtd ] , we also investigate a sfh derived in a very different manner @xcite . the integral of the dtd gives the total number of sne ia per formed stellar mass , @xmath291 . this can be converted into the fraction of intermediate - mass stars that explode as sne ia , @xmath292 , by multiplying by a factor of 47.2 ( for the salpeter imf ) . this assumes that the progenitor mass range for a sn ia is 38(see * ? ? ? * for a discussion ) . for all our model dtds , we set the dtd to zero at epochs earlier than 40myr , the approximate lifetime of an @xmath293 star . we begin by fitting a gaussian dtd , with @xmath294 , to the volumetric data , following @xcite ( see also * ? ? ? we fit a dtd with parameters fixed at @xmath295gyr and @xmath296 ( i.e. , just adjusting the normalization in the fits ) , as well as a dtd fit with @xmath239 allowed to vary . the results are listed in table [ tab : abchi2 ] and compared to other dtd fits in fig . [ fig : rates_dtd_compare ] . this model has @xmath297 ( @xmath90 is the reduced @xmath91 , the @xmath91 per degree of freedom , @xmath92 ) . allowing @xmath239 as a free parameter in the fits gives @xmath298gyr with a similar @xmath90 the fit quality is slightly better when using cole et al . form of the sfh ( @xmath299 ) . the gaussian dtds therefore provide poor fits to the data , and are not capable of matching the snls , sdf and sdss / loss data simultaneously . in particular , these gaussian dtds predict a decrease in the number of sne at @xmath194 not seen in the combined data set . however , they were originally favored following fits to data including @xmath194 points from _ hst _ searches @xcite which are not included in our analysis due to their lower statistical precision compared to the sdf study . as a consistency check we also replace the sdf data with the @xcite data ( adjusted downwards by 15% to account for 19bg - lie events ) in our fits we find that the @xmath90 does not improve ( table [ tab : abchi2 ] ) . theoretically , if sne ia are dominated by a single channel , the dtd will likely decline with age . in the single degenerate channel , sne ia at 10gyr should be rare , since 1 secondaries have small envelopes to donate and must rely on only the most massive primaries @xcite . a power law dtd ( i.e. , @xmath300 with @xmath301 ) with a low time - delay cut - off is expected in the double degenerate scenario ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) , and has been explained post - hoc in the single degenerate channel using a mixture of contributions @xcite . furthermore , models with @xmath301 seem to provide a good match to a variety of recent observational data @xcite . we fit both @xmath302 and free @xmath243 dtds to the snls+external data . the results can be found in table [ tab : abchi2 ] and fig . [ fig : rates_dtd_compare ] . our best - fit value is @xmath8 ( @xmath303 ) , consistent with @xmath304 . this broad agreement with @xmath304 holds when considering the other sfh paramterizations ( @xmath9 for the cole et al . form ) . * hereafter phs ) present a simple model relating white dwarf formation rate , which decreases with time following an instantaneous burst of star formation as @xmath305 , resulting in a dtd with @xmath306 . by fitting the sn ia host galaxy data of , demonstrate that @xmath307 , irrespective of the assumed sfh or the detailed mixture of stellar populations . argue that the single - degenerate formation scenario alone is not sufficient to account for all of the observed sne ia ( see also * ? ? ? the model makes an explicit prediction for the evolution of the sn ia rate with redshift , given an input sfh . we fit the model essentially @xmath308 to the data and show the resultant fit in fig . [ fig : rates_dtd_compare ] ( also table [ tab : rates ] ) . we find a @xmath309 , obviously a substantially poorer fit than the generic power - law fit , or power - law dtds with @xmath302 . for completeness we also also test an exponential dtd , i.e. , @xmath310 . when @xmath239 is small , this approximates a simple star - formation dependence , and when large , it approximates a constant dtd . the results are in table [ tab : abchi2 ] ; generally , single exponential dtds provide poor fits to the data , but do still have acceptable @xmath91 . finally , we examine various two - component dtd models . the first is the popular `` @xmath253 '' model , a simple , two - component model of sn ia production that is comprised of a `` prompt '' component that tracks the instantaneous sfr , and a `` delayed '' ( or `` tardy '' ) component that is proportional to @xcite : @xmath311 here , the @xmath252 and @xmath55 coefficients scale the and sfr components , respectively . the prompt component consists of very young sne ia that explode relatively soon ( in the model , immediately ) after the formation of their progenitors , whereas the delayed component ( scaled by @xmath252 ) corresponds to longer delay times and an underlying old stellar population . this model is empirically attractive due to the ease of comparison with readily observable galaxy quantities , such as and sfr . note that this a+b model does not exactly correspond to a dtd however , it can be easily converted to a dtd by using the variation of with time in a ssp . this leads to a dtd with some fraction of sne ia formed immediately ( the @xmath55 component ) , followed by a slightly decreasing fraction to large times ( the @xmath252 component ) . this decrease is @xmath312% from 0.1 to 5gyr , and @xmath313% from 1gyr to 10gyr clearly significantly shallower than a @xmath302 power law . some confusion exists over the exact definition of in eqn . ( [ eq : aplusb ] ) , and hence the definition of @xmath252 . some authors ( e.g. * ? ? ? * ) simply treat as the integral of the sfh , equating it to the total formed mass , . others make corrections for stars that have died , particularly in studies which perform analyses on a galaxy - by - galaxy basis as this quantity is more straight forward to link to observational data ( e.g. , * ? ? ? this latter definition leads to larger @xmath252 values , as @xmath314 will be less than @xmath315 . here , our @xmath252 values refer to , and we pass the cosmic sfh through the pgase.2 routine @xcite , convolving the chosen sfh with a single stellar population , and generating a galaxy sed@xmath316 from which mass @xmath314 can be estimated . the evolving and sfr are used to perform a fit of equation [ eq : aplusb ] to the volumetric rate evolution , with results listed in table [ tab : abchi2 ] . fitting the snls rates alone gives coefficients of @xmath317 and @xmath318 for the @xcite piece - wise sfh , with @xmath319 for @xmath320 . incorporating the external sn ia rate from loss , sdss and sdf yields values of @xmath321 and @xmath12 , with @xmath322 . this fit is compared to other dtds in fig . [ fig : rates_dtd_compare ] . next , we set @xmath266 to investigate the possibility of a pure star - formation dependence , fitting only the prompt ( @xmath55 ) component to the snls data . this results in an upper limit of @xmath323 ( @xmath324 ) , equivalent to the normalized sfh curve plotted in fig . [ fig : rates_comp ] . adding the external data rates to the snls values gives an upper limit of @xmath325 , but again with a very poor fit quality ( @xmath326 ) . while the snls results themselves are marginally consistent with a pure prompt component ( table [ tab : abchi2 ] ) , adding the additional constraints supplied by the low-@xmath4 data yield a very poor @xmath266 fit . the related test of setting @xmath273 and testing for only the delayed component similarly gives very poor fit results ( table [ tab : abchi2 ] ) . as discussed above , these @xmath252 and @xmath55 values depend on the adopted sfh . even ignoring the systematics in the individual sfr measurements to which the sfh model is fit , the type of fit used also introduces considerable uncertainty . using the cole et al . form of the sfh , for example , changes the best - fit values to @xmath327 and @xmath328 , a significant variation with similar quality fit than for the piece - wise form . thus care must be taking in comparing any particular @xmath253 values , or any particular prediction of sn ia rate evolution , without ensuring the consistent use of a sfh and derivation of . clearly , and as well documented in the literature , the @xmath253 model must be a significant approximation to the physical reality in sn ia progenitor systems . in particular , there must be _ some _ delay time for the prompt component , and it is unlikely to act as a delta function in the dtd . we test this by approximating @xmath287 as two discrete bins in time , i.e. , a step function with a value @xmath279 at times @xmath329 , and @xmath280 at @xmath330 . we choose @xmath331myr ( following , e.g. , @xcite ) , and used a sigmoid function to ensure the dtd was continuous when crossing @xmath332 . this dtd also provides a good fit to the data ( table [ tab : abchi2 ] ) , with a significant detection of the two components @xmath279 and @xmath280 are @xmath333 at 5@xmath334 in all three sfhs considered ( and typically @xmath335 ) . we experimented with making this function more general , by allowing @xmath336 to vary . however , these fits were not constraining , although they prefer @xmath337gyr , and the fit parameters are highly correlated ( i.e. , as @xmath332 decreases , @xmath279 increases to ensure a similar fraction of sne ia are generated from the `` prompt '' component ) . one direct outcome of this simple two - bin dtd model is that , for our default cosmic sfh , while @xmath338% of sne ia originate from the prompt component integrated over cosmic time , at @xmath27 the prompt component accounts for only @xmath312% of sne ia . these fractions remain fairly constant out to @xmath339gyr . we also explored more bins in @xmath340 e.g. , the three bin dtd of @xcite and @xcite but , while these were consistent with our data , they did not provide improved fits over the two - bin dtd , and again , the parameters themselves were not well constrained . we now compare our dtds inferred from the volumetric data with other independent estimates from the literature , comparing both the parametric form of our best - fit dtds , as well as the normalization . in fig . [ fig : dtd ] , we plot our inferred dtds , with normalizations from the best - fits to the volumetric sn ia rates , and compare to other empirical determinations of the dtd from the literature from a variety of methods @xcite . the data points come from the analysis of sn ia rates in galaxy clusters @xcite , the reconstruction of the dtd from an analysis of the loss sn ia host galaxy spectra @xcite , and a similar analysis of the host galaxies of sdss sne ia @xcite . where necessary we convert these external measurements to a salpeter imf from a diet - salpeter imf @xcite for @xcite and a @xcite imf for @xcite . we also adjust the @xcite and @xcite results downwards as their measurements presumably include sn1991bg - like events which are not included in our snls analysis ( the brandt et al . analysis does not include these sne , and their stretch range is well matched to our analysis see their figure 2 ) . furthermore , we correct the @xcite dtd points upwards by 0.26dex to account for stellar masses in brandt et al . that are 0.26dex too high due to a normalization issue in the vespa code used to derive them ( d. maoz , private communication ; r. tojeiro , private communication ) . although the generic shape of the power - law dtd inferred from the volumetric rate data matches the external dtd data well ( fig . [ fig : dtd ] ) , it is clear that the best - fit normalization required to reproduce the volumetric rate data differs to that required to fit some of the external samples . the dtds inferred from volumetric rate data are generally consistent with the @xcite analysis , and the first and third bins of the @xcite data . however , the normalization of the best - fit dtd to the @xcite cluster data lies significantly above our best - fit dtd . integrating our best - fit power - law dtds gives @xmath341 ( @xmath342% ) depending on the sfh , in good agreement with similar analyses @xcite . however , this is significantly below the value of @xmath343 obtained by integrating the @xcite `` optimal iron - constraint '' power - law dtd ( for our imf ) , or @xmath344 from the `` minimal iron constraint '' dtd . we can sanity check our normalizations by predicting the sn ia rate in the milky way , given our dtd values . assuming a milky way stellar mass of @xmath345 and a sfr of @xmath346 , our @xmath253 dtds give a predicted milky way normal sn ia rate of 0.22 - 0.25 events per century . this is in good agreement with independent estimates of the actual rate ( @xmath347 to @xmath348 events per century ; e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) given the uncertainties involved . so why is the normalization in the @xcite cluster sn ia dtd ( which is not arbitrary ) different to those derived from volumetric rate data by such a large factor ? several possibilities exist that may explain the discrepancy . the first is that the snls , and by extension all other studies , are missing a significant number of sne ia , a factor of at least four . however , it is difficult to understand how this might occur , given some of the cluster rates used in @xcite are drawn from very similar surveys ( including snls and sdss - sn ) . a second possibility is that the cosmic sfh models used in our analysis over - predict the actual sfr a lower sfh normalization would require a higher dtd normalization to match the volumetric rate data . to test this possibility , we repeat our rates fitting analysis using the sfh of @xcite , a sfh derived from a requirement to match the redshift evolution of . although this agrees with other sfh estimates at @xmath349 , it suggests that high - redshift sfrs could be @xmath350dex lower compared to the models used in this paper . however , even using this sfh , the integrated power - law dtd only gives @xmath351 ( @xmath352% ) , still some distance short of that apparently required by the clusters analysis . other options to adjust the sfh normalization downward , such as reducing the dust extinction corrections applied to the various star - formation indicators which make up the sfh compilations , are probably not viable given the long - established and significant evidence for obscuration in star - forming galaxies , and the agreement between the different diagnostics ( see discussion in * ? ? ? * ; * ? ? ? a third possibility is that the cluster rates used to derive the dtd of @xcite have some contamination from `` younger '' sne ia , thus increasing their rates above that appropriate for their age , assuming a redshift of formation of 3 . this , and other similar potential systematics from the clusters analysis , are discussed in detail in @xcite . finally , it may well be the case that the assumption of a single dtd is not adequate , given the various indications that there may be more than one progenitor channel . this would suggest that there is not one universal dtd that is independent of redshift , or other variables , such as metallicity . there is an observed correlation between the photometric properties of sne ia and their host environments @xcite . brighter sne ia with slower light - curves tend to originate in late - type spiral galaxies , such that the rates of higher - stretch objects are proportional to star - formation on short ( @xmath353 gyr ) timescales . meanwhile , fainter , faster - declining sne ia are more likely to be associated with older stellar populations . this split seems to extend to the recovered dtds @xcite show that the recovered dtd for low and high stretch sne ia are very different , consistent with the above picture of young sne ia being high stretch and old sne ia low stretch . a larger fraction of high - stretch sne ia might therefore be expected at high redshift tracking the increase in the cosmic sfh and hence preponderance of younger stars . indeed , @xcite find a modest increase in the average light - curve width out to redshifts of @xmath354 . the rate evolution for low - stretch sne ia should therefore demonstrate a correspondingly shallower increase with redshift than that of higher - stretch objects . we investigate this trend in fig . [ fig : scomp ] , splitting the snls sample at the median stretch value of @xmath355 . the measured rates with simple weighted errors are plotted separately for objects with @xmath356 ( open circles ) and those with @xmath198 ( filled circles ) . we also show the @xmath357 rate measurement from @xcite . the samples from the analysis in this paper exhibit a comparable rise in their rates with redshift , with power - law slopes of @xmath358 and @xmath359 . extrapolating the fits to each of the samples back to @xmath27 gives the following fractions of sne ia at @xmath27 in each group : 17% ( @xmath357 ) , 39% ( @xmath356 ) , and 44% ( @xmath198 ) . the @xmath357 fraction is consistent with the loss sn1991bg - like fraction of 15% @xcite , given that true 1991bg - like events have fitted stretches @xmath360 @xcite . the fraction of very low stretch events with @xmath357 sne ia shows only a small increase with increasing redshift ( see also * ? ? ? * ) , although the uncertainties are very large . the stretch - split rates are only considered out to @xmath361 , beyond which redshift the stretch errors become large ( @xmath362 ) , and the lower efficiencies of the redder , lower - stretch sne ia can bias the observed stretch evolution by driving up the @xmath363 rates ( fig . [ fig : objeffs ] ) . fig . [ fig : sratio ] shows the ratio of the rates split by stretch for @xmath364 . to compare this observed data with any expected evolution , we need model stretch distributions for `` old '' and `` young '' sne , together with a mechanism for predicting the relative evolution of these two components with redshift . for the former , we take the two stretch distributions for young and old sne ia from @xcite : the old component is represented by a gaussian with @xmath365 and @xmath366 , while the young component has @xmath367 and @xmath368 . to estimate the relative redshift evolution of the old and young components , we use the best - fitting a+b values from table [ tab : abchi2 ] ( assigning @xmath252 to the old sne and @xmath55 to the young sne ) , as well as the power - law dtd . for this latter dtd , we assign sne born at @xmath369gyr in the dtd to the young component , and sne born at @xmath370gyr to the old component . together with a sfh , these models then predict the relative fraction of low and high stretch sne as a function of redshift , over - plotted in fig . [ fig : sratio ] . we also vary the @xcite stretch distributions by adjusting @xmath371 by @xmath372 for the two components ; these are shown as the hashed gray areas in the figure . as expected , the predicted ratios show a smooth decline from a large fraction of low - stretch sne at @xmath373 . as the relative contribution from delayed sne ia to the rates decreases with increasing redshift , so too should the dominance of lower - stretch objects ( see also fig . 1 in * ? ? ? the prediction based on the power - law dtd shows a shallower evolution with redshift , reflecting the extended age of the young component relative to the simplistic @xmath253 model . however , fig . [ fig : scomp ] and fig . [ fig : sratio ] are surprising . if broad - lightcurve sne ia favor a young environment and narrow - lightcurve events favor an old environment , as has been well established , then the ratio of narrow to broad sne ought to be changing as star formation increases with redshift . but all of the predictions are a relatively poor match to the snls sample in the higher redshift bins . we vary @xmath374 by @xmath375 to assess the sensitivity of our results to the stretch split value ( default of 1.0 ) , but find no significant improvement in the agreement with the predicted model as compared with the straight - line fit at the weighted mean ratio . the lack of an observed evolution may be due to several factors . the first is that the age - split between low- and high - s sne ia may be more subtle than previously appreciated . alternatively , the main effect may be dominated by @xmath357 sne ia , which do show a different evolutionary trend with redshift ( fig . [ fig : scomp ] ) . the limited time baseline of only @xmath376gyr from @xmath377 may also be a factor , and of course limitations of the method ( e.g. , the arbitrary cutoff of for sne to be `` young '' or `` old '' in the power - law dtd ) could mask any real effect . some other aspects of fig . [ fig : sratio ] are better understood . the @xmath253 model over - predicts low - s sne at @xmath27 because it has only 20% fewer sne in the dtd at 12gyr compared to 1gyr , in apparent contrast to the data and to the @xmath378 model , where the rate falls by an order of magnitude over this baseline . these excess old sne ia from @xmath379 show up after a 10 gyr delay at @xmath27 . however , the difference in the predictions between the @xmath253 and power - law models is not large , so this unphysical assumption can not provide the only solution for the lack of observed evolution in fig . [ fig : sratio ] . in this paper , we have probed the volumetric rate evolution of `` normal '' @xmath380 sne ia using a sample of @xmath109 events from the supernova legacy survey ( snls ) in the range @xmath381 , @xmath0 of which have been confirmed spectroscopically . the snls rates increase with redshift as ( 1+@xmath4)@xmath5 with @xmath6 , and show no evidence of flattening beyond @xmath111 . due to spectroscopic incompleteness and the decrease in detection efficiency for the snls sample , a rollover in the slope can not be ruled out beyond @xmath186 based on the snls data alone . as a significant component of the sn ia rate is linked with young stellar populations , an increasing fraction of sn ia events may suffer the effects of host extinction at higher redshifts . in our rate calculation method , the effect of sn color is factored directly into the detection efficiency determinations : detection recovery is evaluated empirically according to the observed sn color regardless of its cause . redder objects at a given redshift have lower detection efficiencies , and are correspondingly more heavily weighted in the rates determination . combining the snls data with that from other sn ia surveys , we fit various simple delay - time distributions ( dtds ) to the volumetric sn ia rate data . dtds with a single gaussian are not favored by the data . we find that simple power - law dtds ( @xmath300 ) with @xmath301 ( @xmath8 to @xmath9 depending on the parameterization of the cosmic sfh ) can adequately explain all the sn ia volumetric rate data , as can two - component models with a prompt and delayed channel . these models can not be separated with the current volumetric rate data . integrating these different dtds gives the total number of sne ia per solar mass formed ( excluding sub - luminous @xmath357 events ) of @xmath382 ( assuming a salpeter imf ) , depending on the star formation history and dtd model . this is in good agreement with other similar analyses , but lies significantly below the number expected from dtds derived from cluster sn ia rates . the use of other techniques , such as fitting the sfh of individual galaxies @xcite , or observing a simplified subset of galaxies @xcite , use more information , and in principle ought to be more reliable . however , each technique has significant drawbacks , such as contamination @xcite , limitations of sed fitting codes @xcite , and the assumption that all cluster galaxies formed at @xmath383 in a delta - function of star - formation @xcite . therefore , our results are an important complementary constraint . by presenting an evolution in the sn ia rate over a large redshift baseline done self - consistently by a single survey we have for the first time mitigated the primary drawback of this method having to combine myriad rate determinations from multiple surveys , all done with different assumptions and biases , sometimes disparate by large factors @xcite . we also find no clear evidence for a difference in the rate evolution for snls samples with @xmath13 and @xmath384 out to @xmath361 , although the stretch evolution model from @xcite can not be ruled out conclusively . stretch evolution plays a more significant role in the sub - luminous population @xcite , which show a much flatter evolution than the @xmath364 sample . next generation surveys such as dark energy survey ( des ) , pan - starrs , palomar transient factory ( ptf ) , and skymapper , many of which are already underway , are finding thousands of sne ia ( in comparison to the @xmath33 in this study ) . statistical rate determinations ought to improve , but systematic difficulties will remain , as not all sne can be spectroscopically confirmed . however , large number statistics will allow the construction of sub - samples larger than the three ( split by stretch ) analyzed here . comparison of the relative rates of sne with different properties and in different environments may ultimately improve deduced dtds , and allow for the construction of different dtds for subsets of sne ia . we are sincerely grateful to the entire queued service observations team and staff at cfht for their patience and assistance throughout the snls real - time observing period . we are particularly indebted to pierre martin , jean - charles cuillandre , kanoa withington , and herb woodruff . canadian collaboration members acknowledge support from nserc and ciar ; french collaboration members from cnrs / in2p3 , cnrs / insu and cea . ms acknowledges support from the royal society . this work is based on observations obtained with megaprime / megacam , a joint project of cfht and cea / dapnia , at the canada - france - hawaii telescope ( cfht ) which is operated by the national research council ( nrc ) of canada , the institut national des sciences de lunivers of the centre national de la recherche scientifique ( cnrs ) of france , and the university of hawaii . this work is based in part on data products produced at the canadian astronomy data centre as part of the canada - france - hawaii telescope legacy survey , a collaborative project of nrc and cnrs . this work is based in part on observations obtained at the gemini observatory , which is operated by the association of universities for research in astronomy , inc . , under a cooperative agreement with the nsf on behalf of the gemini partnership : the national science foundation ( united states ) , the science and technology facilities council ( united kingdom ) , the national research council ( canada ) , conicyt ( chile ) , the australian research council ( australia ) , cnpq ( brazil ) and conicet ( argentina ) . gemini program ids : gs-2003b - q-8 , gn-2003b - q-9 , gs-2004a - q-11 , gn-2004a - q-19 , gs-2004b - q-31 , gn-2004b - q-16 , gs-2005a - q-11 , gn-2005a - q-11 , gs-2005b - q-6 , gn-2005b - q-7 , gn-2006a - q-7 , gn-2006b - q-10 , and gn-2007a - q-8 . observations made with eso telescopes at the paranal observatory under program ids 171.a-0486 and 176.a-0589 . 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 present a measurement of the volumetric type ia supernova ( sn ia ) rate ( ) as a function of redshift for the first four years of data from the canada - france - hawaii telescope ( cfht ) supernova legacy survey ( snls ) . this analysis includes @xmath0 spectroscopically confirmed and more than @xmath1 additional photometrically identified sne ia within the redshift range @xmath2 . the volumetric evolution is consistent with a rise to @xmath3 that follows a power - law of the form ( 1+@xmath4)@xmath5 , with @xmath6 . this evolutionary trend in the snls rates is slightly shallower than that of the cosmic star - formation history over the same redshift range . we combine the snls rate measurements with those from other surveys that complement the snls redshift range , and fit various simple sn ia delay - time distribution ( dtd ) models to the combined data . a simple power - law model for the dtd ( i.e. , @xmath7 ) yields values from @xmath8 to @xmath9 depending on the parameterization of the cosmic star formation history . a two - component model , where is dependent on stellar mass ( ) and star formation rate ( sfr ) as @xmath10 , yields the coefficients @xmath11 and @xmath12 . more general two - component models also fit the data well , but single gaussian or exponential dtds provide significantly poorer matches . finally , we split the snls sample into two populations by the light curve width ( stretch ) , and show that the general behavior in the rates of faster - declining sne ia ( @xmath13 ) is similar , within our measurement errors , to that of the slower objects ( @xmath14 ) out to @xmath15 .
_ spitzer _ spectroscopy has revealed that @xmath1% of smgs are sb - dominated in the mid - infrared ( @xcite ; @xcite ) . we have focussed on the remaining @xmath2% that show signs of harboring powerful agn and use _ spitzer_-irs spectroscopy to study a sample of 8 smgs @xmath3jy at 24@xmath4 m from the scuba half degree extragalactic survey ( shades ; @xcite ) selected on the basis of an irac color - selection ( @xmath5 ; i.e. likely power - law mid - infrared seds ; see fig . 1 ) . the full analysis will be presented in coppin et al.(apj , submitted ) , and our main results are as follows : * there are signs of sf from pah features in _ all _ of our smgs , from which we derive redshifts between 2.53.4 , demonstrating the power of the mid - ir to determine redshifts when the optical counterparts are too faint to study with current facilities . * the irs spectra show signs of _ both _ sb and agn activity in our @xmath5 smg sample , with a continuous distribution of agn fractions in the mid - ir . overall , smgs selected in this way tend to have more dominant agn - components in the mid - ir than typical smgs , with a median agn - fraction of 58% . although , extrapolation to the far - ir reveals that the agn is bolometrically unimportant in the majority of smgs , indicating that the level of agn contamination in the overall smg population is probably @xmath6% . for comparison , typical smgs have @xmath7% agn contribution in the mid - infrared , while they are from a very similar sb @xmath8 class to our sample of @xmath9 . * when literature sources are taken into account , a colour - selection of @xmath10 is a better description overall for defining the boundary between sb and agn - dominated smgs , with a small amount of scatter across this division ( see fig . 1 ) . * our results are thus consistent with the evolutionary scenario @xcite , with all smgs undergoing a ` transitional ' agn - dominated phase with a duty cycle of @xmath2% . our sample of agn - dominated smgs could be at a slightly later stage of evolution than sf - dominated smg systems , with the sf still occurring but where the agn has now begun to heat the dust appreciably in the smg as the bh undergoes a period of rapid growth . we would like to thank the conference organizers , stfc , nsf , nasa and the royal society for financial support , and also our collaborators from shades who contributed to this work . this work is based in part on observations made with the _ spitzer space telescope _ , which is operated by the jet propulsion laboratory , california institute of technology under a contract with nasa .	it is widely believed that ultraluminous infrared ( ir ) galaxies and active galactic nuclei ( agn ) activity are triggered by galaxy interactions and merging , with the peak of activity occurring at @xmath0 , where submillimetre galaxies are thousands of times more numerous than local ulirgs . in this evolutionary picture , submillimetre galaxies ( smgs ) would host an agn , which would eventually grow a black hole ( bh ) strong enough to blow off all of the gas and dust leaving an optically luminous qso . to probe this evolutionary sequence we have focussed on the missing link sources , which demonstrate both strong starburst ( sb ) and agn signatures , in order to determine if the sb is the main power source even in smgs when we have evidence that an agn is present from their irac colours . the best way to determine if a dominant agn is present is to look for their signatures in the mid - infrared with the _ spitzer _ irs , since often even deep x - ray observations miss identifying the presence of agn in heavily dust - obscured smgs . we present the results of our audit of the energy balance between star - formation and agn within this special sub - population of smgs where the bh has grown appreciably to begin heating the dust emission . address = icc , durham university , uk address=_spitzer fellow _ , national optical astronomy observatory , tuscon , az , usa address = nsf astronomy and astrophysics postdoctoral fellow , the observatories of the carnegie institution for science , pasadena , ca , usa address = department of physics , durham university , uk address = scottish universities physics alliance , institute for astronomy , university of edinburgh , royal observatory , uk
the importance of the cosmological light - cone effect in the future redshift surveys , including the two - degree field ( 2df ) and the sloan digital sky survey ( sdss ) , has been well recognized recently ( e.g. , matarrese et al . 1997 ; matsubara , suto , & szapudi 1997 ; nakamura , matsubara , & suto 1998 ; moscardini et al . 1998 ; de laix & starkman 1998 ) . in a series of previous papers ( yamamoto & suto 1999 , paper i ; nishioka & yamamoto 1999a , paper ii ; suto et al . 1999 , paper iii ) , we have extensively explored and discussed various aspects of the cosmological light - cone effect on the two - point correlation functions , @xmath1 , of objects at high redshifts . extending the procedure described in these papers , we develop here a formula to predict the power spectrum of cosmological objects located on the light cone , @xmath2 . inhomogeneities of the matter distribution might affect the light propagation , i.e. , the gravitational lensing effect . in fact , the angular correlation functions are significantly affected by the gravitational lensing through the magnification bias . the spatial correlations which we study in the present paper , however , are fairly insensitive to the effect ( e.g. , moessner , jain , & villumsen 1998 ) . therefore we ignore lensing effect below . the plan of the paper is as follows . in 2 , we define an estimator for the power spectrum constructed from a sample of cosmological sources on a light cone , which generalizes the conventional power spectrum defined on the constant - time hypersurface . by computing the ensemble average of the estimator , we derive the power spectrum on the light cone . a mathematical derivation is outlined in appendix a. while this rigorous procedure assumes linear theory of density perturbations in real space , the effects of nonlinear gravitational growth and redshift - space distortion are important in practice ( papers ii and iii ) . thus we incorporate those effects as described in 3 . then we apply our theoretical results to galaxy and quasar samples with the sdss spectroscopic samples specifically in mind ( 4 ) . finally 5 is devoted to discussion and conclusions . throughout this paper we use the units in which the light velocity @xmath3 is unity . in this section we describe our theoretical formulation for the power spectrum on a light - cone hypersurface which proceeds fairly parallel to paper i. for simplicity , we focus on the spatially - flat friedmann - lematre universe , whose line element is expressed in terms of the conformal time @xmath4 as @xmath5 . \label{metric}\ ] ] we normalize the scale factor to be unity at present , @xmath6 . all cosmological objects observed from redshift surveys are located on the light - cone hypersurface of the space - time ( [ metric ] ) , defined by an observer . we locate the fiducial observer at the origin of the coordinates ( @xmath7 , @xmath8 ) , and therefore an object at @xmath9 and @xmath4 on the light - cone hypersurface of the observer satisfies a simple relation @xmath10 . then the ( real - space ) position of the source on the light - cone hypersurface is specified by @xmath11 , where @xmath12 is a unit vector along the line - of - sight . in order to avoid confusion , we introduce a radial coordinate @xmath13 instead of @xmath9 , and write the metric of the three - dimensional real space as @xmath14 let us denote the number density field of the sources on the light cone by @xmath15 , which is simply related to the comoving number density of objects at a conformal time @xmath4 and at a position @xmath11 as @xmath16 introducing the mean _ observed _ ( comoving ) number density @xmath17 at time @xmath4 and the density fluctuation of luminous objects @xmath18 , we write @xmath19 . \label{defdelta}\ ] ] then equation ( [ eq : nlc1 ] ) is rewritten as @xmath20~\bigr\|_{\eta\rightarrow \eta_0-r , ~\chi\rightarrow r } . \label{nlc}\ ] ] in what follows , we define @xmath21 for convenience . note that @xmath17 is different from the mean number density of the objects @xmath22 at @xmath4 by the selection function which depends on the luminosity function of the objects and thus on the magnitude - limit of a survey ( see paper i and 4 below ) . following the conventional treatment of the power spectrum in cosmology ( e.g. , feldman , kaiser , & peacock 1994 ) , we introduce the number density field of cosmological objects on the light cone : @xmath23^{1 \over 2 } } , \label{1}\ ] ] where @xmath24 denotes the number density of the synthetic catalog without structure , which has the mean number density @xmath25 . to be specific , it satisfies @xmath26 where we neglect the poisson noise term . a power spectrum on the light cone , which can be estimated from a given survey , is written as @xmath27 where @xmath28 is the volume of a thin shell with the radius @xmath29 in the fourier space , and @xmath30 is the fourier transform of equation ( [ 1 ] ) : @xmath31 e^{i\bf k\cdot r } \over \displaystyle \left[\int d^3{\bf r } \n0lc(r)^2\right]^{1 \over 2}}~.\ ] ] finally the power spectrum on the light cone , @xmath32 , can be computed by taking the ensemble average of the above estimator ( [ eq : pkobs ] ) : @xmath33 [ n^{\rm lc}(r_2,\bft_2)-n_s^{\rm lc}(r_2,\bft_2)]\bigr > e^{i { \bf k } \cdot ( { \bf r}_1-{\bf r}_2 ) } \over \displaystyle\int d^3{\bf r } n_0^{\rm lc}(r)^2 } . \label{eq : defpklc}\end{aligned}\ ] ] for definiteness , we use a superscript lc explicitly to indicate the power spectrum of _ objects on the light cone _ throughout the paper . the power spectrum without the superscript denotes that of _ mass _ defined on the constant - time hypersurface . also subscripts @xmath34 and @xmath35 indicate those in real and redshift spaces , and subscripts lin and nl refer to those in linear and nonlinear models . assuming linear bias between density fluctuations of mass and luminous objects and also linear theory for growth of density fluctuations , we find that equation ( [ eq : defpklc ] ) reduces to the following expression : @xmath36^{-1 } \int dr_1r_1 \int dr_2 r_2 \n0lc(r_1)\n0lc(r_2 ) d_1(\eta_0-r_1)d_1(\eta_0-r_2 ) \nonumber\\ & ~ & \hspace{0.1cm}\times{1\over 2\pi k}\int^\infty_0dk_1k_1 p_{\rm r , lin}(k_1,\eta_0 ) b(k_1;\eta_0-r_1)b(k_1;\eta_0-r_2 ) \pi(r_1,r_2,k_1,k)~ , \label{plc}\end{aligned}\ ] ] where @xmath37 is the linear growth rate normalized unity at present , @xmath38 , @xmath39 is the linear bias factor , and @xmath40 with the cosine integral function @xmath41 defined as @xmath42 the derivation of equation ( [ plc ] ) from equation ( [ eq : defpklc ] ) is quite tedious but straightforward which we outline in appendix a. while equation ( [ plc ] ) for power spectrum on the light cone is exact as long as linear theory of density fluctuations and a linear bias model are adopted , it looks fairly complicated . therefore it is instructive to consider a practical approximation to equation ( [ plc ] ) . one can show that the function @xmath43 ( eq.[[defpi ] ] ) is peaked around @xmath44 and @xmath45 . in fact , if one replaces the function as @xmath46 equation ( [ plc ] ) reduces to @xmath47 with @xmath48^{-1 } { \int dr r^2 n_0^{\rm lc}(r){}^2 d_1(\eta_0-r)^2 } b(k;\eta_0-r)^2~. \label{alphad}\ ] ] in an extreme case of no evolution without biasing , i.e. , @xmath49 , @xmath50 becomes unity , and @xmath0 is equivalent to @xmath51 as expected . thus if the approximation ( [ deltaasym ] ) is justified , @xmath50 quantifies the degree of the light - cone effect on a specified survey , which sensitively depends on the observed selection function and the assumed time - evolution of linear bias . figure [ fig : pklcapprox ] compares the power spectra on a light cone , equations ( [ plc ] ) and ( [ plcb ] ) . it should be noted that @xmath0 in the exact formula becomes significantly larger than @xmath52 for @xmath53 with @xmath54 being the depth of the survey volume . this artifact is not particular to the light - cone effect , but simply originates from the fact that the power on scales comparable to , or larger than , the finite size of the survey volume can not be properly evaluated . thus except for the range @xmath55 , where the estimate of the power is fairly reliable for a given observational sample , figure [ fig : pklcapprox ] shows that the approximation ( [ plcb ] ) reproduces the exact formula very accurately . therefore in what follows , we will incorporate several important effects ( redshift - space distortion , nonlinear density and velocity evolution , and the observational selection function ) on the basis of the approximation . while the results in figure [ fig : pklcapprox ] do not include the selection function , i.e. , @xmath56=constant . is assumed , we also examined the case with the proper selection functions for galaxies and quasars ( 4 ) , and made sure that our approximation agrees with the exact formula within several percents at worst . this level of accuracy is sufficiently good in the light of the other approximations employed in describing the nonlinearity of density and velocity fields , most notably , the uncertainty of the model of bias . the results presented in the previous section are unrealistically simplified since we have neglected several important effects including the redshift - space distortion due to the peculiar velocity effect ( kaiser 1987 ) , nonlinear evolution of mass density fluctuations and of peculiar velocity dispersions ( finger - of - god ) , evolution of the bias parameter ( e.g. , fry 1996 ) , and the observational selection function which depends on a specific survey strategy . in this section we describe our modeling to the first three effects which are the necessary ingredients in completing the theoretical predictions . the selection function is discussed in the next section . strictly speaking , inclusion of these effects invalidates the exact derivation of the formula ( [ plc ] ) unfortunately . therefore we implement a _ phenomenological _ correction to these effects as described below , but we believe that this procedure is important and useful in practice , especially given the uncertainty of the bias evolution itself . the cosmological observation is possible only in redshift space . the redshift - space distortion in linear theory , first discussed by kaiser ( 1987 ) , is easily included in our formula for power spectrum on the light cone in real space . in linear theory , the direction averaged power spectrum in redshift space , @xmath57 is related to that in real space , @xmath58 , as @xmath59 p_{\rm r , lin}(k , z ) . \label{defps}\ ] ] the above relation ( kaiser 1987 ) assumes a distant - observer approximation , and scale - independent linear bias , @xmath60 , as well . then @xmath61 is defined by @xmath62 see matsubara & suto ( 1996 ) , hamilton ( 1998 ) and paper iii for extensive discussion on the redshift - space distortion . substituting equation ( [ defps ] ) in equation ( [ plc ] ) yields the expression for @xmath63 which is approximated as @xmath64 \over { \displaystyle \int dr r^2 \n0lc(r)^2 } } . \hspace{0.5 cm } \label{eq : aslin}\end{aligned}\ ] ] a similar expression for the linear redshift - space distortion on two - point correlation functions is derived in paper ii . nonlinear evolution of mass density field introduces the additional @xmath29-dependence on the light - cone power spectrum . while the full description of the nonlinear gravitational evolution of density fields is almost impossible , fairly accurate fitting formulae for @xmath65 have been obtained ( e.g. , peacock & dodds 1994,1996 ) . we attempt to correct for the nonlinear density evolution simply by adopting the fitting formulae , and then obtain an approximate expression for @xmath66 as follows : @xmath67 finally an effect of the nonlinear velocity field , _ finger - of - god _ , should be taken into account in order to make a testable theoretical prediction . again this has been extensively discussed for the power spectrum on the constant - time hypersurface ( cole , fisher & weinberg 1994 ; mo , jing , & brner 1997 ; paper iii ; magira , jing & suto 1999 ) . in particular , cole et al . ( 1994 ) proposed the following correction for the finger - of - god effect : @xmath68 p_{\rm r , nl}(k , z ) . \label{eq : psnl}\ ] ] if the velocity distribution function is approximated by an exponential model with scale - independent one - dimensional dispersion of the peculiar velocity @xmath69 , which is indicated observationally ( davis & peebles 1983 ) and also from numerical simulations ( efstathiou et al . 1988 ; ueda , itoh & suto 1993 ; magira et al . 1999 ) , the above functions are explicitly given by @xmath70 with @xmath71 . on large scales , @xmath69 can be well approximated by a fitting formula proposed by mo , jing & brner ( 1997 ) . the validity and limitation of the approximation are discussed in detail by magira et al . combining equation ( [ eq : psnl ] ) with those fitting formulae , one can compute the nonlinear power spectrum on the light cone in redshift space as @xmath72 \over p_{\rm r , lin}(k , z=0 ) { \displaystyle \int dr r^2 \n0lc(r)^2 } } . \hspace{1 cm } \label{eq : asnl}\end{aligned}\ ] ] in properly predicting the power spectra on the light cone , the selection function should be specified . in this subsection , we describe the selection functions of galaxies and quasars with the upcoming sdss spectroscopic samples in mind . for galaxies , we adopt a b - band luminosity function of the apm galaxies ( loveday et al . 1992 ) fitted to the schechter function : @xmath73 with @xmath74 , @xmath75 , and @xmath76 . then the comoving number density of galaxies at @xmath77 which are brighter than the limiting magnitude @xmath78 is given by @xmath79 , \end{aligned}\ ] ] where @xmath80 ^ 2 10^{2.2 - 0.4 b_{\rm lim}},\ ] ] and @xmath81 $ ] is the incomplete gamma function . for quasars , we compute a selection function for b - band magnitude limited samples following paper i on the basis of the luminosity function by wallington & narayan ( 1993 ; see also nakamura & suto 1997 ) . for definiteness , we consider scdm ( standard cold dark matter ) and lcdm ( lambda cold dark matter ) models , which have @xmath82 @xmath83 and @xmath84 , respectively . these sets of cosmological parameters are chosen so as to reproduce the observed cluster abundance ( kitayama & suto 1997 ) . we use the fitting formulae of peacock & dodds ( 1996 ) and mo , jing , & brner ( 1997 ) for the nonlinear power spectrum @xmath85 and the peculiar velocity dispersions @xmath86 , respectively . figure [ fig : pklcgq ] plots the theoretical predictions for power spectra on the light cone . we adopt the b - band limiting magnitude @xmath87 and @xmath88 for galaxies and quasars , respectively , so as to roughly match the sdss and 2df spectroscopic samples . for illustrative purposes , the results are divided by the corresponding linear power spectrum on the constant - time hypersurface in real space at @xmath89 , @xmath90 . for simplicity we adopt a scale - independent linear bias model of fry ( 1996 ) ; @xmath91 and @xmath92 for galaxies and quasars ( @xmath93 in eq.[[frym ] ] below ) . purely linear theory predictions , @xmath63 , are plotted in dotted lines , which use the cdm transfer function by bardeen et al.(1986 ; bbks ) and the linear redshift - space distortion ( [ eq : plcslin ] ) by kaiser ( 1987 ) . we show two different versions of @xmath94 ; both adopt the peacock & dodds ( 1996 ; pd ) nonlinear spectrum , one , plotted in dashed lines , uses linear redshift - space distortion , and the other ( solid lines ) uses the nonlinear model ( [ eq : plcsnl ] ) with the velocity dispersion formula by mo , jing & brner ( 1997 ) . in linear regime , the light - cone effect suppresses the amplitude of spectrum in real space due to the smaller fluctuation at larger @xmath77 unless the possible evolution of bias exceeds the linear growth rate of the mass fluctuations . the redshift - space distortion , on the other hand , tends to increase the observable power spectrum in redshift space . thus the overall effect of the linear redshift - space distortion on the light cone is to increase ( decrease ) the amplitude of @xmath95 for galaxy ( quasar ) samples in a @xmath29-independent manner . the situation becomes complicated in nonlinear regimes , which produces the additional @xmath29-dependent behavior . while the nonlinear evolution of density field enhances the amplitude of both @xmath66 and @xmath94 relative to their counterparts in linear theory , the suppression due to the finger - of - god is much stronger . consequently @xmath94 is smaller than @xmath96 for any @xmath29 , and in fact the suppression is stronger for larger @xmath29 . since recent numerical simulations ( mo , jing , & brner 1997 ; jing 1998 ; suto et al . 1999 ; magira et al . 1999 ) confirmed that the nonlinear fitting formulae of density and velocity fields are accurate and reliable within several percents , these conclusions are generic as long as the bias is time - independent . the bias , however , is predicted to show strong time evolution . furthermore the realistic bias is expected to be neither time - independent nor linear ( fry 1996 ; mo & white 1996 ) ; it may even not be deterministic ( tegmark & peebles 1998 ; dekel & lahav 1999 ; taruya , koyama & soda 1999 ) . we will discuss this problem in detail elsewhere , and focus on the effect of time - evolution of bias in the next subsection . as discussed in previous subsection , the most uncertain aspect of the theoretical predictions of @xmath2 is the model of bias . since it is unlikely that any reliable theoretical model for bias is established in near future , all that one can try is to explore the possible effects by adopting a simple parametric model . while several phenomenological models of evolution of bias have been proposed in the literature ( e.g. , moscardini , et al . 1998 ; matarrese , et al . 1997 ) , we adopt the following simple model : @xmath97^p } [ b_0 - 1 ] , \label{frym}\ ] ] with @xmath98 is the bias parameter at present . when the constant index @xmath99 is unity , the model reproduces the time - evolution of the lowest - order bias coefficient which is discussed by fry ( 1996 ) on the basis of the continuity equation of mass and galaxy density fields . figure [ fig : alphabias ] plots @xmath100 and @xmath101 for @xmath93 ( fry s model ) and @xmath102 ( regarded as an extreme example for comparison ) as a function of @xmath103 . the selection functions for the galaxy and quasar samples are identical to those adopted in figure [ fig : pklcgq ] . for shallow samples like the sdss galaxies , theoretical predictions are fairly insensitive to the applied model of evolution of bias ; the cosmological light - cone effect on these samples is dominated by the redshift - space distortion . for deeper samples , however , the amplitude is very sensitive to the evolution of bias ; especially when the bias evolves faster than the linear growth rate ( @xmath104 ) , the redshift - space distortion becomes relatively negligible on linear scales . in this paper we have developed a theoretical formulation to compute the power spectrum of cosmological objects on the light - cone hypersurface . on the basis of the exact formula which works in linear theory and in real space , we have obtained a useful approximate expression valid on scales less than the survey volume size ( further discussions for the validity of the approximation will be given elsewhere ; nishioka & yamamoto 1999b ) . in linear theory , the light - cone effect simply suppresses ( in general ) the amplitude of the observable power spectrum ( as in the case of the two - point correlation functions ; see papers i , ii and iii ) . then we improved the approximate formula by including the nonlinear evolution of density fields , and linear and nonlinear redshift - space distortion phenomenologically but in a manner fully consistent with the recent numerical simulations . these nonlinearities produce the additional scale - dependence in the power spectrum on the light cone compared with those defined on the constant - time hypersurface . applying this expression to the galaxy and quasar samples selected so as to match the upcoming sdss spectroscopic samples , for instance , we have quantitatively evaluated the degree of the light - cone effect . with these example , we have demonstrated that we have developed a fairly reliable and useful method to compute the power spectrum on the light cone _ given an evolution model of bias_. in fact , it is clear that the evolution of bias sensitively changes the behavior of the power spectrum on the light cone , especially for quasar samples . while our present analysis assumes the simplest bias model , linear bias , we plan to extend our formulation for the general bias model including the stochastic bias models ( dekel & lahav 1999 ; taruya , koyama , & soda 1999 ) we thank y.kojima , t.matsubara and m.sasaki for useful discussions and comments , and y.p.jing and h.magira for providing numerical routines to compute fitting formulae for nonlinear density and velocity fields . this research was supported in part by the grants - in - aid by the ministry of education , science , sports and culture of japan to resceu ( 07ce2002 ) and to k.y . ( 11640280 ) , by the supercomputer project ( no.98 - 35 , and no.99 - 52 ) of high energy accelerator research organization ( kek ) in japan , and by the inamori foundation . * references * # 1;#2;#3;#4 ; # 1 , # 2 , # 3 , # 4 # 1;#2;#3;#4 ; # 1 , # 2 ( # 3 : # 4 ) # 1;#2 ; # 1 , # 2 . # 1;#2;#3;#4;#5;#6 ; # 1 , # 2 # 3 , ( # 4 : # 5 ) , # 6 bardeen , j.m . , bond , j.r , kaiser , n. , & szalay , a.s . 1986 ; apj;304;15 ( bbks ) ; cole , s. , fisher , k. b. , & weinberg , d. h. 1995;mnras;275;515 ; davis , m. , & peebles , p.j.e . 1983;apj;267;465 ; dekel , a. , & lahav , o. 1999;apj;520;24 ; de laix , a. a. , & starkman , g. d. 1998;mnras;299;977 ; efstathiou , g. , frenk , c.s . , white , s.d.m . , & davis , m. 1988;mnras;235;715 ; feldman , h. a. , kaiser , n. , & peacock , a. a. 1994;apj;426;23 ; fry , j. n. 1996;apj;461;l65 ; hamilton , a.j.s . 1998 ; in `` the evolving universe . selected topics on large - scale structure and on the properties of galaxies '' , ( kluwer : dordrecht ) , p.185 ; jing , y.p . 1998;apj;503;l9 ; kaiser , n. 1987;mnras;227;1 ; kitayama , t. , & suto , y. 1997;apj;490;557 ; loveday , j . , peterson , b.a . , efstathiou , g. , & maddox , s.j . 1992;apj;390;338 ; magira , h. , jing , y.p . , & suto , y. 1999;apj , in press ; matarrese , s. , coles , p. , lucchin , f. , & moscardini , l. 1997 ; mnras;286;115 ; matsubara , t. , & suto , y. 1996;apj;470;l1 ; matsubara , t. , suto , y. , & szapudi , i . 1997;apj;491;l1 ; mo , h.j . , jing , y.p . , & brner , g. 1997;mnras;286;979 ; mo , h.j . , & white , s.d.m . 1996;mnras;282;347 ; moessner , r. , jain , b. , & villumsen , j. v. 1998;mnras;294;291 ; moscardini , l. , coles , p. , lucchin , f. , & matarrese , s. 1998 ; mnras;299;95 ; nakamura , t.t . , matsubara , t. , & suto , y. 1998;apj;494;13 ; nakamura , t.t . , & suto , y. 1997;prog . phys.;97;49 ; nishioka , h. , & yamamoto , k. 1999a;apj;520;426(paper ii ) ; nishioka , h. , & yamamoto , k. 1999b;in preparation ; peacock , j.a . , & dodds , s.j . 1994;mnras;267;1020 ; peacock , j.a . , & dodds , s.j . 1996;mnras;280;l19 ( pd ) ; suto , y. , magira , h. , jing , y. p. , matsubara , t. , & yamamoto , k. 1999;prog.theor.phys.suppl . ; 133;183 ( paper iii , astro - ph/9901179 ) ; taruya , a. , koyama , k. , & soda , j. 1999;apj;510;541 ; tegmark , m. , & peebles , p.j.e . 1998;apj;500;l79 ; ueda , h. , itoh , m. , & suto , y. 1993;apj;408;3 ; wallington , s. , & narayan , r. 1993;apj;403;517 ; yamamoto , k. , & suto , y. 1999;apj;517;1 ( paper i ) ; * appendix * in this appendix , we outline the derivation of equation ( [ plc ] ) from equation ( [ eq : defpklc ] ) : @xmath105 [ n^{\rm lc}(r_2,\bft_2)-n_s^{\rm lc}(r_2,\bft_2)]\bigr > e^{i { \bf k } \cdot ( { \bf r}_1-{\bf r}_2 ) } \over \displaystyle\int d^3{\bf r } n_0^{\rm lc}(r)^2 } . \hspace{0.4 cm } \label{eq : appdefpklc}\end{aligned}\ ] ] with equations ( [ nlc ] ) and ( [ 13]),the numerator of the integrand in equation ( [ eq : appdefpklc ] ) becomes @xmath106 [ n^{\rm lc}(r_2,\bft_2)-n_s^{\rm lc}(r_2,\bft_2)]\bigr > \nonumber \\ & & \hspace{3.5cm}=\n0lc(r_1 ) \n0lc ( r_2 ) \bigl<\delta(\eta_0-r_1,r_1,\bft_1)\delta(\eta_0-r_2,r_2,\bft_2 ) \bigr>. \label{3.5}\end{aligned}\ ] ] we expand the fluctuations @xmath107 in equation ( [ defdelta ] ) as @xmath108 where @xmath109 is the normalized harmonics on the flat space : @xmath110 with @xmath111 being the spherical bessel function and @xmath112 being the spherical harmonics on a unit two - sphere . to proceed further , we adopt two assumptions ; a linear bias model in fourier space : @xmath113 and linear growth of the density fluctuations : @xmath114 with @xmath37 being the linear growth rate normalized to be @xmath38 . in the above expressions , we denote the density fluctuations by @xmath115 with cdm density fluctuation specifically in mind , although not essential at all . then the power spectrum of mass density fluctuations at present , @xmath116 , is defined through @xmath117 substituting equations ( [ 6 ] ) to ( [ 10 ] ) , one can write the third term in equation ( [ 3.5 ] ) as @xmath118 using the expansion of the plane wave in terms of the spherical harmonics : @xmath119 and equation ( [ dd ] ) , equation ( [ eq : appdefpklc ] ) is explicitly written as @xmath120^{-1 } \int dr_1 r_1 ^ 2 \int d\omega_{\vec \gamma_{1 } } \int dr_2 r_1 ^ 2 \int d\omega_{\vec \gamma_{2}}\n0lc(r_1)\n0lc(r_2 ) \nonumber \\ & ~~ & \times { 2\over \pi } \int^{\infty}_0 dk_1k_1 ^ 2p_{\rm r , lin}(k_1,\eta_0 ) b(k_1;\eta_0-r_1)b(k_1;\eta_0-r_2)d_1(\eta_0-r_1)d_1(\eta_0-r_1 ) \nonumber \\ & ~~&\times \sum_{l , m}j_l(k_1r_1)j_l(k_1r_2 ) y^_{lm}(\omega_{\vec\gamma_1})y_{lm}(\omega_{\vec\gamma_2 } ) \nonumber \\ & ~~ & \times ~ 4\pi\sum_{l_1m_1}i^{l_1}j_{l_1}(kr_1 ) y_{l_1m_1}(\omega_{\vec\gamma_1})y^_{l_1m_1}(\omega_{\hat{\bf k } } ) \nonumber \\ & ~~ & \times ~ 4\pi\sum_{l_2m_2}i^{-l_2}j_{l_2}(kr_2 ) y^_{l_2m_2}(\omega_{\vec\gamma_2})y_{l_2m_2}(\omega_{\hat{\bf k } } ) . \label{16}\end{aligned}\ ] ] integrating equation ( [ 16 ] ) over @xmath121 and @xmath122 results in @xmath123^{-1 } \int dr_1r_1 ^ 2 \int dr_2 r_2 ^ 2 \n0lc(r_1)\n0lc(r_2 ) \nonumber \\ & & \hspace{1.5cm}\times{2\over \pi}\int^\infty_0dk_1k_1 ^ 2 p_{\rm r , lin}(k_1,\eta_0 ) b(k_1;\eta_0-r_1)b(k_1;\eta_0-r_2)d_1(\eta_0-r_1)d_1(\eta_0-r_2 ) \nonumber \\ & & \hspace{1.5 cm } \times(4\pi)^2\sum_{l , m}j_l(k_1r_1)j_l(k_1r_2)j_l(kr_1)j_l(kr_2 ) y_{lm}(\omega_{\hat{\bf k}})y^*_{lm}(\omega_{\hat{\bf k } } ) . \label{17}\end{aligned}\ ] ] using the mathematical formula : @xmath124 one can simplify equation ( [ 17 ] ) as follows : @xmath125^{-1 } \int dr_1r_1 ^ 2 \int dr_2 r_2 ^ 2 \n0lc(r_1)\n0lc(r_2 ) d_1(\eta_0-r_1)d_1(\eta_0-r_2 ) \nonumber \\ & & \hspace{1.5cm}\times{2\over \pi}\int^\infty_0dk_1k_1 ^ 2 p_{\rm r , lin}(k_1,\eta_0 ) b(k_1;\eta_0-r_1)b(k_1;\eta_0-r_2 ) \nonumber \\ & & \hspace{1.5cm}\times \sum_l(2l+1)j_l(k_1r_1)j_l(kr_1)j_l(k_1r_2)j_l(kr_2 ) . \label{19}\end{aligned}\ ] ] applying the formula : @xmath126p_l(x ) \label{64}\ ] ] and then @xmath127 equation ( [ 19 ] ) is written as @xmath128^{-1 } \int dr_1r_1 ^ 2 \int dr_2 r_2 ^ 2 \n0lc(r_1)\n0lc(r_2 ) d_1(\eta_0-r_1)d_1(\eta_0-r_2 ) \nonumber \\ & & \hspace{.5cm}\times{2\over \pi}\int^\infty_0dk_1k_1 ^ 2 p_{\rm r , lin}(k_1,\eta_0 ) b(k_1;\eta_0-r_1)b(k_1;\eta_0-r_2 ) \sum_l(2l+1 ) \nonumber \\ & & \hspace{.5cm}\times { 1\over 2k_1 } \int^1_{-1}dx_1(r_1 ^ 2+r_2 ^ 2 - 2r_1r_2x_1)^{-{1\over2 } } \sin[k_1(r_1 ^ 2+r_2 ^ 2 - 2r_1r_2x_1)^{1\over2}]p_l(x_1 ) \nonumber \\ & & \hspace{.5cm}\times { 1\over 2k } \int^1_{-1}dx_2(r_1 ^ 2+r_2 ^ 2 - 2r_1r_2x_2)^{-{1\over2 } } \sin[k(r_1 ^ 2+r_2 ^ 2 - 2r_1r_2x_2)^{1\over2}]p_l(x_2 ) \nonumber \\ & = & \left[\int dr r^2 \n0lc(r)^2\right]^{-1 } \int dr_1r_1 ^ 2 \int dr_2 r_2 ^ 2 \n0lc(r_1)\n0lc(r_2 ) d_1(\eta_0-r_1)d_1(\eta_0-r_2 ) \nonumber \\ & & \hspace{.5cm}\times{1\over \pi k}\int^\infty_0dk_1k_1 p_{\rm r , lin}(k_1,\eta_0 ) b(k_1;\eta_0-r_1)b(k_1;\eta_0-r_2 ) \nonumber \\ & & \hspace{.5cm}\times\int^1_{-1}dx{\sin[k_1(r_1 ^ 2+r_2 ^ 2 - 2r_1r_2x)^{1/2 } ] \sin[k(r_1 ^ 2+r_2 ^ 2 - 2r_1r_2x)^{1/2}]\over r_1 ^ 2+r_2 ^ 2 - 2r_1r_2x } . \label{70}\end{aligned}\ ] ] finally changing the variable @xmath129 to @xmath130 , equation ( [ 70 ] ) reduces to @xmath131^{-1 } \int dr_1r_1 \int dr_2 r_2 \n0lc(r_1)\n0lc(r_2 ) d_1(\eta_0-r_1)d_1(\eta_0-r_2 ) \nonumber \\ & & \hspace{0.5cm}\times{1\over \pi k}\int^\infty_0dk_1k_1 p_{\rm r , lin}(k_1,\eta_0 ) b(k_1;\eta_0-r_1)b(k_1;\eta_0-r_2 ) \int^{r_1+r_2}_{\|r_1-r_2\|}dy { \sin k_1y\sin ky \over y } . \label{72}\end{aligned}\ ] ] rewriting the last integral of the above expression in terms of the the cosine integral function : @xmath132 equation ( [ 72 ] ) yields the final expression ( [ plc ] ) , namely , @xmath131^{-1 } \int dr_1r_1 \int dr_2 r_2 \n0lc(r_1)\n0lc(r_2 ) d_1(\eta_0-r_1)d_1(\eta_0-r_2 ) \nonumber\\ & & \hspace{1.5cm}\times{1\over 2\pi k}\int^\infty_0dk_1k_1 p_{\rm r , lin}(k_1,\eta_0 ) b(k_1;\eta_0-r_1)b(k_1;\eta_0-r_2 ) \pi(r_1,r_2,k_1,k ) , \label{plca}\end{aligned}\ ] ] where @xmath133	we examine observational consequences of the cosmological light - cone effect on the power spectrum of the distribution of galaxies and quasars from upcoming redshift surveys . first we derive an expression for the power spectrum of cosmological objects in real space on a light cone , @xmath0 , which is exact in linear theory of density perturbations . next we incorporate corrections for the nonlinear density evolution and redshift - space distortion in the formula in a phenomenological manner which is consistent with recent numerical simulations . on the basis of this formula , we predict the power spectrum of galaxies and quasars on the light cone for future redshift surveys taking account of the selection function properly . we demonstrate that this formula provides a reliable and useful method to compute the power spectrum on the light cone _ given an evolution model of bias_. 0lcn^lc_0 0lcn^lc_0 hupd-9907 + resceu-16/99 + utap-328/99 + +
consider the following stochastic differential equation describing the kinetics of the proportion @xmath0 of one of two possible forms of certain proteins @xmath1 where @xmath2 is interaction coefficient of the two proteins , @xmath3 is the amplitude of the random gaussian perturbation , and @xmath4 is the standard brownian motion . the small circle @xmath5 before @xmath6 denotes the stochastic integral of stratonovich sense ( @xcite ) . there is no explicit solution to this equation , wherefore numerical computations simulating the propagation of @xmath0 is needed . one of the most well - known numerical methods for solving stochastic differential equations ( sdes ) is the euler - maruyama method , which is , however , only consistent to sdes of it type . for sdes of stratonovich type , the consistent method is the midpoint rule , which , when applied to ( [ 1.1 ] ) , takes the form : @xmath7 where @xmath8 is the uniform time step size , and @xmath9 obeying the gaussian distribution @xmath10 and independent for different @xmath11 . the mean - square convergence of this method can be proved ( @xcite ) . the deterministic midpoint rule is an a - stable method suitable for dealing with stiff equations . for example , for the deterministic version of the equation ( [ 1.1 ] ) , i.e. , the equation ( [ 1.1 ] ) with @xmath12 @xmath13 @xmath14 is an asymptotically stable solution as @xmath15 . with @xmath16 we get the linearized equation of ( [ 1.3 ] ) @xmath17 which can be very stiff as @xmath18 is large , e.g. , @xmath19 . then in this case , the original non - linear equation ( [ 1.3 ] ) is also stiff , meaning that its numerical simulation may need the choice of very small step size @xmath20 . to illustrate this , we use the euler method @xmath21 and the midpoint rule @xmath22 to solve ( [ 1.3 ] ) numerically , and observe the effect in fig . [ f1 ] . in fig . [ f1.1 ] , @xmath23 , we see that both the euler method ( blue dotted ) and the midpoint rule ( red dash - dotted ) produce significant oscillating simulation error , while in fig . [ f1.2 ] , where @xmath24 , the midpoint rule creates a reasonable numerical solution remaining close to the true solution . note that there is no explicit true solution for ( [ 1.3 ] ) , we just use the midpoint rule with a tiny step size @xmath25 to approximate the true solution ( blue solid ) . however , if the step size is small enough , e.g. @xmath26 as used in fig . [ f1.3 ] , both of the two methods give very accurate simulations which coincide visually with the true solution . in fact , as @xmath18 gets larger and larger , the needed @xmath20 for effective computation becomes smaller and smaller . this illustrates that the simulation of the equation ( [ 1.3 ] ) is sensitive to the time step size @xmath20 , due to stiffness of the equation . meanwhile , it can be seen that the midpoint rule is more stable than the euler method . for the stochastic protein kinetic equation ( [ 1.1 ] ) , @xmath14 is also a stochastically asymptotically stable solution if @xmath15 . this can be seen from its linearized stochastic differential equation with @xmath27 ( @xcite ) @xmath28 the solution of which is @xmath29 . thus the lyapunov exponent @xmath30 of ( [ 1.7 ] ) is ( @xcite ) @xmath31 due to @xmath32 with probability 1 . at the same time , similar to its deterministic counterpart ( [ 1.4 ] ) , the equation ( [ 1.7 ] ) can also be very sensitive to its simulation time step size @xmath20 as @xmath18 is large , e.g. @xmath19 . in other words , we say ( [ 1.7 ] ) is stiff for large @xmath18 . in this case , the original non - linear equation ( [ 1.1 ] ) is also stiff ( @xcite ) . this can be seen by numerical experiments , which we show in section 3 , where we see that although the stochastic midpoint rule ( [ 1.2 ] ) is applied , cautious choice of small time step size @xmath20 is still needed , which increases computational cost and accumulates round - off errors , especially for large time intervals . therefore it is meaningful to investigate effective measures to reduce the stiffness of the stochastic differential equations , such as the equation ( [ 1.1 ] ) , for efficient numerical simulation of such equations . for the purpose mentioned above , we employ the technique of variable transformation . [ lem2.1 ] for @xmath15 , the linearized stochastic differential equation ( [ 1.7 ] ) is equivalent to the following linear sde @xmath33 via the variable transformation @xmath34 * proof . * since the ordinary differential chain rule holds for sdes of stratonovich type , a direct calculation yields @xmath35\\ & = -zdt-\frac{\sigma}{1+\lambda}z\circ dw(t ) . \end{split}\ ] ] note that , for large @xmath18 , the equation ( [ 1.7 ] ) is much more stiff than its equivalent equation ( [ 2.1 ] ) , since the lyapunov exponent for ( [ 2.1 ] ) is @xmath36 , obtained in the same way as in ( [ 1.8 ] ) . we thus find a way of reducing the stiffness of ( [ 1.7 ] ) via the variable transformation ( [ 2.2 ] ) . here , @xmath15 is to guarantee that @xmath37 is the stochastically asymptotically stable solution of ( [ 1.7 ] ) , and that the transformation ( [ 2.2 ] ) is meaningful at @xmath37 . applying the same transformation to the original non - linear stochastic protein kinetic equation ( [ 1.1 ] ) , we obtain the following result . [ prop2.2 ] for @xmath15 , the stochastic protein kinetic equation ( [ 1.1 ] ) is equivalent to the following sde @xmath38 via the variable transformation @xmath39 the linearized equation of ( [ 2.4 ] ) at its stochastically asymptotically stable solution @xmath40 is the equation ( [ 2.1 ] ) . * proof . * the inverse transformation of ( [ 2.5 ] ) is @xmath41 using the differential chain rule together with ( [ 2.5])-([2.6 ] ) we obtain @xmath42\\ & = \frac{-1}{1+\lambda}[(x+\lambda xx)dt+\sigma xx\circ dw(t)]\\ & = ( -x+\frac{\lambda}{1+\lambda}x^{2+\lambda})dt-(\frac{\sigma}{1+\lambda}x-\frac{\sigma}{1+\lambda}x^{2+\lambda})\circ dw(t ) . \end{split}\ ] ] the condition @xmath15 can ensures that the transformation ( [ 2.5 ] ) and its inverse ( [ 2.6 ] ) are meaningful at @xmath14 and @xmath40 , respectively . meanwhile , it makes @xmath40 a stochastically asymptotically stable solution of ( [ 2.4 ] ) . the linearized equation of ( [ 2.4 ] ) at @xmath40 is @xmath43 which is just the equation ( [ 2.1 ] ) . + from proposition [ prop2.2 ] , it is clear that the equation ( [ 2.4 ] ) is much less stiff than the original equation ( [ 1.1 ] ) , since the linearized equation ( [ 2.1 ] ) of ( [ 2.4 ] ) is much less stiff than the linearized equation ( [ 1.7 ] ) of ( [ 1.1 ] ) for @xmath18 large ( @xcite ) . thus , for the simulation of the stochastic protein kinetic equation , we can firstly apply the stochastic midpoint rule to the transformed equation ( [ 2.4 ] ) to get @xmath44 , and then use the inverse transform ( [ 2.6 ] ) to get back to @xmath45 . in the following section , we perform numerical tests to illustrate the difference the transformation ( [ 2.5 ] ) makes , that is , different degree of stiffness of ( [ 1.1 ] ) and ( [ 2.4 ] ) , reflected in the dependence on time step size @xmath20 in the numerical simulations of them using the stochastic midpoint rule . in this section we compare the effect of numerical simulation of the stochastic protein kinetic equation based on the original equation ( [ 1.1 ] ) and its transformed equation ( [ 2.4 ] ) , respectively . we apply the stochastic midpoint rule to both equations , with varying time step sizes for @xmath19 and @xmath46 . the results are shown in fig . [ f2 ] , for which we take @xmath47 , @xmath48 , and the number of iterations in each time step for the realization of the implicit stochastic midpoint rule is 10 . + in fig . [ f2.1 ] and [ f2.2 ] , @xmath19 . as @xmath49 , the midpoint rule applied to the equation before transform , i.e. , the original equation ( [ 1.1 ] ) loses accuracy severely , as shown by the blue dotted line in fig . [ f2.1 ] , while that applied to the equation after transform , i.e. , the equation ( [ 2.4 ] ) together with the inverse transformation ( [ 2.6 ] ) produces much better simulation as illustrated by the red dash - dotted line in the same figure . however , as @xmath20 is small enough , e.g. @xmath26 , the midpoint rule applied to both equations gives good numerical results , which can be seen in fig . [ f2.2 ] . as the stiffness increases by enlarging the absolute value of @xmath2 , for example , @xmath46 , we see in fig . [ f2.3 ] that the time step size @xmath26 loses its effectiveness if the numerical simulation is performed on the original equation ( [ 1.1 ] ) without transformation , while remains valid if the numerical simulation is performed on the transformed equation ( [ 2.4 ] ) . however , for a much smaller time step size @xmath25 , both simulations work fairly well again , with the cost of much more computations , as shown by fig . [ f2.4 ] . note that , there is no explicit true solution for ( [ 1.1 ] ) or ( [ 2.4 ] ) . the solution lines ( blue solid ) are simulated by the midpoint rule approximation based on the original equation ( [ 1.1 ] ) with @xmath25 in fig . [ f2.1]- [ f2.3 ] , and @xmath50 in fig . [ f2.4 ] . to conclude , the numerical results show superiority of reducing stiffness of the sde ( [ 1.1 ] ) by variable transformation in the numerical computation of the equation . it permits larger time step sizes , and therefore reduces computational costs and increases computational robustness . the more general stochastic protein kinetic equations involve a varying parameter @xmath51 in the drift part @xmath52 in which case @xmath14 is not a stochastic stationary solution if @xmath53 . however , if we still employ the variable transformation ( [ 2.5 ] ) , obtaining the transformed equation @xmath54 instead of ( [ 2.4 ] ) , we can still observe the effect of stiffness - reduction in the numerical tests for @xmath55 $ ] , as shown by fig . [ f3 ] . as @xmath18 is small , or @xmath56 , the transformation is not recommended . in fact , this method of reducing stiffness can be generalized to a more general class of sdes , e.g. , of the form @xmath57 where the functions @xmath58 and @xmath59 are regular enough for the following discussion . this class of equations may cover many stochastic differential equation models in biology , chemistry , physics and so on . assume that there exists a constant @xmath60 such that @xmath61 , and @xmath62 is a stochastically stationary solution of ( [ 4.1 ] ) . thus , with @xmath63 , the linearized equation of ( [ 4.1 ] ) is ( @xcite ) @xmath64 if @xmath65 is large , then the equation ( [ 4.2 ] ) is stiff , and so is the original equation ( [ 4.1 ] ) . we search for a variable transformation @xmath66 such that @xmath67 which is an ordinary differential equation with solution @xmath68 for the purpose of having inverse transformation , we take @xmath69 for @xmath70 , and @xmath71 for @xmath72 . in both cases we have the following transformed equation of ( [ 4.2 ] ) @xmath73 which is less stiff than ( [ 4.2 ] ) for large @xmath65 . based on these , we perform the transformation @xmath74 for @xmath75 and @xmath76 for @xmath77 on the equation ( [ 4.1 ] ) , to get its transformed equation @xmath78 for @xmath75 and @xmath79 for @xmath77 . note that , if @xmath80 switches sign in the time interval of observation , then the transformation has no global inverse , which is a limitation of this method . to show that ( [ 4.12 ] ) or ( [ 4.13 ] ) is less stiff than the original equation ( [ 4.1 ] ) , we need to have the linearized equation of them at @xmath81 , which is just the equation ( [ 4.9 ] ) for both ( [ 4.12 ] ) and ( [ 4.13 ] ) , for @xmath82 . the less stiffness of ( [ 4.9 ] ) than ( [ 4.2 ] ) implies the less stiffness of ( [ 4.12 ] ) and ( [ 4.13 ] ) than ( [ 4.1 ] ) . note that the condition @xmath82 is also to guarantee that the drift parts of both ( [ 4.12 ] ) and ( [ 4.13 ] ) tends to @xmath83 as @xmath84 , and that @xmath62 is a stochastically asymptotically stable solution of ( [ 4.1 ] ) . the author is supported by the nnsfc ( no.11071251 , no.91130003 , no . 11471310 ) and the 2013 headmaster funds of ucas . klden p.e . , platen e. , numerical solution of stochastic differential equations , springer - verlag berlin heidelberg ( 1992 ) . mao x. , stochastic differential equations and their applications , horwood , chichester ( 1997 ) . milstein g.n . , numerical integration of stochastic differential equations , kluwer academic publishers ( 1995 ) .	the stochastic protein kinetic equations can be stiff for certain parameters , which makes their numerical simulation rely on very small time step sizes , resulting in large computational cost and accumulated round - off errors . for such situation , we provide a method of reducing stiffness of the stochastic protein kinetic equation by means of a kind of variable transformation . theoretical and numerical analysis show effectiveness of this method . its generalization to a more general class of stochastic differential equation models is also discussed . + * ams subject classification : * 65c20 , 65c30 . + * key words : * numerical methods for sdes ; stochastic protein kinetic equation ; stiffness ; midpoint rule .
* early universe physics * is the likely source of fluctuations to input into the cosmic structure formation problem ( @xcite ) . we want to measure the cmb ( and large scale structure ) response to these initial fluctuations . the goal is the lofty one of peering into the physical mechanism by which the fluctuations were generated , by learning : the statistics of the fluctuations , whether gaussian or non - gaussian ; the mode , whether adiabatic or isocurvature scalar perturbations , and whether there is a significant component in gravitational wave tensor perturbations ; the power spectra for these modes ( albeit over only 3 decades in wavenumber for cmb probes , fig . [ fig : probes ] , and fewer for large scale structure ( lss ) probes ) . * fluctuation generation mechanisms : * the contenders are ( 1 ) `` zero point '' oscillations in scalar and tensor fields that must be there in the early universe if quantum mechanics is applicable and ( 2 ) topological defects which may arise in the inevitable phase transitions expected in the early universe . * inflation : * for quantum oscillations to be important , a period of accelerated expansion seems essential , for then the comoving hubble length @xmath2 shrinks in size , freezing the time - incoherent noise into time - coherent patterns for structure formation . a major challenge for inflation models is a `` natural '' explanation of why post - inflation fluctuations should be @xmath3 in size , as revealed by cobe and also lss observations . many many variants of the basic inflation theme have been proposed , sometimes with radically different consequences for the appearance of the cmb sky , which is used in fact to highly constrain the more baroque models . a rank - ordering of inflation possibilities : ( 1 ) adiabatic curvature fluctuations with nearly uniform scalar tilt over the observable range , slightly more power to large scales ( @xmath4 ) than `` scale invariance '' ( @xmath5 ) gives , a predictable nonzero gravity wave contribution with tilt similar to the scalar one , and tiny mean curvature ( @xmath6 ) ; ( 2 ) same as ( 1 ) , but with a tiny gravity wave contribution ; ( 3 ) same as ( 1 ) but with a subdominant isocurvature component of nearly scale invariant tilt ( the case in which isocurvature dominates is ruled out by @xmath7 ) ; ( 4 ) radically broken scale invariance with weak to moderate features ( ramps , mountains , valleys ) in the fluctuation spectrum ( strong ones are largely ruled out by @xmath7 ) ; ( 5 ) radical breaking with non - gaussian features as well ; ( 6 ) `` open '' inflation , with quantum tunneling producing a negatively - curved ( hyperbolic ) space which inflates , but not so much as to flatten the mean curvature ( @xmath8 , not @xmath9 , where @xmath10 ) ; ( 7 ) quantum creation of compact hyperbolic space from `` nothing '' with volume @xmath11 which inflates , with @xmath12 , not @xmath9 , and @xmath13 of order @xmath14 ; ( 8) flat ( @xmath15 ) inflating models which are small tori of scale @xmath13 with @xmath13 a few @xmath2 in size . it is quite debatable which of the cases beyond ( 2 ) are more or less plausible , with some claims that ( 4 ) is supersymmetry - inspired , others that ( 6 ) is not as improbable as it sounds ( see cohn , these proceedings , for a nice discussion ) . it is the theorists job to push out the boundaries of the inflation idea and use the data to select what is allowed . * defects : * gradients in the disordered field energy left by a phase transition are smoothed on the growing scale over which causal communication can occur . topological knot - like or string - like field configurations disappear slowly enough that they can act as isocurvature seed perturbations to drive the growth of fluctuations in the total mass density . the inherent nonlinearity of these defects in the field implies many issues can only be answered with numerical simulations , and these are subject to the same computational - size - to - resolution limitations that plague the rest of nonlinear cosmology . much analytic progress in understanding the basic observable features of defect models is being made and renewed effort is being put into doing large enough simulations to predict cmb anisotropies in both texture and string theories . * hydro : * although hydrodynamic and radiative processes are expected to play important roles around collapsed objects and may bias the galaxy distribution relative to the mass , a global role in obscuring the early universe fluctuations by late time generation on large scales now seems unlikely . ( 95% cl ) , constrain the product @xmath16 of filling factor @xmath17 and bubble formation scale @xmath18 , to values too small for a purely hydrodynamic origin . if supernovae were responsible for the blasts , the accompanying presupernova light radiated would have been much in excess of the explosive energy ( more than a hundred - fold ) , leading to much stronger restrictions . ] * transport : * cosmological radiative transfer is on a firm theoretical footing . together with a gravity theory and the transport theory for the other fields and particles present ( baryons , hot , warm and cold dark matter , coherent fields , `` dynamical '' cosmological `` constants '' , _ etc . _ ) , we propagate initial fluctuations from the early universe through photon decoupling into the ( very ) weakly nonlinear phase , and predict _ primary anisotropies _ , those calculated using either linear perturbation theory ( for inflation - generated fluctuations ) , or , in the case of defects , linear response theory . the sources driving their development are listed in the table . * cosmic parameters : * even simple gaussian inflation - generated fluctuations for structure formation have a large number of early universe parameters we would wish to determine : power spectrum amplitudes at some normalization wavenumber @xmath19 for the modes present , @xmath20 ; `` tilt '' shape functions @xmath21 , usually chosen to be constant or with a logarithmic correction , @xmath22 . the transport problem is dependent upon physical processes , and hence on physical parameters . a partial list includes various mean energy densities @xmath23 , the hubble parameter @xmath24 , the number of relativistic neutrinos , the abundance of primordial helium , and parameters characterizing the ionization history of the universe , the compton optical depth from a reheating redshift @xmath25 to the present . for a given model , the early universe power spectrum amplitude measures are uniquely related to late - time power spectrum measures of relevance for the cmb , such as the quadrupole or bandpowers for various experiments , or to large scale structure observations , such as the _ rms _ density fluctuation level on the @xmath26 ( cluster ) scale , @xmath27 . the arena in which theory battles observation is the anisotropy power spectrum figure . [ fig : clpow ] illustrates how primary @xmath28 s vary with cosmic parameters . they are normalized to the 4-year _ dmr_(53 + 90 + 31)(a+b ) data . a `` standard '' cdm model , with @xmath29=1 , @xmath30=1 , @xmath31=50 , @xmath32=0.05 , and a 13 gyr age , is the upper solid curve . it has @xmath33 , far from the @xmath34 target value derived from cluster abundance observations . an ( almost indistinguishable ) dotted curve has the same parameters except that it includes a light neutrino with @xmath35 ( and @xmath36 ) . it has @xmath37 . the upper dashed curve is a vacuum - dominated model with @xmath31=75 and the 13 gyr age ( and @xmath38 , @xmath39 , @xmath40 ) . it has @xmath41 , which is ok for cluster abundances . an open cdm model has the @xmath28 peak shifted to higher @xmath42 ; the one shown has @xmath43 and the 13 gyr age , with @xmath30=0.33 , @xmath44=0.30 , @xmath45 , and @xmath46 . by @xmath47 , @xmath30 is down to 0.055 at this age . the lower solid curve is a cdm model with reionization at @xmath48 , and almost degenerate with it is a tilted cdm model ( @xmath29=0.95 but otherwise standard ) . even the nearly degenerate hot / cold and cdm models shown should be distinguishable by cobras / samba . * comba : * spurred on by the promise of percent - level precision in cosmic parameters from cmb satellites , a considerable fraction of the cmb theoretical community with boltzmann transport codes compared their approaches and validated the results to ensure percent - level accuracy up to @xmath49 @xcite . the arena shifted from figures like fig . [ fig : clpow ] to @xmath50 figures with tiny vertical range . we look forward to the happy day when such a relative difference figure will be used to reveal the remaining tiny residuals in the best fit theoretical model . an important goal for comba was speed , since the parameter space we wish to constrain has many dimensions . most groups have solved cosmological radiative transport by evolving a hierarchy of coupled moment equations , one for each @xmath42 . although the equations and techniques were in place prior to the cobe discovery for scalar modes , and shortly after for tensor modes , to get the high accuracy with speed has been somewhat of a challenge . there are alternatives to the moment hierarchy for the transport of photons and neutrinos . in particular the entire problem of photon transport reduces to integral equations in which the multipoles with @xmath51 are expressed as history - integrals of metric variables , photon - bunching , doppler and polarization sources , as in the table . the fastest comba - validated code uses this method ( seljak , these proceedings ) . * secondary : * secondary anisotropies , with sources listed in the table , arise from nonlinear structures . they are a nuisance foreground to be subtracted to get at the _ primary _ primary ones , but also invaluable probes of shorter - distance aspects of structure formation theories , full of important cosmological information . the effect of lensing is to smooth slightly the doppler peaks and troughs of fig . [ fig : clpow ] . @xmath52 s from quadratic nonlinearities in the gas at high redshift are concentrated at high @xmath42 , but for most viable models are expected to be a small contaminant . scattering from gas in moving clusters also has a small effect on @xmath52 , although is measurable in individual clusters . power spectra for the thermal sz effect from clusters are larger ; examples in fig . [ fig : clpow ] are for a cluster - normalized @xmath27=0.7 hot / cold hybrid model ( solid , @xmath53=0.3 ) and an @xmath29=0.8 tilted cdm model . although @xmath54 may be small , because the power for such non - gaussian sources is concentrated in hot or cold spots the signal is detectable , and often has been . @xmath52 for a typical dusty primeval galaxy model is also shown , the larger ( arbitrarily normalized ) part a shot - noise effect for galaxies with dust distributed over @xmath55 , the smaller a contribution associated with clustering . similar spectra are expected for other extragalactic point sources , radio galaxies . * phenomenology : * we have progressed from the tens of pixels of early @xmath7 experiments through the thousands for dmr and sk , and will soon be dealing with tens of thousands for long duration balloon experiments and eventually millions for the map and cobras / samba satellites . how best to analyze statistically these data sets is a subject being developed largely by cmb theorists . theorists have also taken on an increasingly phenomenological role in lss studies and many of the same techniques are being applied to both @xmath7 and lss data sets . this is opening up into a major subfield , a trend which we should strongly encourage . finding nearly optimal strategies for data projection , compression and analysis which will allow us to disentangle the primary anisotropies from the galactic and extragalactic foregrounds and from the secondary anisotropies induced by nonlinear effects will be the key to realizing the theoretically - possible precision on cosmic parameters and so to determine the winner ( and losers ) in theory space . particularly powerful is to combine results from different cmb experiments and combine these with lss and other observations . almost as important as the end - product is the application of the same techniques to probing the self - consistency and cross - consistency of experimental results . * current state : * current band - powers , shown in fig . [ fig : clpow ] , broadly follow inflation - based expectations , but may still include residual signals . consistency with the primary anisotropy frequency spectrum has been shown for dmr and for the smaller angle experiments , but over a limited range . that the level is @xmath3 provides strong support for the gravitational instability theory . to get the large scale structure of cobe - normalized fluctuations right provides encouraging support that the initial fluctuation spectrum was not far off the scale invariant form that inflation ( and defect ) models prefer . that there appears to be power at @xmath56 suggests the universe could not have reionized too early . * large scale structure : * we have always combined cmb and lss data in our quest for viable models . fig . [ fig : probes ] shows how the two are connected . as we have seen , the dmr data precisely determines @xmath27 for each model considered . for the cobe - normalized density power spectra to thread the `` eye of the needle '' associated with cluster abundances severely constrains the parameters determining them . similar constrictions arise from galaxy - galaxy and cluster - cluster clustering observations . smaller angle cmb data ( sp94 , sk95 ) are consistent with these models ( bond and jaffe , these proceedings ) , and will soon be powerful enough for the cmb by itself to offer strong selection , but this will definitely not diminish the combined lss - cmb phenomenology . * ultra - large scale structure : * the `` beyond our horizon '' land in fig . [ fig : probes ] is actually partly accessible because long waves contribute gentle gradients to our observables . constraints on such `` global parameters '' as average curvature are an example , @xmath57 , though not yet very restrictive . one may also probe whether a huge bump or deficit in power exists just beyond @xmath58 , but this has not been much explored . the remarkable non - gaussian structure predicted by stochastic inflation theory would likely be too far beyond our horizon for its influence to be felt . the bubble boundary in hyperbolic inflation models may be closer and its influence shortly after tunneling occurred could have observable consequences for the cmb . theorists have also constrained the scale of topology in simple models ; we ( pogosyan , sokolov and i ) find @xmath59 for flat 3-tori and @xmath60 for flat 1-tori from dmr ( see de oliveira - costa , these proceedings ) . a number of groups are now trying to constrain the compact hyperbolic topologies . * futures : * let us look forward to the day phenomenological theorists will have optimally - analyzed ldbs / vsa / cbi / chispi / map / cobras / samba and we know the power spectrum and cosmic parameters to wonderful precision . what will it mean ? it may not be clear . take inflation as an example . there will be attempts , undoubtedly optimal ones , to reconstruct the inflaton s potential , but all of our cmb and lss observations actually access only a very small region of the potential surface , and even this will be fuzzily determined if we allow too much freedom in parameter space . still even a fuzzy glimpse is worth the effort . most fun will be when phenomenology teaches us that non - baroque inflation and defect models fail and howling packs of theorists go hunting for the elusive generator following trails well marked by the data . 0.1 cm bond , j.r . 1996 , _ theory and observations of the cosmic background radiation _ , in `` cosmology and large scale structure '' , les houches session lx , august 1993 , ed . r. schaeffer , elsevier science press , and references therein . bertschinger , e. , bode , p. , bond , j.r . , coulson , d. , crittenden , r. , dodelson , s. , efstathiou , g. , gorski , k. , hu , w. , knox , l. , lithwick , y. , scott , d. , seljak , u. , stebbins , a. , steinhardt , p. , stompor , r. , souradeep , t. , sugiyama , n. , turok , n. , vittorio , n. , white , m. , zaldarriaga , m. 1995 , itp workshop on _ cosmic radiation backgrounds and the formation of galaxies _ , santa barbara . + @xmath61 & `` naive '' sachs - wolfe effect : gravitational potential + @xmath62 & photon bunching ( acoustic ) : @xmath63 effect , isocurvature effect + @xmath64 & linear - order thompson scattering ( doppler ) + @xmath65 & integrated sachs - wolfe effect + & subdominant anisotropic stress and polarization terms + + @xmath66 & gravity waves ( two polarizations ) + & subdominant polarization terms + + @xmath67 & linear weak lensing : 2d shear tensor + @xmath68 & rees - sciama effect : linear response to @xmath69 of nonlinear structure + @xmath70 & nonlinear thompson scattering : quadratic - order ( vishniac ) effect , + @xmath71 & `` kinematic '' sz effect ( moving cluster / galaxy ) + @xmath72 & thermal sz effect : compton cooling from nonlinear gas ( @xmath73 ) + @xmath74 & redshifted dust emission , pregalactic / protogalactic ( @xmath75 ) + + & extragalactic radio sources : falling , flat , rising + & iras sources and extrapolations to moderate @xmath76 + & galactic bremsstrahlung , synchrotron + & galactic dust : regular , cool , strange ? ? +	0.0 cm the terrain that theorists cover in this cmb golden age is described . we ponder early universe physics in quest of the fluctuation generator . we extoll the virtues of inflation and defects . we transport fields , matter and radiation into the linear ( primary anisotropies ) and nonlinear ( secondary anisotropies ) regimes . we validate our linear codes to deliver accurate predictions for experimentalists to shoot at . we struggle at the computing edge to push our nonlinear simulations from only illustrative to fully predictive . we are now phenomenologists , optimizing statistical techniques for extracting truths and their errors from current and future experiments . we begin to clean foregrounds . we join cmb experimental teams . we combine the cmb with large scale structure , galaxy and other cosmological observations in search of current concordance . the brave use all topical data . others carefully craft their prior probabilities to downweight data sets . we are always unbiased . we declare theories sick , dead , ugly . sometimes we cure them , resurrect them , rarely beautify them . our goal is to understand how all cosmic structure we see arose and what the universe is made of , and to use this to discover the laws of ultrahigh energy physics . theorists are humble , without hubris . epsf # 1@xmath0 ^ # 1#1 # 1#1 # 1#1 # 1#1 c 3mp@xmath1 m # 1to 0pt#1 canadian institute for advanced research cosmology programcanadian institute for theoretical astrophysics , toronto , ontario , canada .
on large matrices are a cornerstone of computational linear algebra . with a few exceptions such as the approximate lanczos and power methods , most algorithms used by practitioners aimed to optimize speed of computation under the constraint of obtaining an exact result . recently , spurred by the seminal paper of frieze et al . @xcite , there has been a greater interest in finding algorithms which sacrifice the precision of the result for a gain in the speed of execution . consider the low - rank approximation problem ; i.e. , finding a matrix @xmath0 of rank at most @xmath1 which approximates a given matrix @xmath2 . the best matrix @xmath0 , best in the sense that it minimizes @xmath3 for any unitarily invariant norm ( e.g. , spectral or frobenius norms ) , can be obtained by computing the singular value decomposition ( svd ) of @xmath2 . ( throughout this paper , we adopt the frobenius norm ; we use the notation @xmath0 to denote the best rank-@xmath1 approximation to @xmath2 and @xmath4 to denote an approximation to it it will be easy to avoid confusion with @xmath5 , which is used to denote the @xmath6 column of @xmath2 ) but in some instances , evaluating the svd , which scales as @xmath7 where @xmath8 is the largest dimension of @xmath2 , may be too costly . frieze et al . in @xcite showed that @xmath0 can be reasonably well approximated by computing the svd of a subset of the columns of @xmath2 only , where the columns are sampled according to their relative powers i.e . , @xmath9 , with the expected error coming from using the approximation @xmath4 instead of @xmath0 being of the form @xmath10 in subsequent papers it has been argued that the additive error term in @xmath11 may be large , and thus other sampling techniques have been introduced to obtain relative approximation error bounds ( see , e.g. , @xcite ) . in this paper , we address the sparse representation of linear operators for the approximation of matrix products . an important object that will appear in our study is the so - called nystrm method ( see section [ sec : background ] ) to find a low - rank approximation to a positive kernel . this method , familiar to numerical analysts , has nowadays found applications beyond its original field , most notably in machine learning . in previous work @xcite , we proposed an approach for low - rank approximation in such applications ; here we show that the task of finding optimal sparse representation of linear operators , in order to evaluate their product , is related to the nystrm extension of a certain positive definite kernel . we will the use this connection to derive and bound the error of two new algorithms , which our simulations indicate perform well in practice . related to our work is that of @xcite for matrix products . in that paper , drineas et al . showed that a randomized algorithm sampling columns @xmath5 and rows @xmath12 of @xmath2 and @xmath13 in proportion to their relative powers @xmath14 and @xmath15 yields an expected error of @xmath16 . notice that this bound does not involve a low - rank approximation of @xmath2 or @xmath13 . in contrast , we obtain a randomized algorithm bound in which the approximating rank @xmath1 of a kernel related to @xmath2 and @xmath13 appears explicitly . the methods mentioned above are all _ adaptive _ , in the sense that they require some knowledge about @xmath2 and @xmath13 . a very simple non - adaptive method is given by an application of the johnson - lindenstrauss lemma : it is easy to show @xcite that if @xmath17 is a @xmath18 matrix with independent unit normal entries and @xmath19 , then for @xmath20 we have that @xmath21 letting @xmath22 denote the @xmath6 row of @xmath2 and @xmath23 the @xmath24 column of @xmath13 , we observe the following element - wise relation : @xmath25 and thus we see that approximating @xmath26 by @xmath27 yields a good result with high probability . later we compare this method to our algorithms described below . the remainder of this paper is organized as follows . in section [ sec : background ] , we briefly review the nystrm method used to approximate positive definite matrices @xcite . in section [ sec : mat_approx ] , we introduce the problem of approximating a matrix product and highlight two key aspects : the issue of best subset selection and the issue of optimal rescaling . we then solve the optimal rescaling problem and analyze a randomized and a deterministic algorithm for subset selection ; we conclude with simulations and a brief discussion of algorithmic complexity . to provide context for our results , we first introduce the so - called nystrm method to approximate the eigenvectors of a symmetric positive semi - definite ( spsd ) matrix . the nystrm method , familiar in the context of finite element methods , has found many applications in machine learning and computer vision in recent years ( see , e.g. , @xcite and references therein ) . we give here a brief overview : let @xmath28 \times [ 0,1 ] \rightarrow { \mathbb{r}}$ ] be a positive semi - definite kernel and @xmath29 , @xmath30 , denote pairs of eigenvalues and eigenvectors such that @xmath31}k(x , y)f_i(y)dy=\lambda_if_i(x).\ ] ] the nystrm extension is a method to approximate the eigenvectors of @xmath32 based on a discretization of the interval @xmath33 $ ] . define the @xmath34 points @xmath35 by @xmath36 with @xmath37 , so that the @xmath35 s are evenly spaced along the interval @xmath33 $ ] . then form the _ gram matrix _ @xmath38 , which in turn is used to approximate by a finite - dimensional spectral problem @xmath39 the nystrm extension then uses these @xmath40 to give an estimate @xmath41 of the @xmath6 eigenfunction as follows : @xmath42 this method can also be applied in the context of matrices . let @xmath43 be an @xmath44 spsd matrix , partitioned as @xmath45,\ ] ] where @xmath46 and @xmath1 is typically much smaller than @xmath8 . it is then possible to approximate @xmath1 eigenvectors and eigenvalues of @xmath43 by using the eigendecomposition of @xmath47 as follows . define @xmath48 and @xmath49 with @xmath50 orthogonal and @xmath51 diagonal . the nystrm extension then tells us that an approximation for @xmath1 eigenvectors in @xmath52 is given by @xmath53.\ ] ] these approximations @xmath54 and @xmath55 in turn yield an approximation @xmath56 to @xmath43 as follows : @xmath57.\ ] ] the quality of this approximation can then be measured as the ( e.g. , frobenius ) norm of the schur complement of @xmath47 in @xmath43 : @xmath58 let @xmath59 and @xmath60 . we use the notation @xmath5 to denote the columns of @xmath2 and @xmath61 the rows of @xmath13 . we can write the product @xmath26 as the sum of rank - one matrices as follows : @xmath62 our approach to estimate the product @xmath26 , akin to model selection in statistics , will consist of keeping only a few terms in the sum of ; this entails choosing a subset of columns of @xmath2 and of rows of @xmath13 , and rescaling their outer products as appropriate . to gain some basic insight into this problem , we may consider the following two extreme cases with @xmath63 and @xmath64 . first , suppose that the vectors @xmath65 and @xmath66 are collinear . then @xmath67 , hence we can recover the product without error by only keeping one term of the sum of and rescaling it appropriately _ provided _ that we know the correlation between @xmath65 and @xmath66 . at the other extreme , if @xmath65 and @xmath66 are orthogonal , rescaling will not decrease the error no matter which term in is kept . hence we see that there are two key aspects to the problem of sparse matrix product approximation as formulated above : optimal model selection : : + which rows / columns should be retained ? optimal reweighting : : + how should these rows / columns be rescaled ? as we show below , the latter of these problems can be solved exactly for a relatively low complexity . for the former , which is combinatorial in nature and seemingly much harder to solve , we give an efficient approximation procedure . we first consider the problem of optimal reweighting , conditioned upon a choice of subset . in particular , suppose that an oracle gives us the best subset @xmath68 of cardinality @xmath1 to estimate the product @xmath26 . without loss of generality , we assume that @xmath69 . we then have the following result characterizing how well one can estimate the product @xmath26 : [ th : maintheo]let the @xmath44 spsd matrix @xmath43 be defined as @xmath70 ( i.e. , @xmath71 , where @xmath72 is the hadamard or entrywise product of matrices ) and have the partition @xmath73,\ ] ] where @xmath74 without loss of generality , and @xmath47 is the corresponding principal submatrix . then the best approximation to the product @xmath26 using the terms @xmath75 is given by @xmath76 where @xmath77 and @xmath78 moreover , if @xmath79 is the @xmath80 matrix with all entries equal to one , then the squared approximation error in frobenius norm is given by @xmath81 with @xmath82 the schur complement of @xmath47 in @xmath43 . this result tells us how well we can approximate the product @xmath26 _ granted _ that we know only a few rows / columns of @xmath2 and @xmath13 , and their correlations with the remaining rows and columns . it also allows us to characterize the best subset @xmath83 of size @xmath1 ; it is the subset that minimizes @xmath84 . given the subset @xmath83 of @xmath85 , we seek the best scaling factors @xmath86 to minimize the squared approximation error @xmath87 . we can write the squared error as @xmath88 by distributing the product and using the linearity of the trace , we get @xmath89 where we made use of the following equality : @xmath90 we now work towards rewriting in a more manageable form . first , using the fact that @xmath91 , we see that @xmath92 by combining and , we have @xmath93 similarly , using , we get after an easy computation that @xmath94 we now rewrite in a more manageable form : @xmath95 the optimal weight vector is now obtained by setting the gradient of to zero . hence we obtain @xmath96 which proves the first part of the statement . for the second part , first notice that if @xmath97 $ ] is the vector whose entries are all one , we have the following expression for @xmath98 : @xmath99[\mathbf{1 } ] .\ ] ] hence , at the optimum , the error is @xmath100^t [ q_j \ y]^t { q_j}^{-1}[q_j \ y][\mathbf{1}]\\ & = { \\| { ab } \\| } ^2- [ \mathbf{1}]^t \widetilde q [ \mathbf{1 } ] , \end{aligned}\ ] ] where we see that @xmath56 is the _ nystrm approximation of _ @xmath43 as described in section [ sec : background ] . using lemma [ lem : normhad ] below , we have @xmath101^t(q-\widetilde q)[\mathbf{1}],\ ] ] which finishes the proof of the theorem . the proof of the second part of theorem [ th : maintheo ] is based on the identity proven below : [ lem : normhad ] let @xmath2 and @xmath13 be real matrices of dimensions @xmath102 and @xmath103 , respectively , and let @xmath79 be the @xmath104 matrix with all entries equal to one . the following identity holds : @xmath105 recall that we can write the product @xmath26 as a sum of @xmath8 rank - one terms as follows : @xmath106 we thus have , by definition of the frobenius norm , that @xmath107 using the invariance of the trace with respect to cyclic permutations , the last equation yields @xmath108 and the relation is proved . having shown a solution to the optimal reweighting problem according to theorem [ th : maintheo ] , we now turn our attention to the companion problem of optimal subset selection . in order to minimize the approximation error , we have to find the subset @xmath83 whose associated schur complement @xmath109 has the lowest possible power along the one - dimensional subspace of @xmath110 spanned by the vector @xmath97 $ ] . determining the eigenvectors and eigenvalues of this schur complement , and relating them to @xmath2 and @xmath13 , is not an easy task . here we present two approximations : one based on a random choice of subsets , and an alternative `` greedy '' approach which yields a worst - case error bound . we first discuss a random oracle which outputs a subset @xmath83 with probability @xmath111 defined below . recall our earlier definition of the matrix @xmath112 according to theorem [ th : maintheo ] ; this approach is motivated by the expression of the resultant squared error , conditioned upon having chosen a subset @xmath113 , as @xmath114 . since @xmath109 is positive definite , we have that @xmath115 is larger than the largest eigenvalue of @xmath109 , and we can bound this error as follows : @xmath116 note that equality is obtained when @xmath117 , and hence this bound is tight . we have investigated in @xcite an algorithm to minimize @xmath118 , which has been shown to be effective in the context of low - rank covariance matrix approximation . returning to our random oracle , note that both @xmath119 and @xmath120 are positive definite , and thus by the schur theorem @xcite , @xmath43 is also positive definite . from this , we conclude : 1 . there exists a matrix @xmath121 such that @xmath122 ; 2 . all the principal minors @xmath123 of @xmath43 are positive . consequently , we assume here that an oracle returns a subset @xmath124 with probability @xmath125 where @xmath126 is a normalizing constant , and the second fact above ensures that this probability distribution is well defined . we may then adapt the following result from the proof of theorem 1 in @xcite : [ th : randomorac ] let @xmath127 be a positive quadratic form with eigenvalues @xmath128 . if @xmath129 is chosen with probability @xmath130 , then @xmath131 combining with leads , via jensen s inequality , to an upper bound on the average error of this approach to random subset selection : @xmath132 where @xmath133 is defined via the relation @xmath134 , and @xmath135 denotes the optimal rank-@xmath1 approximation to @xmath133 obtained by truncating its singular value decomposition . despite the appearance of the term @xmath136 in this bound , it serves to relate the resultant approximation quality to the ranks of @xmath2 and @xmath13 , a feature reinforcing the well - foundedness of the accompanying algorithm we present below . in particular , if @xmath137 , then the approximation error is zero as expected . for practical reasons , we may also wish to relate this error to the eigenvalues of @xmath2 and @xmath13 . to this end , let @xmath138 and @xmath139 be two @xmath44 matrices , @xmath140 , and let @xmath141 ( resp . @xmath142 , @xmath143 ) be the singular values of @xmath138 ( resp . @xmath139 , @xmath144 ) sorted in non - increasing order . we then have the following majorization relation @xcite : @xmath145 in particular , if @xmath146 , @xmath147 , and @xmath148 , then the singular values of @xmath43 , @xmath138 , and @xmath139 are the squares of the singular values of @xmath149 and @xmath13 respectively : @xmath150 we may then conclude from that @xmath151 although the approach presented above relies on an oracle to sample in proportion to @xmath123 , we will subsequently outline a realizable algorithm based on these results . recall that theorem [ th : maintheo ] indicates we should ensure that the diagonal terms of @xmath152 are kept as small as possible . hence , as a _ deterministic _ approximation to the optimal subset selection procedure , we may take @xmath83 such that it contains the indices of the @xmath1 largest terms @xmath153 . while yielding only a worst - case error bound , this approach has the advantage of being easily implementable ( as it does not require sampling according to @xmath123 ) ; it also appears to perform well in practice @xcite . this greedy algorithm proceeds as follows : given matrices @xmath154 and @xmath155 and a positive integer @xmath156 : 1 . set @xmath157 , @xmath158 , and take @xmath159 to be the indices of the @xmath1 largest elements of @xmath160 . set @xmath161 as @xmath162 , for @xmath163 . 3 . set @xmath164 as @xmath165 , for @xmath166 . 4 . set @xmath167 and @xmath168 for @xmath166 . return @xmath169 as an approximation to @xmath26 . since the error term is the sum of all the terms in the schur complement , we can look to bound its largest element . to this end , we have the following result : [ prop : greedy ] the largest entry in @xmath109 is smaller than the largest diagonal element of @xmath152 in . this lemma confirms that a good error - minimization strategy is to make sure that the diagonal terms of @xmath152 are as small as possible , or equivalently to take @xmath83 such that it contains the indices of the @xmath1 largest @xmath153 as per algorithm [ alg : pseudocodemult ] . matrix product approximation error using the power rescaling of applied to each of the four subset selection algorithms described in sec . [ sec : results ] ] the proof of lemma [ prop : greedy ] is based on the following set of simple results : [ lem : lardiag ] if @xmath43 is a positive definite matrix , then @xmath170 is positive and on the diagonal of @xmath43 . since @xmath43 is positive definite , we know there exists a matrix @xmath171 such that @xmath172 by the cauchy - schwartz inequality , we have @xmath173 from which we deduce that one of the following inequalities has to be satisfied : @xmath174 now if we suppose that @xmath170 is not a diagonal element , the relations of yield a contradiction and hence the largest entry of @xmath43 is on its main diagonal . the entries of @xmath109 , the schur complement of @xmath47 in @xmath43 , can be characterized explicitly according to the following formula : [ lem : crabtree ] let @xmath175 be a nonsingular leading principal submatrix of @xmath43 obtained by keeping the rows and columns with indices @xmath176 . then @xmath109 , the schur complement of @xmath47 in @xmath43 , is given element - wise by @xmath177 furthermore , it is possible to bound the diagonal entries of @xmath109 as follows : [ lem : fisch ] if @xmath47 is a positive definite matrix , then @xmath178 we are now ready to give the proof of lemma [ prop : greedy ] : the preceding two lemmas tell us that the diagonal entries of @xmath109 are bounded by @xmath179 ( i.e. , the largest diagonal element of @xmath152 , according to the partition of ) . and using lemma [ lem : lardiag ] , we know that every entry of @xmath109 is bounded by these diagonal entries . lemma [ prop : greedy ] can be further refined to give a worst - case error bound for deterministic matrix product approximation , conditioned on a choice of subset @xmath83 and the corresponding optimal reweighting procedure . appealing to the inequality of arithmetic and geometric means to further bound the elements of @xmath109 , the results of theorem [ th : maintheo ] and lemmas [ lem : lardiag][lem : fisch ] yield : @xmath180 matrix product approximation error using the optimal rescaling of theorem [ th : maintheo ] applied to the subset selection algorithms described in sec . [ sec : results ] ] we now present preliminary experimental results and discuss the computational complexity of the algorithms under consideration . three sets of experiments were performed , in which we compared the performance of four subset selection methods : a baseline uniform sampling on @xmath1-subsets ; sampling according the row / column powers @xcite ; sampling in proportion to the @xmath1-principal minors of @xmath43 according to ; and selecting greedily according to step 1 of algorithm [ alg : pseudocodemult ] . we also compared the choice of reweighting following subset selection , in one case applying the optimal reweighting of theorem [ th : maintheo ] and in the other simply reweighting according to the row / column powers ( see @xcite ) : @xmath181 to test these various experimental conditions , we drew 200 random matrices @xmath182 and @xmath183 in total , each having independent unit normal entries . we then averaged the error of the randomized algorithm over 20 trials per matrix product , and report relative error in db as @xmath184 for each test condition . in the first set of experiments , shown in figure [ fig : fig1 ] , we compare the four different algorithms for subset selection described above , applied in conjunction with a reweighting according to row / column powers . the highest - error method in this case corresponds to choosing the subset @xmath83 uniformly at random , and thus should be understood as a baseline measure of performance as a function of approximant rank @xmath1 . it can also be seen that sampling @xmath83 according to the relative powers of the row / columns of @xmath2 and @xmath13 , and sampling via a metropolis - hastings algorithm ( with independent proposal distributions taken in proportion to row / column powers ) , yield similar results , with both improving upon the baseline performance . the best results in this case are obtained by the greedy subset selection method indicated by step 1 of algorithm [ alg : pseudocodemult ] . in a second set of experiments , we followed the same procedure as above to compare subset selection procedures , but applied the optimal reweighting of theorem [ th : maintheo ] rather than a rescaling according to row / column powers . performance in this case is ( as expected ) seen to be better overall , but with the ordering of the methods unchanged . as our final experiment , we compare the method of algorithm [ alg : pseudocodemult ] ( greedy subset selection followed by optimal rescaling ) to two non - adaptive methods : choosing row / columns of @xmath2 and @xmath13 uniformly at random and rescaling according to @xmath185 , and the simple johnson - lindenstrauss random projection approach outlined in section [ sec : intro ] . these non - adaptive methods can be seen to yield significantly worse performance than algorithm [ alg : pseudocodemult ] , suggesting its potential as a practical method of selecting sparse representations of linear operators that yield low approximation errors for the resultant matrix products . we conclude with a brief discussion of the algorithmic complexity of algorithm [ alg : pseudocodemult ] . first , assume without loss of generality that @xmath186 , and recall that straightforward matrix multiplication requires @xmath187 operations , though the best algorithm known so far ( the coppersmith - winograd algorithm @xcite ) can perform this operation in @xmath188 . evaluating @xmath160 in algorithm [ alg : pseudocodemult ] requires the computation of @xmath189 inner products of @xmath190-or @xmath191-dimensional vectors , and hence requires @xmath192 operations . extracting the @xmath1 largest elements of a set of size @xmath190 , as is necessary to construct @xmath83 , can be done efficiently using a variation on the quicksort algorithm ( see @xcite ) in @xmath193 . the matrix @xmath43 is symmetric and its diagonal is a restriction of @xmath160 . hence it requires the computation of an additional @xmath194 inner products , and thus @xmath195 operations . evaluating @xmath98 requires @xmath196 operations , taking into account the fact that @xmath1 terms of the sum also appear in @xmath43 . finally , evaluating @xmath197 can be done using gaussian elimination in @xmath198 operations . hence the overall complexity is given by @xmath199 . matrix product approximation error using non - adaptive random projections ( johnson - lindenstrauss ) , non - adaptive subset selection ( uniform ) , and adaptive subset selection ( algorithm [ alg : pseudocodemult ] ) ] c. k. i. williams and m. seeger , `` using the nystrm method to speed up kernel machines , '' in _ neural information processing systems _ , t. g. dietterich , s. becker , and z. ghahramani , eds . , pp . mit press , 2001 .	thus far , sparse representations have been exploited largely in the context of robustly estimating functions in a noisy environment from a few measurements . in this context , the existence of a basis in which the signal class under consideration is sparse is used to decrease the number of necessary measurements while controlling the approximation error . in this paper , we instead focus on applications in numerical analysis , by way of sparse representations of linear operators with the objective of minimizing the number of operations needed to perform basic operations ( here , multiplication ) on these operators . we represent a linear operator by a sum of rank - one operators , and show how a sparse representation that guarantees a low approximation error for the product can be obtained from analyzing an induced quadratic form . this construction in turn yields new algorithms for computing approximate matrix products . = 1
although the high energy physics community is unanimous in considering the standard model ( sm ) to be incomplete for many and well known reasons , such as the higgs mass instability under quantum corrections or the unexplained hierarchy observed in the masses of the fermionic sector ; it is impossible to deny its enormous ( and , to some extent , unexpected ) experimental success . indeed , after years of scrutiny and new physics search , first at the lep and the tevatron and now at the lhc , the sm has not been still seriously challenged by collider experiments . the only exception today may be the @xmath0 deviation recently reported by cdf in the top - quark forward - backward asymmetry ( @xmath1)@xcite ( see also @xcite and @xcite ) . the fact that the process under question involves the top quark makes it even more interesting since , due to its mass which is near the electroweak scale , it has been suspected for a long time that the top - quark plays a special rle in nature . a natural candidate to explain a deviation in @xmath1 is an axigluon ( other alternatives beyond the sm have been considered in @xcite,@xcite,@xcite , @xcite , @xcite,@xcite,@xcite and @xcite ) . in fact , this is an expected axigluon signal @xcite . it is true that it has been argued that the present signal can not be explained by an axigluon due to other experimental constraints over it @xcite . nevertheless , bai , hewett , kaplan and rizzo have argued that a heavy and broad axigluon can be a viable explanation of the @xmath1 while remaining invisible in the dijet spectrum @xcite . in their paper , they parameterize the axigluon interaction as @xmath2 where @xmath3 is the usual qcd coupling constant a @xmath4 measures the deviation of the axigluon coupling constant from qcd . in order to evade experimental constraint it proves useful to consider that @xmath4 may take different values for light quarks ( @xmath5 ) and for the top quark ( @xmath6 ) . the analysis presented in @xcite shows that in order to have a broad axigluon it must occur that @xmath7 . additionally , in order to escape detection in contact interaction searches it must happen that @xmath8 ( this kind of non - universality has been also reported in @xcite ) . finally , @xmath9 must be negative in order to reproduce correctly the sign of @xmath1 . the construction of a model that can incorporate all these conditions , specially the different coupling constants for different generations , has proven to be difficult . as a solution , the authors of @xcite propose a two - site model but they have to include exotic vector - like quarks in order to satisfy the @xmath8 constrain . in this paper , we address the construction of a viable axigluon model not by adding new fermions but by extending the gauge sector of the model . in section [ sec : effmodel ] we show that an effective model , based on @xmath10 but with four octet fields transforming as gauge bosons , has enough structure to accommodate all the desire features of the axigluon . in section [ sec:4site ] , we show that the essential characteristics of the effective model can be obtained from a four - site model with delocalized fermions . finally , in section [ sec : summary ] , we summarize our conclusions . we start by recalling an effective extended chiral model we have previously proposed@xcite . in this model , we consider an usual @xmath11 with two pairs of vectors fields : @xmath12 and @xmath13 transforming as gauge fields of @xmath14 , and @xmath15 and @xmath16 transforming as gauge fields of @xmath17 . the gauge sector of the model ( including a non - linear sigma model sector which is introduced to produce the breaking of @xmath11 to @xmath18 ) is described by the following lagrangian : @xmath19 where @xmath20\\ g_{r\mu\nu } & = & \partial_{\mu}r_{\nu}-\partial_{\nu}r_{\mu}-ig_{r}\left[r_{\mu},\ : r_{\nu}\right]\\ \rho_{l\mu\nu } & = & \partial_{\mu}l_{\nu}-\partial_{\nu}l_{\mu}-ig'_{l}\left[l_{\mu},\ : l_{\nu}\right]\\ \rho_{r\mu\nu } & = & \partial_{\mu}r_{\nu}-\partial_{\nu}r_{\mu}-ig'_{r}\left[r_{\mu},\ : r_{\nu}\right]\\ d_{\mu}u & = & \partial_{\mu}u - ig_{l}l_{\mu}u+ig_{r}ur_{\mu}\end{aligned}\ ] ] in the unitary gauge a non - diagonal mass matrix is explicitly generated . although the resulting mass matrix can be exactly diagonalized , in order to obtain easily manipulable expressions we will assume that @xmath21 and we will write our results in the first order in @xmath22 . the eigenvalues , that is , the masses of the physical states are : @xmath23 thus , the physical spectrum is composed of a ( exactly ) massless gluon , an axigluon and two degenerate ( at this level of approximation ) heavy gluon and axigluon . the normalized mass eigenvectors can be written as : @xmath24 because we have now two fields that transform as gauge fields for each group ( @xmath12 and @xmath13 for @xmath25 and @xmath15 and @xmath16 for @xmath26 ) any combination of the form @xmath27 and @xmath28 , where @xmath29 is a new arbitrary parameter . in principle , it can take different values in the right and left sectors but , for simplicity we will assume it is the same in both cases . with this structure it is possible to construct generalized covariant derivatives . this means that the lagrangian describing the gauge interaction of quarks can be written as : @xmath30 using lagrangian ( [ eq : mylagferm ] ) and the definition of the physical fields , we can obtain the terms of the lagrangian that couple the gluon and the axigluon to quarks @xmath31 where : @xmath32 and @xmath33 interestingly , the mechanism described just above is related with fermion delocalization in deconstruction theory @xcite . now we have an axigluon with a modified coupling to quarks . nothing in the model prevents that the value of @xmath4 varies from one generation to another . this implies that in this framework we can easily satisfy the necessary conditions ( in particular , @xmath8 constrain ) for making the axigluon compatible with existing data and , at the same time , explain the @xmath1 data from cdf . the key features of the model described in this section are the presence of two pairs of vector bosons which transform like gauge fields and delocalized fermions . these properties can be naturally implemented in a four - site model . the construction of such a model is the subject of next section . our simple four site model is based on the gauge group @xmath34 and three link fields ( @xmath35 , @xmath36 and @xmath37 ) as shown in figure [ fig : group ] . we assume that the scalar fields develop vacuum expectation values ( @xmath38 and @xmath39 ) breaking down the gauge symmetry to @xmath18 which we identify with the usual color group . the lagrangian for the bosonic sector can be written as : @xmath40 where @xmath41\\ d^{\mu}\sigma_1 & = & \partial_{\mu}\sigma_1 -iga_{1\mu}\sigma_1 + ig\sigma_1 a_{2\mu } \\ d^{\mu}u & = & \partial_{\mu}u -iga_{2\mu}u + igua_{3\mu } \\ d^{\mu}\sigma_2 & = & \partial_{\mu}\sigma_2 -iga_{3\mu}\sigma_2 + ig\sigma_2 a_{4\mu}\end{aligned}\ ] ] notice that in order to simplify our equations we have assume that all groups share the same coupling constant @xmath42 . as usual , in the unitary gauge ( @xmath43 ) a non - diagonal mass matrix for the gauges bosons explicitly appears . this mass matrix can be exactly diagonalized , nevertheless , in benefice of simplicity , we work in the limit where @xmath44 . in this limit , the physical spectrum is composed of a ( exactly ) massless gluon @xmath45 , a light ( in the sense that its mass is proportional to @xmath46 ) axigluon @xmath47 and two degenerate heavy states ( @xmath48 and @xmath49 ) with masses proportional to @xmath50 . the original gauge fields can be written in terms of the physical fields as follows : @xmath51 now we turn our attention to quarks . a crucial feature of our effective model described in the previous section was the fact that the fermions coupled to a certain linear combination of gauge bosons . in the language of deconstruction theory , it correspond to `` fermion delocalization '' . the main idea of fermion delocalization @xcite is that the low energy properties of the fermions ( for example the color interaction of quarks ) originate from the contribution of many sites of the underlying gauge group . in other words , the current couples to a linear combination of gauge fields @xmath52 with the constrain @xmath53 in order to preserve gauge invariance under the whole group . as shown in figure [ fig : group ] , we decide to couple left handed quarks to @xmath54 and @xmath55 ( with weights @xmath56 and @xmath57 respectively ) and right handed quarks to @xmath58 and @xmath59 ( with weights @xmath57 and @xmath56 respectively ) . in this way , the interaction lagrangian for quarks can be written as : @xmath60+\nonumber\\ & & + gj_{r\mu}^a\left[xa^{a\mu}_4+(1-x)a^{a\mu}_2\right]\end{aligned}\ ] ] where @xmath61 and @xmath62 are the quark left - handed and right - handed currents , respectively . notice that we have chosen , for simplicity , the same delocalization parameter @xmath56 for right handed and left handed quarks . in terms of the physical fields the interaction lagrangian can be written as : @xmath63 where @xmath64 is the usual strong interaction coupling constant . our delocalization pattern has provoked the appearance of a axigluon interaction which depends on the delocalization parameter @xmath56 . as in our effective model , nothing prevents that different quark generations can be differently delocalized . in this way , we can select appropriated values of @xmath56 for the top quark and for the light quarks in order to consistently explain the @xmath1 measurements . in summary , we have shown that an extended chiral color model with extra color octet spin-1 fields has enough structure to describe an axigluon with appropriate coupling to quarks in such a way that it can be consistent with dijet and contact interactions searches and , at the same time , explain the @xmath1 anomaly reported by cdf . the fact that this effective model can be re - obtained from a four - site model , as we have shown , is interesting from the theoretical point of view since , in principle , it may be related , via dimensional deconstruction , to five dimensional models . this work has been partially supported by fondecyt grant 1070880 , the conicyt grant `` southern theoretical physics laboratory '' act-91 , the conicyt grant `` instituto para estudios avanzados en ciencia y tecnologa '' act-119 and fondecyt grant 1110167 . tgd t. aaltonen _ et al . _ [ cdf collaboration ] , `` forward - backward asymmetry in top quark production in ppbar collisions at sqrts=1.96 tev , '' phys . rev . * 101 * ( 2008 ) 202001 . [ arxiv:0806.2472 [ hep - ex ] ] . v. m. abazov _ et al . _ [ d0 collaboration ] , `` first measurement of the forward - backward charge asymmetry in top quark pair production , '' phys . * 100 * ( 2008 ) 142002 . [ arxiv:0712.0851 [ hep - ex ] ] . i. dorsner , s. fajfer , j. f. kamenik , n. kosnik , `` light colored scalars from grand unification and the forward - backward asymmetry in t t - bar production , '' phys . * d81 * ( 2010 ) 055009 . [ arxiv:0912.0972 [ hep - ph ] ] .	very recently , it has been shown by other authors that the cdf measurement of the top quark forward - backward asymmetry can be explained by means of a heavy and broad axigluon . in order to work , this mechanism needs that the axigluon coupling to the top quark must be different than the coupling to light quarks and both must be stronger than the one predicted in classical axigluon models . in this paper , we argue that this kind of axigluon can be accommodated in an extended chiral color model we proposed previously . additionally , we show that the desired features can be derived from a simple four - site model with delocalized fermions .
contemporary communications security relies on assumptions about an adversary s computational ability , a property that is uncontrolled and can rapidly change . in contrast , quantum key distribution ( qkd ) provides a mechanism to establish future - proof secure communications between two parties by exploiting fundamental quantum mechanics to generate a secret , random encryption key common to both parties @xcite . while commercial application of this technology is growing , with current technology direct qkd links on the ground can not reach distances beyond a few hundred kilometers due to optical losses @xcite . quantum repeaters @xcite promise to be an essential component of future long - distance qkd networks , but such devices are not ready for operational integration @xcite . alternatively , an orbiting satellite could serve as an untrusted node @xcite linking two ground stations simultaneously and facilitating key distribution without acquiring the key itself or as trusted node @xcite exchanging individual keys with each ground station and broadcasting the combination of two keys to allow two ground stations to establish a shared key . although the satellite is privy to the keys , by requiring only one link at a time the trusted node satellite benefits from a simpler design and , with a suitable orbit , allows key distribution between two parties located anywhere on earth . a satellite qkd link will require optical pointing mechanisms compatible with the requirements of a quantum link i.e . , capable of preserving the quantum information while maximizing the average transmission ( but not necessarily the signal consistency required for classical optical communication ) . recent demonstrations of qkd using moving _ transmitters _ @xcite imitate the case of a transmitting satellite platform , albeit without a traveling platform reaching the angular speed of a low - earth - orbit ( leo ) satellite . in contrast , a satellite - based quantum _ receiver _ is somewhat less complex to develop , and adds flexibility to utilize multiple different source technologies , including attenuated laser pulses and entangled photons @xcite . to show the viability of such an approach , we demonstrate an optical qkd link to a moving receiver platform traveling at equivalent angular speed of a satellite . our demonstration improves upon previous moving - transmitter - qkd demonstrations by employing a coincidence algorithm incorporating time - of - flight correction , implementing automated correction of polarization drifts ( including those due to the movement of the tracking system ) incurred in the fiber from the source to the transmitter , and fully conducting the post - processing steps of decoy - state qkd necessary to extract a secure key . and the length of the road traveled during the test was @xmath0 . map data : google , digitalglobe . ] our qkd receiver ( bob ) is located in the cargo area of a pickup truck driven along a road approximately from the transmitter ( alice)see fig . [ fig.overview_diagram ] . the experiment was performed at night to reduce background light contributions and avoid daytime traffic . to maintain the link we utilize wavelength beacon lasers and cmos cameras , in conjunction with a tracking algorithm operating at , at both transmitter and receiver . over the link we transmit intensity- and polarization - modulated pulses , implementing bb84 qkd with decoy states @xcite . simultaneously , the truck maintains an angular speed exceeding the maximum angular speed of a typical leo satellite platform @xcite , thereby supporting the feasibility of performing qkd from a ground station to an orbiting satellite . qkd states are generated by a weak coherent pulse ( wcp ) source located in a ground - floor laboratory in the research advancement center 1 building on the university of waterloo campus . our wcp source produces photon pulses at wavelength through sum - frequency generation ( sfg ) , combining an pulsed ti : sapphire laser with a continuous wave ( cw ) laser that is intensity and polarization modulated using fiber - based electro - optical intensity and phase modulators @xcite . the resulting pulses possess the same pulse rate and pulse width ( @xmath1 ) as the mode - locked ti : sapphire laser while having the same polarization as the cw laser . the benefit of using sfg to create pulses is that the short coherence length of the laser removes the phase correlation between the pulses ( an inherent property of a mode - locked laser ) while allowing the use of fast and stable fiber - based electro - optical modulators designed for the telecom - band . the pulses are sent to the transmitter ( fig . [ fig.transmitter_receiver ] , top ) , via an optic fiber , where they are collimated to a beam with @xmath2 waist by a 1-inch diameter , focal - length lens . the small beam waist was chosen to maintain simplicity of the transmitter ( by not requiring large optics ) , and despite the additional loss caused by diffraction , the total loss remains significantly lower than that expected in a satellite uplink . to correct for phases and rotations arising from the fiber transmission ( which vary in time due to both temperature fluctuations in the building and mechanical stress on the fiber caused by the motion of the transmitter ) , the transmitter is equipped with a polarization characterization and compensation system . we use a modified optical chopper where upon each of six open slots has been placed a polarizer aligned to pass horizontal ( h ) , vertical ( v ) , diagonal ( d ) , antidiagonal ( a ) , right- ( r ) , or left - circular ( l ) polarization . some of the closed slots of the chopper wheel were removed to improve signal transmission . as a result , of the signal pulses pass the chopper wheel unobstructed , are polarized ( either absorbed or passed , with probability dependent on the state overlap with the orientation of the polarizer ) and the remaining are blocked . a beam - splitter is used to send of the passed signals to a multi - mode fiber which is coupled to a single - photon detector . a reference signal is given from the chopper wheel at each rotation to identify the position of the wheel ( assuming a constant rotation speed between each rotation ) . the proportions of detected polarized signals are assessed each second and used to tomographically characterize the states at the transmitter . to return the measured states to the desired h , v , d , and a polarization states for qkd transmission , we optimize a theoretical model of a set of wave plates in rotation mounts , consisting of two quarter - wave plates on either side of a half - wave plate . we then apply the optimized rotations to actual wave plates mounted in motorized rotation stages . beam and includes a chopper wheel with polarizer films followed by a reflective beam splitter with a fiber coupler in the reflected port , allowing us to tomographically characterize the polarization state of some of the transmitted photons . this characterization allows us to implement , in real time using a set of motorized wave plates , a compensation to the polarization drift of the states caused by the fiber to the transmitter . only the signals that passed through the open slots of the chopper wheel are used for key generation . the receiver , which is mounted on the back of a pickup truck , implements a passive - basis - choice photon polarization measurement . ] with this , in contrast to other polarization alignment approaches @xcite , we are able to compensate , prior to the qkd signal leaving the telescope , _ any _ phase or rotation which may be induced by the fiber , directly applying a unique compensation solution at all times . our system can also compensate polarization drifts caused by other birefringent elements , such as glass and mirrors , as well as ( slowly ) moving elements such as those in a pointing mechanism , provided they are located before the polarization characterization . moreover , given additional relative orientation information , our compensation optimizer could also easily correct for misalignment between receiver and transmitter frames . for our demonstration this was not implemented , as the receiver frame was inherently sufficiently aligned to the transmitter . the receiver ( fig . [ fig.transmitter_receiver ] , bottom ) consists of a 2-inch diameter input lens ( focal length ) followed by a diameter collimating lens ( focal length ) . a beam - splitter is used to implement a passive basis choice between the h / v basis ( reflected port ) and the d / a basis ( transmitted port ) . the polarization measurement is performed using polarizing beam - splitter cubes ( pbs ) . an extra pbs ( rotated around the beam path ) is added at the reflected port of each measurement pbs to suppress noise . in each of the four measurement outputs , the light is focused into multi - mode fibers ( core diameter ) using a 1-inch diameter , focal - length collection lens and a diameter , focal - length focusing lens . the four fibers are connected to silicon avalanche photodiodes which detect the photons . the field of view of the receiver is . given some work to integrate the optical elements in a more compact form to minimize mass and volume , the simplicity of the overarching optical design of the analyzer makes it appropriate for implementation on a satellite platform . for our tests , the receiver platform is suspended on the truck with a small inner - tube and held in place with vertically sliding metal bars and elastic straps . the transmitter and the receiver are each mounted on an orthogonally oriented pair of rotation stages , one with a horizontal ( azimuth ) plane of rotation and the other with a vertical ( elevation ) plane of rotation . on each motorized platform , a camera and three beacon lasers are included , allowing for the transmitter and receiver to each use the other s beacon signal deviation , measured on their camera , to control their motors . the pointing algorithm is based on velocity estimation of the beacon spot on the camera and its deviation from a nominal center , and adjusts the velocity of the motors to match the estimated velocity plus a proportional error term to reduce the deviation . both of these corrections are based on a weighted average ( exponentially decaying ) of previous spot velocity estimates , in order to suppress high - frequency jitter . the size of the optical spot at the receiver had a measured diameter of about . following link acquisition , the total link loss at the receiver averaged , mostly due to geometric effects , with other effects including atmospheric transmittance , detector efficiency and losses in optical components accounting for only of loss . we calculate that the geometric effects contribute loss from diffraction alone ( due to the small beam waist at the transmitter output ) , with an average of of additional loss due to pointing error of the transmitter and atmospheric turbulence , and an average of loss from the receiver pointing error . the receiver pointing error contributes more loss than the transmitter because road and motor - induced vibrations of the truck under motion result in increased jitter . these vibrations would be absent on a satellite platform . during the experiment the truck is driven at a speed of , corresponding to an angular speed of , above the maximum angular speed of a leo satellite platform at altitude : at zenith . the angular deviation of the motor positions , as measured by the imaged beacon spot on the camera , is shown in fig . [ fig.pointing_offset_40km_h ] ( top ) . initial acquisition and stabilization required approximately following first acquisition of the beacon . the mean deviation after stabilization of the link was measured to be at the transmitter and at the receiver . the extra deviation at the receiver is due to the jitter from the vibrations of the truck . ( a pointing test with the truck stationary showed similar mean deviation to the measured transmitter deviation , , at both parties . ) figure [ fig.pointing_offset_40km_h ] ( bottom ) shows the angular velocity of the motors during the test . both the transmitter and the receiver averaged an azimuthal angular speed above , with the transmitter ( alice ) showing a more consistent angular speed . ) during the test . the increased deviation in the azimuthal axis at the transmitter ( alice ) is caused by the jitter in the mounting frame which can move more easily in the horizontal plane than the vertical plane . before , alice s motors are not moving because she has yet to acquire bob s beacon signal . the gray dotted line represents the maximum angular speed of a leo satellite ( ) . ] the time of flight ( fig . [ fig.time-of-flight ] ) is calculated from per - second gps coordinates recorded during the experiment , and is subtracted from detection event times assuming a linear time - of - flight variation and extrapolating the position of the truck from the most recent per - second data using the velocity , also given by the gps receiver . we observe a time - of - flight variation of around . a satellite time - of - flight variation , which could be as high as near the horizon ( approaching zero at zenith , where the count rates are highest ) , could be corrected similarly , but may require greater fidelity , e.g. a higher - order interpolation incorporating predicted orbital positions . figure [ fig.qber_and_counts_40km_h ] shows the signal quantum bit error ratio ( qber ) , the decoy qber , and the count rate during the test . the count rate increases once the link stabilizes after initial acquisition ( which takes @xmath3 ) , briefly reaching over . even after acquisition , there remains a large amount of fluctuation due to the low pointing accuracy of the receiver , which varies outside the field of view of the receiver . both the signal and decoy qber decrease as the counts increase , with a minimum measured signal qber of . intrinsic qber of the source is the primary contributor to the measured qber . we indirectly measure the intrinsic qber via the polarization compensation system , which measures the polarization states arriving at the transmitter and predicts those states after the optimized correction is applied by the wave plates . the measured qber before compensation and the predicted qber after compensation are shown in fig . [ fig.polnator_qber_40km_h ] . before compensation , the qber varies around , increasing over the duration of the test , while the predicted qber after compensation stays constant at around . we conclude that the measured qber at the receiver when the count rate is above is mostly caused by the intrinsic qber found at the transmitter , with the additional owing to background noise at the receiver . while the pre - compensation qber observed during the of data acquisition did not vary wildly , the polarization rotation induced by the motion of the optical fiber can be large and is effectively unpredictable . in addition , the observed long - term polarization drift was significant , necessitating optimization prior to each experimental attempt . these factors justify our usage of the real - time compensation system . . ] we examine the of data where the counts within each second total more than , effectively applying a signal - to - noise filter @xcite a filter with finer grain could potentially improve the results , given the high - frequency fluctuations of the count rate ( fig . [ fig.qber_and_counts_40km_h ] ) . post - processing steps , including error correction ( using low - density parity check codes ) and privacy amplification ( based on reduced - toeplitz - matrix two - universal hashes ) , are performed after the transmission , and secure key is extracted . the measured qkd parameters are reported in table [ tab.moving_receiver_qkd_results ] . ( for details of our post - processing approach , see ref . @xcite . ) from these data , we extract of secure key , not considering finite - size statistical effects . a longer link duration would be required in order to extract secure key when including finite - size statistics to extract secure key with finite - size statistics to ten standard deviations ( a common heuristic @xcite ) , a total of of data at the measured system performance would be required . .experimentally measured qkd parameters of the moving receiver test . the parameters are based on a coincidence window using data where the received counts exceeded . the small window allows us to increase the signal - to - noise ratio ( and thus reduce the qber ) at the cost of reduced raw key rate . [ cols="^,^ " , ] improvement to the intrinsic qber of the system could drastically reduce this link duration requirement . for example , in a similar system we recently demonstrated qkd at high loss , showing an asymptotic key rate of at a loss of @xcite ( above the measured of our system ) , while qbers of @xmath4 have previously been achieved in other wcp qkd sources @xcite . following the completion of the experiment , modelling @xcite showed that this qber is largely due to reduced purity , efficiency imbalance between the two sfg crystals , and deviations in the modulator phase for two polarization states . this analysis was performed using measured state data obtained from the polarization compensation system , from which we calculate the purity and fidelity of the states at the transmitter . the mean purity and fidelity of the states was found to be and respectively . our modeling shows a significant imbalance in the sfg efficiency , with one crystal showing a normalized efficiency of compared to in the other crystal , a difference of more than a factor 2 . in addition , the modulator phases for two states ( vertical and diagonal ) show notable deviations from the desired values , explaining the lower fidelities measured in these two states when compared to the other two states . another important effect that increased the intrinsic qber is the reduced purity of the state , which may have been caused by unstable phase modulators , or unstable voltages applied to the modulators , in one polarity the states with lower purity , vertical and diagonal , correspond to different applied voltage polarities than the states with higher purity , h and a @xcite . other possible causes are spreading of the polarization components due to birefringence ( given the short temporal width of the pulses ) , frequency - hopping of the laser , and noise due to background light . accounting for this reduced purity , our modeled states predict an intrinsic qber of , comparable to the predicted by the polarization compensation system . the asymptotic - limit result of our experiment is sufficient to show the viability of qkd to a moving receiver platform travelling at an angular speed of up to . of course , the useful optical contact time to a satellite ( within a single pass ) is limited to only a few hundred seconds . however , given sufficiently precise tracking , key extraction while accounting for finite - size fluctuation is nevertheless feasible under this limited time @xcite . we have thus demonstrated the feasibility of using a moving receiver platform for qkd . we constructed a custom pointing system , and used it to maintain a free - space quantum link over which bb84 decoy - state quantum signals were exchanged to a truck traveling at angular speeds consistent with a leo satellite . following timing analysis and post - processing steps , this was sufficient to generate asymptotically secure key . with reasonable improvements to the system , such as reducing the source intrinsic qber , and the addition of a second - stage fine - pointing mechanism ( both of which are largely engineering challenges with demonstrated solutions ) we expect to achieve appreciable secure key bit rates with the inclusion of finite - size statistics . such improvements will also allow us to demonstrate qkd to a moving receiver over longer distances and higher losses , completing necessary steps towards achieving long - distance satellite - mediated qkd networks . the authors would like to thank christian barna , alexander chuchin and jennifer fernick for assistance during the experimental tests . support for this work by cfi , cifar , cryptoworks21 , csa , feddev ontario , industry canada , ontario research fund and nserc is gratefully acknowledged . s.k . acknowledges support from mike & ophelia lazaridis fellowship .	technological realities limit terrestrial quantum key distribution ( qkd ) to single - link distances of a few hundred kilometers . one promising avenue for global - scale quantum communication networks is to use low - earth - orbit satellites . here we report the first demonstration of qkd from a stationary transmitter to a receiver platform traveling at an angular speed equivalent to a altitude satellite , located on a moving truck . we overcome the challenges of actively correcting beam pointing , photon polarization and time - of - flight . our system generates an asymptotic secure key at . 10 v. scarani , h. bechmann - pasquinucci , n. j. cerf , m. dusek , n. ltkenhaus , and m. peev , `` the security of practical quantum key distribution , '' * 81 * , 13011350 ( 2009 ) . r. ursin , f. tiefenbacher , t. schmitt - manderbacha , h. weier , t. scheidl , m. lindenthal , b. blauensteiner , t. jennewein , j. perdigues , p. trojek , b. oemer , m. fuerst , m. meyenburg , j. rarity , z. sodnik , c. barbieri , h. weinfurter , and a. zeilinger . , `` free - space distribution of entanglement and single photons over 144 km , '' nature physics * 3 * , 481486 ( 2007 ) . d. stucki , n. walenta , f. vannel , r. t. thew , n. gisin , h. zbinden , s. gray , c. r. towery , and s. ten , `` high rate , long - distance quantum key distribution over 250 km of ultra low loss fibres , '' new j. phys . * 11 * , 075003 ( 2009 ) . y. liu , t .- y . chen , j. wang , w .- q . cai , x. wan , l .- k . chen , j .- h . wang , s .- b . liu , h. liang , l. yang , c .- z . peng , k. chen , z .- b . chen , and j .- w . pan , `` decoy - state quantum key distribution with polarized photons over 200 km , '' * 18 * , 85878594 ( 2010 ) . z. yan , d. r. hamel , a. k. heinrichs , x. jiang , m. a. itzler , and t. jennewein , `` an ultra low noise telecom wavelength free running single photon detector using negative feedback avalanche diode , '' rev . sci . intrum . * 83 * , 073105 ( 2012 ) . b. korzh , c. c. w. lim , r. houlmann , n. gisin , m. j. li , d. nolan , b. sanguinetti , r. thew , and h. zbinden , `` provably secure and practical quantum key distribution over 307 km of optical fibre , '' nature photonics * 9 * , 163168 ( 2015 ) . c. h. bennett , g. brassard , c. crpeau , r. jozsa , a. peres , and w. k. wootters , `` teleporting an unknown quantum state via dual classical and einstein - podolsky - rosen channels , '' * 70 * , 18951899 ( 1993 ) . h .- j . briegel , w. dr , j. i. cirac , and p. zoller , `` quantum repeaters : the role of imperfect local operations in quantum communication , '' * 81 * , 5932 ( 1998 ) . c. simon , m. afzelius , j. appel , a. b. de la giroday , s. j. dewhurst , n. gisin , c. y. hu , f. jelezko , s. krll , j. h. mller , j. nunn , e. s. polzik , j. g. rarity , h. de riedmatten , w. rosenfeld , a. j. s. n. , skld , r. m. stevenson , r. thew , i. a. walmsley , m. c. weber , h. weinfurter , j. wrachtrup , and r. j. young , `` quantum memories , '' the european physical journal d * 58 * , 122 ( 2010 ) . n. sangouard , c. simon , h. de riedmatten , and n. gisin , `` quantum repeaters based on atomic ensembles and linear optics , '' * 83 * , 33 ( 2011 ) . r. j. hughes , w. t. buttler , p. g. kwiat , s. k. lamoreuax , g. l. morgan , j. e. nordholt , and c. g. peterson , `` quantum cryptography for secure satellite communications , '' in _ aerospace conference proceedings _ , ( ieee 2000 ) pp . 191200 . z. tang , r. chandrasekara , y. y. sean , c. cheng , c. wildfeuer , and a. ling , `` near - space flight of a correlated photon system , '' scientific reports * 4 * , 6366 ( 2014 ) . w. t. buttler , r. j. hughes , s. k. lamoreaux , g. l. morgan , j. e. nordholt , and c. g. peterson , `` daylight quantum key distribution over 1.6 km , '' * 84 * , 5652 ( 2000 ) . j. e. nordholt , r. j. hughes , g. l. morgan , c. g. peterson , and c. c. wipf , `` present and future free - space quantum key distribution , '' free - space laser communication technologies xiv * 4635 * , 116126 ( 2002 ) . j. g. rarity , p. r. tapster , p. m. gorman , and p. knight , `` ground to satellite secure key exchange using quantum cryptography , '' new j. phys . * 4 * , 82 ( 2002 ) . r. ursin , t. jennewein , j. kofler , j. m. perdigues , l. cacciapuoti , c. j. de matos , m. aspelmeyer , a. valencia , t. scheidl , a. acin , c. barbieri , g. bianco , c. brukner , j. capmany , s. cova , d. giggenbach , w. leeb , r. h. hadfield , r. laflamme , n. ltkenhaus , g. milburn , m. peev , t. ralph , j. rarity , r. renner , e. samain , n. solomos , w. tittel , j. p. torres , m. toyoshima , a. ortigosa - blanch , v. pruneri , p. villoresi , i. walmsley , g. weihs , h. weinfurter , m. zukowski , and a. zeilinger . , `` space - quest , experiments with quantum entanglement in space , '' europhysics news * 40 * , 2629 ( 2009 ) . p. villoresi , t. jennewein , f. tamburini , m. aspelmeyer , c. bonato , r. ursin , c. pernechele , v. luceri , g. bianco , a. zeilinger , and c. barbieri , `` experimental verification of the feasibility of a quantum channel between space and earth , '' new j. phys . * 10 * , 033038 ( 2008 ) . r. etengu , f. m. abbou , h. y. wong , a. abid , n. nortiza , and a. setharaman , `` performance comparison of bb84 and b92 satellite - based free space quantum optical communication systems in the presence of channel effects , '' j. opt . commun . * 32 * , 3747 ( 2011 ) . r. hughes and j. nordholt , `` refining quantum cryptography , '' science * 333 * , 15841586 ( 2011 ) . e. meyer - scott , z. yan , a. macdonald , j .- p . bourgoin , h. hbel , and t. jennewein , `` how to implement decoy - state quantum key distribution for a satellite uplink with 50-db channel loss , '' * 84 * , 062326 ( 2011 ) . j. yin , y. cao , s .- b . liu , g .- s . pan , j .- h . wang , t. yang , z .- p . zhang , f .- m . yang , y .- a . chen , c .- z . peng , and j .- w . pan , `` experimental quasi - single - photon transmission from satellite to earth , '' * 21 * , 2003220040 ( 2013 ) . s. nauerth , f. moll , m. rau , c. fuchs , j. horwath , s. frick , and h. weinfurter , `` air - to - ground quantum communication , '' nature photonics * 7 * , 382386 ( 2013 ) . j .- y . wang , b. yang , s .- k . liao , l. zhang , q. shen , x .- f . hu , j .- c . wu , s .- j . yang , h. jiang , y .- l . tang , b. zhong , h. liang , w .- y . liu , y .- h . hu , y .- m . huang , b. qi , j .- g . ren , g .- s . pan , j. yin , j .- j . jia , y .- a . chen , k. chen , c .- z . peng , and j .- w . pan , `` direct and full - scale experimental verifications towards ground - satellite quantum key distribution , '' nature photonics * 7 * , 387393 ( 2013 ) . j .- p . bourgoin , e. meyer - scott , b. l. higgins , b. helou , c. erven , h. hbel , b. kumar , d. hudson , i. dsouza , r. girard , r. laflamme , and t. jennewein , `` a comprehensive design and performance analysis of low earth orbit satellite quantum communication , '' new j. phys . * 15 * , 023006 ( 2013 ) . w. y. hwang , `` quantum key distribution with high loss : toward global secure communication , '' * 91 * , 057901 ( 2003 ) . z. yan , e. meyer - scott , j .- p . bourgoin , b. l. higgins , n. gigov , a. macdonald , h. hbel , and t. jennewein , `` novel high - speed polarization source for decoy - state bb84 quantum key distribution over free space and satellite links , '' j. lightwave technol . * 31 * , 13991408 ( 2013 ) . j. chen , g. wu , y. li , e. wu , and h. zeng , `` active polarization stabilization in optical fibers suitable for quantum keydistribution , '' * 15 * , 1792817936 ( 2007 ) . m. toyoshima , h. takenaka , y. shoji , y. takayama , m. takeoka , m. fujiwara , and m. sasaki , `` polarization - basis tracking scheme in satellite quantum key distribution , '' international journal of optics * 2011 * , 25415418 ( 2011 ) . m. zhang , l. zhang , j. wu , s. yang , x. wan , z. he , j. jia , d. s. citrin , and j. wang , `` detection and compensation of basis deviation in satellite - to - ground quantum communications , '' * 22 * , 98719886 ( 2014 ) . c. erven , b. heim , e. meyer - scott , j .- p . bourgoin , r. laflamme , g. weihs , and t. jennewein , `` studying free - space transmission statistics and improving free - space quantum key distribution in the turbulent atmosphere , '' new j. phys . * 14 * , 123018 ( 2012 ) . j .- p . bourgoin , n. gigov , b. l. higgins , z. yan , e. meyer - scott , a. k. khandani , n. ltkenhaus , and t. jennewein , `` experimental quantum key distribution with simulated ground - to - satellite photon losses and processing limitations , '' phys . rev . a * 92 * , 052339 ( 2015 ) . s .- h . sun , l .- m . liang , and c .- z . li , `` decoy state quantum key distribution with finite resources , '' phys . lett . a * 373 * , 25332536 ( 2009 ) . j .- p . bourgoin , `` experimental and theoretical demonstration of the feasibility of global quantum cryptography using satellites , '' ph.d . thesis , university of waterloo ( 2014 ) .
a typical object of interest in many fields is the sample covariance matrix @xmath9 of a data matrix @xmath10 , @xmath0 , @xmath11 . the matrix @xmath12 can be seen as a sample of size @xmath3 of @xmath13-dimensional data vectors . for fixed @xmath13 one can show , as @xmath3 tends to infinity , that under certain assumptions the eigenvalues of the sample covariance matrix converge to the eigenvalues of the true underlying covariance matrix @xcite . however , the assumption @xmath14 may not be justified if one has to deal with high dimensional data sets , so that it is often more suitable to assume that the dimension @xmath13 is of the same order as the sample size @xmath3 , that is @xmath15 such that @xmath16 for a symmetric matrix @xmath17 with eigenvalues @xmath18 , we denote by @xmath19 the spectral distribution of @xmath17 , where @xmath20 denotes the dirac measure located at @xmath21 . this means that @xmath22 is equal to the number of eigenvalues of @xmath17 that lie in the set @xmath23 . from now on we will call @xmath24 the sample covariance matrix . due to [ npy ] , this change of normalization can be reversed by a simple transformation of the limiting spectral distribution . for notational convenience we suppress the explicit dependence of the occurring matrices on @xmath3 and @xmath13 where this does not cause ambiguity . the distribution of gaussian sample covariance matrices of fixed size was first computed in @xcite . several years later , it was marchenko and pastur @xcite who considered the case where the random variables @xmath25 are more general i.i.d . random variables with finite second moments @xmath26 , and the number @xmath13 of variables is of the same order as the sample size @xmath3 . they showed that the empirical spectral distribution ( esd ) @xmath27 of @xmath24 converges , as @xmath28 , to a nonrandom distribution @xmath29 , called limiting spectral distribution ( lsd ) , given by @xmath30 and point mass @xmath31 if @xmath32 ; in this formula , @xmath33 . here and in the following , convergence of the esd means almost sure convergence as a random element of the space of probability measures on @xmath34 equipped with the weak topology . in particular , the eigenvalues of the sample covariance matrix of a matrix with independent entries do not converge to the eigenvalues of the true covariance matrix , which is the identity matrix and therefore only has eigenvalue one . this leads to the failure of statistics that rely on the eigenvalues of @xmath24 which have been derived under the assumption of fixed @xmath13 , and random matrix theory is a tool to correct these statistics @xcite . in the case where the true covariance matrix is not the identity matrix , the lsd can in general only be given in terms of a nonlinear equation for its stieltjes transform , which is defined by @xmath35 conversely , the distribution @xmath29 can be obtained from its stieltjes transform @xmath36 via the stieltjes perron inversion formula ( ( * ? ? ? * theorem b.8 ) ) , which states that @xmath37)=\frac{1}{\pi}\lim_{\epsilon\to 0^+ } \int_a^b \im m_{\hat{f}}(x+ { \mathrm{i}}\epsilon ) { \mathrm{d}}x.\ ] ] for all continuity points @xmath38 of @xmath39 . for a comprehensive account of random matrix theory we refer the reader to @xcite , and the references therein . our aim in this paper is to obtain a marchenko pastur type result in the case where there is dependence within the rows of @xmath12 . more precisely , for @xmath0 , the @xmath40th row of @xmath12 is given by a linear process of the form @xmath41 here , @xmath42 is an array of independent random variables that satisfies @xmath43 as well as the lindebergtype condition that , for each @xmath44 , @xmath45 clearly , [ z2 ] is satisfied if all @xmath2 are identically distributed . the novelty of our result is that we allow for dependence within the rows , and that the equation for @xmath46 is given in terms of the spectral density @xmath47,\ ] ] of the linear processes @xmath48 only , which is the fourier transform of the autocovariance function @xmath49 potential applications arise whenever data is not independent in time such that the marchenko pastur law is not a good approximation . this includes e.g. wireless communications @xcite and mathematical finance @xcite . note that a similar question is also discussed in @xcite . however , they have a different proof which relies on a moment condition to be verified . furthermore , they assume that the random variables @xmath2 are identically distributed so that the processes within the rows are independent copies of each other . more importantly , their results do not yield concrete formulas except in the ar(1 ) case and are therefore not directly applicable . in the context of free probability theory , the limiting spectral distribution of large sample covariance matrices of gaussian arma processes is investigated in @xcite . before we present the main result of this article , we explain the notation used in this article . the symbols @xmath50 , @xmath51 @xmath34 , and @xmath52 denote the sets of integers , natural , real , and complex numbers , respectively . for a matrix @xmath17 , we write @xmath53 for its transpose and @xmath54 for its trace . finally , the indicator of an expression @xmath55 is denoted by @xmath56 and defined to be one if @xmath55 is true , and zero otherwise ; for a set @xmath57 , we also write @xmath58 instead of @xmath59 . [ maintheorem ] for each @xmath60 , let @xmath61 , @xmath62 , be a linear stochastic process with continuously differentiable spectral density @xmath63 . assume that a. [ maintheoremconditionsz ] the array @xmath42 satisfies conditions , b. [ maintheoremconditionsc ] there exist positive constants @xmath64 and @xmath65 such that @xmath66 for all @xmath67 , c. [ maintheoremconditionsflevelsets ] for almost all @xmath68 , @xmath69 for at most finitely many @xmath70 $ ] , and d. [ maintheoremconditionsdfzero ] @xmath71 for almost every @xmath72 . then the empirical spectral distribution @xmath73 of @xmath6 converges , as @xmath3 tends to infinity , almost surely to a nonrandom probability distribution @xmath29 with bounded support . moreover , there exist positive numbers @xmath74 such that the stieltjes transform @xmath75 of @xmath39 is the unique mapping @xmath76 satisfying @xmath77:f(\omega)=\lambda } \frac{1}{\left\|f'(\omega)\right\| } { \mathrm{d}}\lambda}.\ ] ] the assumptions of the theorem are met , for instance , if @xmath78 is an arma or fractionally integrated arma process ; see [ examples ] for details . , as it stands , does not contain the classical marchenko pastur law as a special case . for if the entries @xmath79 of the matrix @xmath12 are i.i.d . , the corresponding spectral density @xmath63 is identically equal to the variance of @xmath80 , and thus condition is not satisfied . we therefore also present a version of [ maintheorem ] that holds if the rows of the matrix @xmath12 have a piecewise constant spectral density . [ maintheorempiecewise ] for each @xmath60 , let @xmath61 , @xmath62 , be a linear stochastic process with spectral density @xmath63 of the form @xmath81\to{\mathbb{r}}^+,\quad \omega \mapsto \sum_{j=1}^k{\alpha_j { \mathbf{1}}_{a_j}(\omega)},\quad k\in{\mathbb{n}},\ ] ] for some positive real numbers @xmath82 and a measurable partition @xmath83 of the interval @xmath84 $ ] . if conditions of [ maintheorem ] hold , then the empirical spectral distribution @xmath73 of @xmath6 converges , as @xmath28 , almost surely to a nonrandom probability distribution @xmath29 with bounded support . moreover , the stieltjes transform @xmath75 of @xmath39 is the unique mapping @xmath76 that satisfies @xmath85 where @xmath86 denotes the lebesgue measure of the set @xmath87 . in particular , if the entries of @xmath12 are i.i.d . with unit variance , one recovers the limiting spectral distribution of the marchenko pastur law . in applications one often considers processes of the form @xmath88 with mean @xmath89 . if we denote by @xmath90 the @xmath91th column of the matrix @xmath12 , and define the empirical mean by @xmath92 , then the sample covariance matrix is given by the expression @xmath93 instead of @xmath6 . however , by ( * ? ? ? * theorem a.44 ) , the subtraction of the empirical mean does not change the lsd , and thus [ maintheorem , maintheorempiecewise ] remain valid if the underlying linear process has a nonzero mean . the proof of [ maintheorem , maintheorempiecewise ] can easily be generalized to cover noncausal linear processes , which are defined as @xmath94 . for this case one obtains the same result except that the autocovariance function is now given by @xmath95 . if one considers a matrix @xmath12 which has independent linear processes in its columns instead of its rows , one obtains the same formulas as in [ maintheorem , maintheorempiecewise ] except that @xmath7 is replaced by @xmath96 . this is due to the fact that @xmath97 and @xmath98 have the same nontrivial eigenvalues . in [ proofs ] we proceed with the proofs of [ maintheorem , maintheorempiecewise ] . thereafter we present some interesting examples in [ examples ] . in this section we present our proofs of [ maintheorem , maintheorempiecewise ] . dealing with infiniteorder moving average processes directly is dfficult , and we therefore first prove a variant of these theorems for the truncated processes @xmath99 . we define the @xmath100 matrix @xmath101 , @xmath0 , @xmath11 . [ maintheoremmod ] under the assumptions of [ maintheorem ] ( [ maintheorempiecewise ] ) , the empirical spectral distribution of the sample covariance matrix of the truncated process @xmath102 converges , as @xmath3 tends to infinity , to a deterministic distribution with bounded support . its stieltjes transform is uniquely determined by [ eq - stieltjes ] ( [ eq - stieltjespiecewise ] ) . the proof starts from the observation that one can write @xmath103 , where @xmath104 , @xmath0 , @xmath105 , and @xmath106 in particular , @xmath107 . in order to prove convergence of the empirical spectral distribution @xmath108 and to obtain a characterization of the limiting distribution , it suffices , by ( * ? ? ? * theorem 1 ) , to prove that the spectral distribution @xmath109 of @xmath110 converges to a nontrivial limiting distribution . this will be done in [ lemma - lsdhh ] , where the lsd of @xmath110 is shown to be @xmath111 ; the distribution @xmath112 is computed in [ lemma - densitygamma ] if we impose the assumptions of [ maintheorem ] , respectively in [ lemma - densitygammapiecewise ] if we impose the assumptions of [ maintheorempiecewise ] . inserting this expression for @xmath113 into equation ( 1.2 ) of @xcite shows that the esd @xmath108 converges , as @xmath28 , almost surely to a deterministic distribution , which is determined by the requirement that its stieltjes transform @xmath114 satisfies @xmath115 using the explicit formulas of @xmath112 computed in [ lemma - densitygamma , lemma - densitygammapiecewise ] , one obtains [ eq - stieltjes , eq - stieltjespiecewise ] . uniqueness of a mapping @xmath116 solving [ eq - stieltjescharacgamma ] was shown in @xcite . we complete the proof by arguing that the lsd of @xmath117 has bounded support . for this it is enough , by ( * ? ? ? * theorem 6.3 ) , to show that the spectral norm of @xmath110 is bounded in @xmath3 , which is also done in [ lemma - lsdhh ] . [ lemma - lsdhh ] let @xmath118 be the matrix appearing in [ eq - defh ] , and assume that there exist positive constants @xmath119 such that @xmath66 ( assumption of [ maintheorem ] ) . then the spectral norm of the matrix @xmath110 is bounded in @xmath3 . if , moreover , the spectral distribution of the toeplitz matrix @xmath120 converges weakly to some limiting distribution @xmath121 , then the spectral distribution @xmath109 converges weakly , as @xmath28 , to @xmath122 . we first introduce the notation @xmath123 as well as the block decomposition @xmath124 $ ] , @xmath125 . we prove the second part of the lemma first . there are several ways to show that the spectral distributions of two sequences of matrices converge to the same limit . in our case it is convenient to use ( * ? ? ? * corollary a.41 ) which states that two sequences @xmath126 and @xmath127 , either of whose empirical spectral distribution converges , have the same limiting spectral distribution if @xmath128 converges to zero as @xmath3 tends to infinity . we shall employ this result twice : first to show that the lsds of @xmath129 and @xmath130 agree , and then to prove equality of the lsds of @xmath131 and @xmath132 . let @xmath133 ; a direct calculation shows that @xmath134 $ ] , and we will consider each of the two terms in turn . from the definition of @xmath135 it follows that the @xmath136th entry of @xmath137 is given by @xmath138 . the trace of the square of the upper left block of @xmath137 therefore satisfies @xmath139 ^ 2\\ { \leqslant } & \sum_{i , j , k , l=1}^n{\|c_{i+k-1}\|\|c_{j+k-1}\|\|c_{i+l-1}\|\|c_{j+l-1}\|}\\ { \leqslant } & c^4 \sum_{i , j , k , l=2}^{n+1}{i^{-1-\delta}j^{-1-\delta}l^{-1-\delta}k^{-1-\delta}}\\ < & \left[c\zeta(1+\delta)\right]^4<\infty,\end{aligned}\ ] ] where @xmath140 denotes the riemann zeta function . as a consequence , the limit of @xmath141 as @xmath3 tends to infinity is zero . similarly , we obtain for the trace of the square of the off - diagonal block of @xmath137 the bound @xmath142 ^ 2\\ { \leqslant } & \sum_{i=1}^n\sum_{j=1}^{n}\sum_{k = j}^{n - i+1}\sum_{l = j}^{n - i+1}{c_{i+k-1}c_{k - j}c_{i+l-1}c_{l - j}}\\ { \leqslant } & \sum_{i=1}^n\sum_{j=1}^n\sum_{r=0}^n\sum_{s=0}^n{\|c_{i+r+j-1}\|\|c_r\|\|c_{s+j-1}\|\|c_s\|}\\ { \leqslant } & c^4 \sum_{i , j , r , s=1}^{n+1 } { i^{-1-\delta } r^{-1-\delta } s^{-1-\delta } j^{-1-\delta}}\\ < & \left[c\zeta(1+\delta)\right]^4<\infty,\end{aligned}\ ] ] which shows that the limit of @xmath143 is zero . it follows that @xmath144 , as defined in [ eq - deltahh ] , converges to zero as @xmath3 goes to infinity , and therefore that the lsds of @xmath137 and @xmath145 coincide . the latter distribution is clearly given by @xmath146 , and we show next that the lsd of @xmath131 agrees with the lsd of @xmath120 . as before it suffices to show , by ( * ? ? ? * corollary a.41 ) , that @xmath147 converges to zero as @xmath3 tends to infinity . it follows from the definitions of @xmath137 and @xmath132 that @xmath148 can be estimated as @xmath149 ^ 2\\ = & \sum_{i , j=1}^n\left[\sum_{k=\max{(i , j)}}^n{c_{k - i}c_{k - j}}-\sum_{k=\max{(i , j)}}^\infty{c_{k - i}c_{k - j}}\right]^2\\ = & \sum_{i , j=1}^n\sum_{k , l=1}^\infty{c_{k+i-1}c_{k+j-1}c_{l+i-1}c_{l+j-1}}\\ { \leqslant } & c^4\sum_{i , j=2}^{n+1}\sum_{k , l=2}^\infty{i^{-1-\delta}j^{-1-\delta}k^{-1-\delta}l^{-1-\delta } } < \left[c\zeta(1+\delta)\right]^4<\infty.\end{aligned}\ ] ] consequently , @xmath150 converges to zero as @xmath3 goes to infinity , and it follows that @xmath151 . in order to show that the spectral norm of @xmath129 is bounded in @xmath3 , we use gerschgorin s circle theorem ( ( * ? ? ? * theorem 2 ) ) , which states that every eigenvalue of @xmath137 lies in at least one of the balls @xmath152 with centre @xmath153 and radius @xmath154 , @xmath155 , where the radii @xmath154 are defined as @xmath156 . we first note that the centres @xmath157 satisfy @xmath158 ^ 2<\infty.\ ] ] to obtain a uniform bound for the radii @xmath154 we first assume that @xmath159 . then @xmath160 ^ 2<\infty.\end{aligned}\ ] ] similarly we find that , for @xmath161 , @xmath162 ^ 2\end{aligned}\ ] ] is bounded , which completes the proof . in the following two lemmas , we argue that the distribution @xmath112 exists and we prove explicit formulas for it in the case that the assumptions of [ maintheorem ] or [ maintheorempiecewise ] are satisfied . [ lemma - densitygamma ] let @xmath163 be a sequence of real numbers , @xmath164 , and @xmath165 . under the assumptions of [ maintheorem ] it holds that the spectral distribution @xmath166 of @xmath120 converges weakly , as @xmath28 , to an absolutely continuous distribution @xmath112 with bounded support and density @xmath167 we first note that under assumption of [ maintheorem ] the autocovariance function @xmath168 is absolutely summable because @xmath169 ^ 2<\infty.\ ] ] szeg s first convergence theorem ( @xcite and ( * ? ? ? * corollary 4.1 ) ) then implies that @xmath112 exists , and that the cumulative distribution function of the eigenvalues of the toeplitz matrix @xmath132 associated with the sequence @xmath170 is given by @xmath171:f(\omega){\leqslant}\lambda\}),\ ] ] for all @xmath172 such that the level sets @xmath173:f(\omega)=\lambda\}$ ] have lebesgue measure zero . by assumption of [ maintheorem ] , [ eq - cdfgamma ] holds for almost all @xmath172 . in order to prove that the lsd @xmath121 is absolutely continuous with respect to the lebesgue measure , it suffices to prove that the cumulative distribution function @xmath174 is differentiable almost everywhere . clearly , for @xmath175 , @xmath176:\lambda < f(\omega){\leqslant}\lambda+\delta\lambda\}).\ ] ] due to assumption of [ maintheorem ] , the set of all @xmath68 such that the set @xmath177:f(\omega)=\lambda \textnormal { and } f'(\omega)=0\}$ ] is nonempty is a lebesgue null - set . hence it is enough to consider only @xmath172 for which this set is empty . let @xmath178 be the preimage of @xmath172 , which is a finite set by assumption . the implicit function theorem then asserts that , for every @xmath179 , there exists an open interval @xmath180 around @xmath72 such that @xmath63 restricted to @xmath180 is invertible . it is no restriction to assume that these @xmath180 are disjoint . by choosing @xmath181 sufficiently small it can be ensured that the interval @xmath182 $ ] is contained in @xmath183 , and from the continuity of @xmath63 it follows that outside of @xmath184 , the values of @xmath63 are bounded away from @xmath172 , so that @xmath185\\ = & \frac{1}{2\pi}\lim_{\delta\lambda\to 0}\frac{1}{\delta\lambda}{\operatorname{leb}}\left(\bigcup_{\omega\in f^{-1}(\lambda)}\{\omega'\in i_\omega : \lambda < f(\omega'){\leqslant}\lambda+\delta\lambda\}\right)\\ = & \frac{1}{2\pi}\sum_{\omega\in f^{-1}(\lambda)}\lim_{\delta\lambda\to 0}\frac{1}{\delta\lambda}{\operatorname{leb}}\left(\{\omega'\in i_\omega : \lambda < f(\omega'){\leqslant}\lambda+\delta\lambda\}\right).\end{aligned}\ ] ] in order to further simplify this expression , we denote the local inverse functions by @xmath186 $ ] . observing that the lebesgue measure of an interval is given by its length , and that the derivatives of @xmath187 are given by the inverse of the derivative of @xmath63 , it follows that @xmath188 = & \frac{1}{2\pi}\sum_{\omega\in f^{-1}(\lambda)}\lim_{\delta\lambda\to 0}\frac{1}{\delta\lambda}\left\|f_\omega^{-1}(\lambda+\delta\lambda)-f_\omega^{-1}(\lambda)\right\|\\ = & \frac{1}{2\pi}\sum_{\omega\in f^{-1}(\lambda)}\left\|\frac{{\mathrm{d}}}{{\mathrm{d}}\lambda}f_\omega^{-1}(\lambda)\right\|\\ = & \frac{1}{2\pi}\sum_{\omega\in f^{-1}(\lambda)}\frac{1}{\left\|f'(\omega)\right\|}.\end{aligned}\ ] ] this shows that @xmath174 is differentiable almost everywhere with derivative @xmath189 . it remains to argue that the support of @xmath112 is bounded . the absolute summability of @xmath190 implies boundedness of its fourier transform @xmath63 . the claim then follows from [ eq - cdfgamma ] , which shows that the support of @xmath191 is equal to the range of @xmath63 . [ lemma - densitygammapiecewise ] let @xmath192 be the piecewise constant spectral density of the linear process @xmath193 , and denote the corresponding autocovariance function by @xmath194 . under the assumptions of [ maintheorempiecewise ] it holds that the spectral distribution @xmath166 of @xmath120 converges weakly , as @xmath28 , to the distribution @xmath195 . without loss of generality we may assume that @xmath196 . as in the proof of [ lemma - densitygamma ] one sees that @xmath112 exists , and that @xmath197 is given by @xmath198:f(\omega){\leqslant}\lambda\}),\quad\forall \lambda\in[0,2\pi]\backslash \bigcup_{j=1}^k \{\alpha_j\}.\ ] ] the special structure of @xmath63 thus implies that @xmath199 , where @xmath200 is the largest integer such that @xmath201 . since @xmath174 must be rightcontinuous , this formula holds for all @xmath172 in the interval @xmath84 $ ] . it is easy to see that the function @xmath174 is the cumulative distribution function of the discrete measure @xmath202 , which completes the proof . of theorems and it is only left to show that the truncation performed in [ maintheoremmod ] does not alter the lsd , i.e. that the difference of @xmath73 and @xmath203 converges to zero almost surely . by ( * corollary a.42 ) , this means that we have to show that @xmath204 converges to zero . to this end we show that @xmath205 has a limit , and that @xmath206 converges to zero , both almost surely . by the definition of @xmath12 and @xmath207 we have @xmath208 we shall prove that the variances of @xmath206 are summable . for this purpose we need the following two estimates which are implied by the cauchy schwarz inequality , the assumption that @xmath209 is finite , and the assumed absolute summability of the coefficients @xmath163 : [ eq - fubini ] @xmath210 @xmath211 therefore we can , by fubini s theorem , interchange expectation and summation to bound the variance of @xmath206 as @xmath212 considering separately the terms where @xmath213 and @xmath214 , we can write @xmath215 for the expectation in the first sum not to be zero , @xmath216 must equal @xmath217 and @xmath218 must equal @xmath219 , in which case its value is unity . the expectation in the second term can always be bounded by @xmath220 , so that we obtain @xmath221 due to [ npy ] and the assumed polynomial decay of @xmath222 there exists a constant @xmath223 such that the right hand side is bounded by @xmath224 , which implies that @xmath225 and therefore , by the first borel cantelli lemma , that @xmath206 converges to a constant almost surely . in order to show that this constant is zero , it suffices to shows that the expectation of @xmath206 converges to zero . since @xmath226 , and the @xmath2 are independent , one sees , using [ eq - fubini1 ] and again fubini s theorem , that @xmath227 , which converges to zero because the @xmath228 are squaresummable . we now consider factor @xmath205 of expression and define @xmath229 . then @xmath230 because of @xmath231 and similarly @xmath232 , we have that @xmath233\notag\\ = & { \underbrace}{\sum_{i=1}^p \sum_{t=1}^n \sum_{k = n+1}^\infty \sum_{m = n+1}^\infty c_k c_m z_{i , t - k } z_{i , t - m}}_{={\mathrm{ii}}\to 0\text { a.s.}}\notag\\ \label{eq - trdelta } & + 2\sum_{i=1}^p \sum_{t=1}^n \sum_{k = n+1}^\infty \sum_{m=1}^n c_k c_m z_{i , t - k } z_{i , t - m}.\end{aligned}\ ] ] allows us to apply fubini s theorem to compute the variance of the second term in the previous display as @xmath234 which is , by the same reasoning as we did for @xmath206 , bounded by @xmath235 for some positive constant @xmath223 . clearly , this is summable in @xmath3 . having , by [ eq - fubini1 ] , expected value zero , the second term of [ eq - trdelta ] and , therefore , also @xmath236 both converge to zero almost surely . thus , we only have to look at the contribution of @xmath237 in expression . from [ maintheoremmod ] we know that @xmath203 converges almost surely weakly to some nonrandom distribution @xmath39 with bounded support hence , denoting by @xmath18 the eigenvalues of @xmath238 , @xmath239 almost surely . it follows that , in [ eq - iandii ] , factor @xmath205 is bounded , and factor @xmath206 converges to zero , and so the proof of [ maintheorem , maintheorempiecewise ] is complete . for several classes of widely employed linear processes , [ maintheorem ] can be used to obtain an explicit description of the limiting spectral distribution . in this section we consider the class of autoregressive moving average ( arma ) processes as well as fractionally integrated arma models . the distributions we obtain in the case of ar(1 ) and ma(1 ) processes can be interpreted as oneparameter deformations of the classical marchenko pastur law . given polynomials @xmath240 and @xmath241 , an arma(p , q ) process @xmath242 with autoregressive polynomial @xmath243 and moving average polynomial @xmath244 is defined as the stationary solution to the stochastic difference equation @xmath245 if the zeros of @xmath243 lie outside the closed unit disk , it is well known that @xmath242 has an infiniteorder moving average representation @xmath193 , where @xmath246 are the coefficients in the power series expansion of @xmath247 around zero . it is also known ( @xcite ) that there exist positive constants @xmath248 and @xmath223 such that @xmath249 , so that assumption of [ maintheorem ] is satisfied . while the autocovariance function of a general arma process does not in general have a simple closed form , its fourier transform is given by @xmath250.\ ] ] since @xmath63 is rational , assumptions of [ maintheorem ] are satisfied as well . in order to compute the lsd of @xmath132 , it is necessary , by [ lemma - densitygamma ] , to find the roots of a trigonometric polynomial of possibly high degree , which can be done numerically . we now consider the special case of the arma(1,1 ) process @xmath251 , @xmath252 , for which one can obtain explicit results . by [ eq - specdensarma ] , the spectral density of x is given by @xmath253.\ ] ] implies that the lsd of the autocovariance matrix @xmath132 has a density @xmath191 , which is given by @xmath254:f(\omega)=\lambda } \frac{1}{\left\|f'(\omega)\right\|}\\ = & \frac{1}{\pi(\vartheta+\varphi\lambda)\sqrt{\left[(1+\vartheta)^2-\lambda(1-\varphi)^2\right]\left[\lambda(1+\varphi)^2-(1-\vartheta)^2\right]}}{\mathbf{1}}_{(\lambda_-,\lambda_+)}(\lambda),\end{aligned}\ ] ] where @xmath255 by [ maintheorem ] , the stieltjes transform @xmath256 of the limiting spectral distribution of @xmath6 is the unique mapping @xmath116 that satisfies the equation @xmath257\left[(1+\varphi)^2+m_z(1-\vartheta)^2\right]}}.\notag\end{aligned}\ ] ] this is a quartic equation in @xmath258 which can be solved explicitly . an application of the stieltjes inversion formula then yields the limiting spectral distribution of @xmath6 . if one sets @xmath259 , one obtains an ma(1 ) process ; plots of the densities obtained in this case for different values of @xmath260 and @xmath7 are displayed in [ fig - densitiesma1 ] . similarly , the case @xmath261 corresponds to an ar(1 ) process ; see [ fig - densitiesar1 ] for a graphical representation of the densities one obtains for different values of @xmath262 and @xmath7 in this case . for the special case @xmath263 , @xmath264 , [ fig - histoarma11 ] compares the histogram of the eigenvalues of @xmath6 with the limiting spectral distribution obtained from [ maintheorem ] for different values of @xmath7 . for the stieltjes transform of the limiting spectral distribution of the sample covariance matrix of an arma(1,1 ) process should be compared to ( * ? ? ? * eq . ( 2.10 ) ) , where the analogous result is obtained for an autoregressive process of order one . they use the notation @xmath265 and consider the spectral distribution of @xmath266 instead of @xmath6 . if one observes that this difference in the normalization amounts to a linear transformation of the corresponding stieltjes transform , one obtains their result as a special case of [ eq - stieltjesarma11 ] . in many practical situations , data exhibit longrange dependence , which can be modelled by longmemory processes . denote by @xmath267 the backshift operator and define , for @xmath268 , the ( fractional ) difference operator by @xmath269 a process @xmath270 is called a fractionally integrated arma(p , d , q ) processes with @xmath271 and @xmath272 if @xmath273 is an arma(p , q ) process . these processes have a polynomially decaying autocorrelation function and therefore exhibit longrangedependence , cf . * theorem 13.2.2 ) and @xcite . we assume that @xmath274 , and that the zeros of the autoregressive polynomial @xmath243 of @xmath273 lie outside the closed unit disk . then it follows that @xmath242 has an infiniteorder moving average representation @xmath193 , where the @xmath163 have , in contrast to our previous examples , not an exponential decay , but satisfy @xmath275 , for some @xmath276 . therefore , if @xmath274 , one can apply [ maintheorem ] to obtain the lsd of the sample covariance matrix , using that the spectral density of @xmath270 is given by @xmath277.\ ] ] both authors gratefully acknowledge financial support from technische universitt mnchen - institute for advanced study funded by the german excellence initiative , and from the international graduate school of science and engineering .	we derive the distribution of the eigenvalues of a large sample covariance matrix when the data is dependent in time . more precisely , the dependence for each variable @xmath0 is modelled as a linear process @xmath1 , where @xmath2 are assumed to be independent random variables with finite fourth moments . if the sample size @xmath3 and the number of variables @xmath4 both converge to infinity such that @xmath5 , then the empirical spectral distribution of @xmath6 converges to a nonrandom distribution which only depends on @xmath7 and the spectral density of @xmath8 . in particular , our results apply to ( fractionally integrated ) arma processes , which we illustrate by some examples .
with the accessibility to the rest - frame mid - ultraviolet ( 22003200 ) spectra of distant evolved systems ( @xmath2 ) , the analysis of the ultraviolet morphology of intermediate and late type stars and old systems in the local universe has regained interest ( e.g. @xcite ; @xcite ; and , more recently , @xcite ; @xcite ) . dating galaxies which have gone through null or negligible star formation events over the past few gigayears at these redshifts has fundamental implications not only from an individual object point of view , but also at cosmological scales . the mid - ultraviolet offers a new spectral window which may help to break the so - called age - metallicity degeneracy present in the optical spectra of evolved stellar systems @xcite . in general , the uv analyses of stellar populations have relied on empirical stellar databases , mainly the one constructed upon iue low resolution data @xcite . among the most important efforts to characterize the uv morphology of stars that presumably dominate the uv spectrum of old systems was presented by @xcite and @xcite . they have defined 25 narrow - band indices in the ultraviolet covering two spectral segments : 12301930 and 19503200 , corresponding to the wavelength limits of iue cameras . spectroscopic indices defined in the far - uv @xcite were created to provide alternative tools for the analysis of active star - forming galaxies . the second set of indices @xcite was constructed with the aim of analyzing the properties of prominent spectral features in old populations . in general , these indices focus on wavelength regions that are more sensitive to stellar atmospheric parameters and less affected by the iue instruments artifacts and by the effects of features ( emission and/or absorption ) of interstellar origin . given the prevailing contribution to the mid - uv wavelength range of bright main sequence stars @xcite , the study of this spectral region is of special importance as a diagnostic tool , for instance to date high - redshift galaxies @xcite . it has been recognized , however , that empirical databases are generally composed of stars of the solar neighbourhood and therefore unavoidably carry on the evolutionary imprints of the milky way . in order to cope the paucity of chemical composition of empirical libraries required for the analysis of old stellar systems , many of the population synthesis works have incorporated theoretical libraries of stellar fluxes . amongst the most popular libraries is the kurucz grid of spectral energy distributions ( seds ) which is public through his web site . it has been found that kurucz flux library well represents segments of the mid - uv spectra of stars of intermediate and late type , in particular the continuum measurements ( slopes in the seds ) and some features such as the blend at about 2538 ( lfb00 ) . the limitations of the public kurucz grid of fluxes has been a matter of discussion in several investigations . in this respect @xcite concluded that for the analysis of the integrated uv properties of stellar aggregates , low resolution synthetic fluxes should be preferred over high resolution data . the reason being that use of high resolution theoretical data is precluded by the still unsolved uncertainties in atomic and molecular line data . it is important to note , nevertheless , that kurucz grid of fluxes have a wavelength sampling of 10 and therefore might lack of the appropiate resolution for a direct comparison with , for example , the commonly used iue database . while we agree that , for now , only low resolution ( e.g. broad band indices ) can be safely used , it is important to stress that at low resolution a great deal of information is lost and efforts should concentrate in the improvement of high resolution theoretical data . such an improvement must be based on a quantitative validation process by comparing theoretical data with stellar observations . on this line , @xcite and @xcite have embarked into a project aimed at providing a large theoretical database fully consistent with space telescope imaging spectrograph ( stis ) observations . in this paper we present a quantitative comparison between observed mid - uv features in the form of absorption line and continuum indices and those computed from high resolution synthetic spectra . we start , in section [ sec : theory ] , by investigating the effects of stellar parameters and instrumental resolution on spectroscopic indices from a purely theoretical point of view . in section [ sec : emp_vs_th ] we compare theoretical and observed indices and discuss the extent to which empirical indices are reproduced by theory . section [ sec : calib ] is devoted to provide the equations needed to transform theoretical ( discrepant ) indices to the standard ( iue ) system . this represents the first attempt to derive synthetic indices from high resolution mid - uv spectra . following the definitions given in table [ tab : indices ] , we carried out the calculation of a set of mid - ultraviolet indices of the synthetic stellar seds of the uvblue libraryeps / uvblue / uvblue.html and http://www.inaoep.mx/@xmath3modelos/uvblue/uvblue.html . ] , described in ( * ? ? * hereafter paper i ) . the indices measure either the strength of absorption features or the slope of the continuum . it is important to mention that most of the definitions listed in table [ tab : indices ] correspond to those provided by @xcite . two continuum indices ( s2850 and s2850l ) are analogous to the index defined by lfb00 . the band passes for these two indices and for 2110/2570 are those of bressan ( private communication ) . in column 5 of table [ tab : indices ] we list the atomic species that most contribute to the absorption in the central band of each line index . the list has been assembled from a visual inspection of the relevant absorption features present in the theoretical spectrum with @xmath0/@xmath1/[m / h]=5000/4.0/0.0 . the general procedure followed by @xcite to define the wavelength sequence and measure spectroscopic indices was carried out in a way similar to that described by @xcite . for the calculations presented here , uvblue has been degraded to match iue nominal resolution ( fwhm=6 ) . the line indices are the ratio of the integrated flux @xmath4 in the index passband to the integrated continuum flux @xmath5 , obtained by linear interpolation of the mean flux of the two side bands ; the result is transformed to magnitudes : @xmath6 where @xmath7 and @xmath8 are the limits of the index passband . the continuum indices are given by the ratio of the mean flux of the blue passband , @xmath9 , to the mean flux of the red one , @xmath10 , transformed to magnitudes : @xmath11 in the mid - uv , the spectra of intermediate and late - type stars are strongly affected by line blanketing and therefore it is hard to find spectral regions free from features . for this reason the spectral bands are selected such as to optimize each index for a temperature interval in which the feature is maximum . the homogeneous coverage in stellar parameters , in particular the chemical composition , allows the investigation of their effects on each index . the results are partially presented in table [ tab : uvblueindices ] where we list some synthetic indices ( line indices ) for a subset of atmospheric parameters in uvblue , identified in columns 1 - 3 . the full grid of indices is avaliable in electronic form . in figure [ fig : idxteffgrav ] we plot the index vs. effective temperature ( @xmath0 ) and surface gravity ( @xmath1 ) for synthetic spectra of solar metallicity . the reference labels inscribed in the upper left panel stand for different gravities which range from @xmath12 to 5.0 dex . we would like to stress at this point that the approach in this section is purely theoretical and not meant to analyze the compatibility between theoretical and empirical indices . some general remarks can be drawn just on the basis of a quick - look analysis of the different plots of fig . [ fig : idxteffgrav ] , simply relying on prime physics principles . in all cases it is evident that indices probe a characteristic temperature range at which the corresponding absorption feature ( or pseudo - continuum break ) is maximum . as expected , metal indices for higher ionization states generally peak at warmer effective temperature : this is because stellar layers at the appropriate thermodynamical conditions to produce the feature lie at a shallower optical depth with increasing stellar effective temperature . this is well evident , for instance , looking at the mg ii 2800 and mg i 2852 indices or comparing the fe ii 2402 feature with the fe i 3000 one . in both cases , the higher ionization state is the prevailing one among f stars while the neutral metal becomes more important for the g spectral type . as for gravity effects , one has to consider that the relative partition of a given elemental species between , let us say , first ( @xmath13 ) and neutral ( @xmath14 ) ionization states scales according to the saha equation ( e.g. * ? ? ? * ) as @xmath15 with @xmath16 being the electronic pressure of the stellar plasma . therefore , with increasing @xmath17 ( that is by `` packing '' atoms more efficiently ) one makes easier the ionization process allowing metals to more efficiently feed @xmath18 to the plasma . as a consequence , the correspondingly higher value of @xmath16 leads any given equilibrium partition @xmath19 to shift to slightly higher temperatures . so , in fig . [ fig : idxteffgrav ] a given spectral feature is seen to peak at earlier spectral types with increasing @xmath17 ( see , for instance , the nice trend of fe ii 2402 or mg ii 2800 in the figure ) . in addition , the broader damping wings of high - gravity spectral features mimic the wiping effect of a lower spectral resolution ( see sec . [ sec : resolution ] ) thus favoring weaker index values for dwarf stars compared to giants . these qualitative arguments are summarized in fig . [ fig : trend ] . based on a more careful inspection of fig . [ fig : idxteffgrav ] , we can give below further comments on individual indices : _ fe ii 2332 _ : the absorption in the central band of this index is mainly due to fe ii and fe i lines , although several other elements add significant contribution , as in all other indices that we present here . this index has its maximum for all gravities at about @xmath20 k. it displays a large variation with surface gravity , ranging from 0.5 mag for high gravity stars to 1.5 mag for supergiant spectra . _ fe ii 2402 _ : feature produced by multiplets of fe ii . this index is very sensitive to gravity . at its maximum , at @xmath21 k , the index doubles its value from about 0.8 to 1.5 mag when the gravity decreases from @xmath22 to 1.0 . all gravities peak at about the same temperature ( 6000 k ) . _ bl 2538 _ : the blend is produced by mainly fe , mg , cr , and ni absorption . among line indices , bl 2538 shows the highest sensitivity to surface gravity . it peaks ( index @xmath23 mag ) at the lowest temperature ( 4000 k ) for @xmath22 . for @xmath24 the index drastically drops to negative values . note that , contrary to most indices , for @xmath0 above 4500 k this index increases with increasing gravity . _ fe ii 2609 _ : the absorption in the index band is largely due to fe ii , with contributions by fe i and mn ii . this index has its maximum at about 5000 k , temperature at which it displays a strong sensitivity to gravity with a variation from 1.5 mag , for @xmath22 , to 2.5 mag for @xmath25 . it peaks at about the same temperature for all gravities . _ bl 2720 _ : blend produced by neutral and ionized iron lines , with presence of cr i and v ii . down to @xmath26 k , this index shows little variation with surface gravity , with the exception of the outermost @xmath1 values . similarly to bl 2538 for high gravity , this index reaches its maximum at the lowest temperatures . _ bl 2740 _ : blend of iron and chromium lines . this index is similar to bl 2720 in that it shows marginal dependence with gravity , it sharply increases with decreasing @xmath0 , and it peaks at the lower temperature edge of uvblue . _ mg ii 2800 _ : this index includes the strongest feature in the mid - uv interval , due to mg ii . it sharply increases shortward of 10000 k and peaks at about 6000 k for @xmath27 and 5000 k for @xmath12 . the index shows slight dependence on gravity since it varies about 30% along the full gravity interval . this dependence is restricted for @xmath28 k. _ mg i 2852 _ : this index is produced by a strong mg i absorption line . it displays an enhanced sensitivity to the surface gravity with respect to mg ii 2800 . indices from high gravity spectra peak at @xmath20 k , while for the lowest gravity the maximum value is reached at about 4500 k. _ fe i 3000 _ : the absorption in the index band is dominated by fe i. it appears to be sensitive to @xmath0 in a short range , @xmath29 k , and it is moderately dependent on gravity . _ bl 3096 _ : the blend is produced by fe i , ni i , mg i , and al i. it displays a marginal sensitivity on gravity for high gravity spectra . giants and supergiants are not separated among themselves . _ mg wide _ : this index covers 600 of the mid - uv sed . its central band contains those of bl 2720 , bl 2740 , mg ii 2800 , and mg i 2852 . is has been envisaged as a useful indicator of the effective temperature of the dominant stellar population in evolved systems . defined by three 200 - wide bands , this index is expected to be suitable for the analysis of low resolution data of distant galaxies collected by current large facilities @xcite . the curves depicted in fig . [ fig : idxteffgrav ] show that the index increases monotonically for most gravities up to the low temperature edge ; only for @xmath22 the index saturates at 4500 k ( early k - type stars ) . mg wide is very sensitive to surface gravity with indices derived from lower gravity spectra displaying the highest values . the general behavior of continuum indices ( spectral breaks and broad band indices ) is depicted below . _ 2110/2570 _ : this index primarily measures the strong mg i magnesium break at 2512 . it is interesting to note that down to about 5000 k the index is insensitive to gravity . at lower temperatures the effects of gravity are moderate . for synthetic fluxes with @xmath30 k the index turn to negative values . _ 2609/2660 _ : as indicated by @xcite the spectral break at about 2600 is dominated by iron opacity . in our synthetic spectra indices for all gravities peak at about the same temperature ( 5500 k ) with perhaps a slight shift towards lower temperatures at lower gravities . the index shows a moderate sensitivity to gravity , changing by 50% over the entire @xmath1 interval . the break vanishes for models of 10000 k. _ 2828/2921 _ : this break is among the most prominent features in the mid - uv spectra in cool stars and evolved populations . it displays very low sensitivity to surface gravity for model spectra of @xmath31 k. below this temperature , which marks the peak for high gravity fluxes ( @xmath22 ) , the trends for different gravities clearly separate . at the lowest temperatures the change of the index in the interval @xmath125.0 is about a factor of two . _ 2600 - 3000 , s2850 , and s2850l _ : as can be seen from the bands listed in table [ tab : indices ] , these indices measure the slope between 2600 and 3000 . the differences are the band widths and the centers of the bands of the index 2600 - 3000 . as expected , the overall behavior of the three continuum indices is very similar : all turn from negative to positive values at @xmath32 k , higher gravity model fluxes display larger indices , and all are sensitive to gravity , with an enhanced sensitivity of the index s2850 that varies almost a factor of three from @xmath12 to 5.0 . we have described the general behavior of the 17 line and continuum indices from a purely theoretical point of view . it is worth to remark that the above results are affected by the theoretical limitations of model atmospheres , in particular , the limitations inherent to models constructed under the classical approximations , as is our case . in paper i we have commented on the uvblue caveats and the reader is referred to that paper for more details . the discussion of the impact of uvblue limitations on synthetic indices is deferred to sec . [ sec : emp_vs_th ] . nevertheless , we can anticipate that the overall trends of index values with @xmath0 and surface gravity are similar to those depicted by indices computed from empirical data @xcite . note that we have restricted the analysis to line indices redward of 2300 since both , theoretical and observed spectra , decrease in quality for intermediate and late stellar types . in what follows we describe the effects of chemical composition , which , in the context of stellar populations analyses , might be more important than those of gravity . the uniform metallicity coverage of uvblue ( from [ m / h][email protected] to @xmath34 ) allows the investigation , at least from an exploratory point of view , of the effects of chemical composition on spectroscopic indices . such an investigation is not feasible with current empirical databases since the vast majority of objects have solar or nearly solar metallicity ( see fig . 8 of paper i ) . in fig . [ fig : idxteffmet ] we show the trends of the mid - uv spectral indices as a function of the effective temperature and metallicity for a fixed surface gravity of @xmath35 dex . different line types correspond to different metallicities as denoted in the labels in the upper left panel in the figure . like for gravity effects , a similar trend on spectrophotometric indices has also to be expected by increasing stellar metallicity . in this case , @xmath16 would increase in eq . ( [ eq : saha ] ) merely as a consequence of a higher value of @xmath36 . in addition , in case of thin absorption features , the corresponding index strength would increase reaching a stronger peak for higher values of metal abundance . this is nicely shown in fig . [ fig : idxteffmet ] , specially for the case of metal blends , like the bl 2538 or bl 2740 indices . recalling our previous physical arguments ( see right panel of fig . [ fig : trend ] ) , note that systematically , the loci of the maxima in the plots of fig . [ fig : idxteffmet ] also depend on the metal content , this is , high metallicity indices peak at higher temperatures . because of this , in some indices , namely fe ii 2332 , fe ii 2402 , and mg ii 2800 , the effects of chemical composition are reversed after the index saturates towards low @xmath0 : at the lowest temperatures the low metallicity indices are higher . the peak at different metallicities are shifted by as much as 3000 k if we compare , for the index fe ii 2402 , metal rich indices with those of the lowest chemical composition . the variation of the indices can be as high as a factor of three in the full metallicity interval . the three redder indices , mg ii 2852 , fe i 3000 , and , notably , bl 3096 display marginal variations with chemical composition . this property seems promising for separating the effects of age and metallicity in integrated spectra . an important test that has to be conducted and quite often is neglected is the analisys of sensitivity of spectral indices to instrumental resolution . the comparison of theoretical vs. observed indices ( either stellar or integrated ) or the analyses of indices measured from spectra from data sets collected with different instrumentation should be carried out with data at the same resolution . the risk of not doing so can be , for instance , the incorrect segregation of potential useful indices . the changes due to spectral resolution might have important implications in the calculation of spectroscopic indices , in particular if the indices are used in the study of the integrated properties of galaxies whose spectra are intrinsically broadened by the motions of the stellar component . in figure [ fig : idxres ] we show the effects of resolution on the line and continuum indices discussed in the previous section . in the _ y_-axis we indicate the index difference with respect to its value at a resolution of 6 . for the analysis we have considered two synthetic spectra of @xmath37 , [ m / h]@xmath38 and @xmath39 and 6000 k ( represented by solid and dashed lines , respectively ) , temperatures at which most indices reach their maxima . data points correspond to indices measured in spectra after applying gaussian filters for differents fwhm ranging from 1 to 20 with a step of 1.0 . the open circles in each panel indicate the position of the reference iue resolution ( 6 ) and the resolution of the kurucz public library of fluxes @xcite . in general , one would expect that resolution effects will be modulated by the width of the bands , in the sense that broader bands will be less sensitive to resolution . in a similar way , effects of resolution will depend on the intrinsic width of the spectral lines , this is , whether a line is fully embraced or not within the central bandpass . let us first comment on the line indices . we can identify several properties of the effects of resolution on index values . first , all line indices have their lowest value at the lowest resolution ( 20 ) . the only exception is mg wide , which , however , displays a marginal variation of less 0.002 mag between 6 and 10 . second , all indices have their peak values either at the highest resolution or at about 24 . this is most probably due to the role that features very near to the limit of the bands ( either inside and outside ) play in modulating an index . third , with the exception of the index bl 3096 , all indices vary up to 40% in the full resolution range . bl 3096 doubles its value from 1 to 20 , and changes as much as 0.05 mag ( about 20% ) in the 6 to 10 interval . concerning the continuum indices we find a wide diversity of behaviors , nevertheless in four of them ( 2120/2570 , 2600 - 3000 , 2609/2660 , and s2850l ) the effects of resolution are very weak within the full resolution range , while for the remaning two , 2828/2921 and s2850 , values change up to 0.12 mag and 0.8 mag ( 30 and 40% ) , respectively . note that this latter couple of indices are those defined with the narrowest bands . it is important to stress that the exercise shown here is based only on two model spectra . at other temperatures and chemical compositions the effects might decrease or be enhanced . with the goal of quantitatively compare theoretical indices with indices derived from observed data , we have adopted the database described in paper i to where the reader is referred for more details . in summary , our working sample is based on the @xcite atlas of iue low resolution spectra and the cool star extension by @xcite . a sub sample of 111 stars was secured by imposing that at least one full set of atmospheric parameters be available in @xcite . keeping in mind that there exist differences in the image extraction processes used in the original @xcite atlas and our iue - ines data set , we have , as a first step , verified the extent to which new indices measured in re - processed iue spectra are compatible with those reported by @xcite . the overall result is that most indices display a one to one correlation with the exception of fe ii 2402 where there is a quite high dispersion for points corresponding to giant stars . the most plausible explanations for these differences are , on one side , the different absolute flux calibration curves used in _ ines _ @xcite and _ iuesips _ @xcite and , on the other side , the fact that we used all available images of good quality ( in some cases more than 50 ) while @xcite considered only a few . in addition to this , one has to consider that spectra of cool stars are often quite noisy since their intrinsic uv flux descreases drastically and that at the short wavelength edge the flux is close to the lwp / r cameras sensitivity limits . with the indices computed in the re - processed iue images we have defined the standard system . from the theoretical side , we have created a synthetic data set through a three - linear interpolation of the flux , divided by the corresponding @xmath40 , of the uvblue spectra in the space ( @xmath0 , @xmath1 , [ m / h ] ) . for each star we adopt the parameters listed in table 2 of paper i. as in the case of the synthetic seds presented in the previous section , these spectra were degraded to match iue resolution ( 6 ) . it is important to mention that we have imposed one restriction for the comparison : we have excluded low gravity objects ( @xmath41 ) since , on one side , giant and supergiant stars show the largest discrepancies between observed and theoretical seds according to paper i and , on the other side , intrinsic uv fluxes of low gravity objects at the short wavelength edge of iue are lower that the high gravity counterparts . additionally , evolved objects of low surface gravity are not expected to significantly contribute to the mid - uv flux of early - type systems , one of the main foreseeable applications of synthetic uv indices . the final sample for our comparison is composed of 68 objects whose distribution in @xmath0 and [ m / h ] is displayed in figure [ fig : histo ] . the restriction on wavelength for the material presented in this paper ( @xmath42 , except for the spectral break index 2110/2570 ) ensures a higher signal to noise in observed data . note also that no restriction on chemical composition was imposed . the results of the comparison are displayed in figure [ fig : syniue ] , where we plot the synthetic indices ( @xmath43-axis ) and iue - indices ( @xmath44-axis ) together with a reference 45@xmath45 dotted line . ten indices are overestimated by the results of classical model atmospheres . of these indices two of them ( fe ii 2332 and fe ii 2402 ) represent the worst cases with a large scatter over the whole index range . all the remaining overestimated indices ( bl 2538 , fe ii 2609 , bl 2720 , bl 2740 , mg ii 2800 , mg i 2852 , bl 3096 , and the spectral break 2609/2660 ) exhibit a clear linear correlation ; however , in five of them there is a notable enhancement of scattering on the points for low temperatures stars , not so for the indices bl 2720 , mg i 2852 , and bl 3096 . additionally , it is interesting to note that in three indices with large scatter at low @xmath0 the linear relation appears to be truncated and the empirical indices show a reversal , i.e. indices saturate and decrease after reaching their maxima towards low temperatures . this situation is not modelled by the synthetic indices . two indices ( mg wide and the continuum index 2600 - 3000 ) appear slightly underestimated by theory . finally , five indices appear properly reproduced by uvblue indices : 2110/2570 , fe i 3000 , the spectral break 2828/2921 , and the continuum indices s2850 and s2850l . let us compare our results with those of lotz , ferguson & bohlin ( 2000 , herefater lfb00 ) which represent , to our knowledge , the only attempt so far to compare kurucz - derived synthetic mid - uv indices with observations . they present eight indices all of which are included in our analysis , although , our definition of s2850 is different from that of lfb00 . they concluded that for five of these indices , namely , fe ii 2402 , fe ii 2609 , mg ii 2800 , mg i 2852 , and mg wide , values are overestimated by kurucz model fluxes . they found a good match for bl 2538 and for s2850 ( the latter for stellar models with @xmath46 k ) , while fe i 3000 is underestimated by 0.15 mag . while our results agree for the first four indices , for the rest we obtain contradicting behaviors . the bl 2538 index appears to be also overestimated , particularly at low temperatures , mg wide is somewhat underestimated , and our synthetic fe i 3000 and the slope between 2600 and 3000 match very well empirical indices . there are three plausible explanations for the conflicting results . first , the fact that lfb00 measure theoretical indices from the set of kurucz model fluxes compiled by @xcite which are at a nominal resolution of 10 while our high resolution theoretical data set has been degraded to 6 . as we have explained in section [ sec : resolution ] , this change in resolution can account , at least partially , for the discrepancies . for instance , bl 2538 and fe i 3000 decrease by about 0.04 mag after changing the resolution from 6 to 10 . second , kurucz model fluxes are calculated using the opacity distribution functions ( odfs ) while the uvblue seds account for detailed line opacity . it is clear that the statistical nature of the odfs dilutes the effects of incomplete and/or incorrect line opacities , however , at the price of preventing the possibility of producing spectra at high enough resolution ( see also the discussion in * ? ? ? third , in our comparison we have accounted for the individual stellar parameters while lfb00 provide a comparison of empirical indices obtained from mean stellar iue spectra with those measured from a single - gravity single - metallicity ( solar ) theoretical flux . even though the definition of the s2850 index in lfb00 and our work differs , our comparison indicates that the slopes we measure in uvblue - interpolated spectra closely reproduce empirical results . various agents more can be brandished to explain the discrepancies seen in figure [ fig : syniue ] . as briefly mentioned by lfb00 and listed in paper i as caveats of uvblue , these are : incomplete line lists , which directly affect the heavily blanketed ultraviolet fluxes ; departures from lte and chromospheric heating . additionally , and not explicitly mentioned in the above investigations , is the fact that , even if we had a complete line list , the line parameters ( mainly the oscillator strengths , van der waals damping constants and wavelengths ) might be wrong . we believe that discrepancies in many of the indices can be significantly reduced with the improvement of these three parameters , since , as demonstrated in paper i , individual lines are systematically stronger in theoretical spectra . a thorough spectral line study and their effects on individual index bands is necessary to unambiguously identify the agent(s ) that provoke the discrepancies in the comparison between observed and theoretical indices . such an analysis is beyond the scope of this work , however , we can speculate on the most likely explanations . _ mg wide and 2600 - 3000 _ : the underestimation of these two indices is attributed to a lower opacity in their shared blue band which partially includes the spectral region 26402700 . in paper i we have already pointed out that this region is poorly reproduced by model spectra ; the most plausible reason for this is the lack of many absorption lines in the uvblue line list @xcite . furthermore , in uvblue we did not considered the so - called predicted " lines @xcite which affect the blanketing in this region . the reason for avoiding the use of lines with non - laboratory atomic data is that uvblue is also intended for high resolution analyses and the inclusion of unreliable data greatly increases confusion for line identification . moreover , for mg wide these effects are mixed with the overvalued strength of the mg features comprised in its central band . we do not expect any important contribution to the inconsistencies from the red band since , on one side , we know that within the mid - uv spectral interval line blanketing effects are enhanced towards shorter wavelengths . on the other side , the red and middle bands and partially the blue band of the index fe i 3000 are embraced by the red band of mg wide and 2600 - 3000 , and this index is well reproduced . _ fe ii 2609 , bl 2720 , bl 2740 _ : these indices , which share one side band , are also affected by the incompatibilities between the theoretical and observed seds in the region 26402720 which result in the largest deviations from the one - to - one reference line . _ fe ii 2332 , fe ii 2402 , mg i 2852 _ : these indices are defined in regions where part of the discrepancies can be ascribed to differences in the individual feature intensities themselves . an additional aspect that should be taken into account is that the discrepant results of the two bluer indices might also be ascribed to the low quality of observational data in the wavelength interval used to define them . this is supported by the large scatter present even at low index values . several authors have concentrated on redder indices because of this potential limitation . _ mg ii 2800 _ : the prominent mg ii line at about 2800 is filled by chromospheric emission . the different extents to which stars have chromospheric activity could provide the observed dispersion in the points , particularly at low @xmath0 ( i.e. at the largest value of the index ) . even though we knew in advance that this feature could not be modelled because of the non - thermal heating of the upper atmosphere , we decided to include it for the sake of completness and to explore its behavior , particularly in terms of chemical abundance . in fact we found that different metallicities peak at different temperatures and therefore it could partially explain its paradoxical behavior in stellar populations where the index is smaller with increasing metallicity . if this is true one should explore in more detail the less studied indices fe ii 2332 and fe ii 2402 where these effects arise more pronounced . _ bl 3096 _ : this index is marginally overestimated by uvblue model fluxes until it reaches a value of about 0.25 which roughly corresponds to intermediate k - type stars . larger indices ( i.e. cooler stars ) are apparently not reproduced ; however , there are too few points to draw any conclusion . it is important to stress that to reach a better agreement between theoretical and the empirical results , an improvement of the line parameters is much needed . we anticipate that we have initiated a thorough test of the line parameters for the strongest lines in the mid - uv range ( excluding those that are severely affected by chromospheric emission ) . we have seen that five out of seventeen empirical indices are adequately reproduced by their synthetic counterparts . this result invites to a thorough analysis of the main agents affecting the uv spectra in late - type stars . however , we can still extract useful information from the indices that display a clear linear correlation between theory and observations . this can be done by calibrating the synthetic indices to match empirical ones in a similar way @xcite did for optically defined indices . we recall that no restriction has been imposed on metallicity . hence , the resulting calibration should be in principle applicable to the full chemical composition and effective temperature ranges involved in our iue database , thanks to the improved homogeneity in the parameter space provided by the theoretical spectra . the general underlying goal of the transformation of a set of data to a standard system could be manyfold . for example , it is required when two data sets have not been obtained with the same instrument , i.e. have different resolution . when the flux calibration in both sets is different or has not been conducted at all in one or both and therefore it is necessary to match the different detector responses . a standard system is usually conformed by a sufficiently large sample of stellar observed spectra . this transformation has been a common practice in the analysis of , for example , the collection of optical indices defined by the lick group . for the case of mid - uv indices we have defined , as mentioned previously , the mid - uv standard system as the set of indices measured from the re - processed iue data . in our particular case , the transformation is aimed at transporting synthetic indices to the expected empirical values . in figure [ fig : syniue ] we have represented with a solid line the least squares fit of the function @xmath47 for 11 indices . we constrained the fitted line to include the ( 0,0 ) point , i.e. both the theoretical and empirical indices vanish at the same temperature . however , one can see that , in any case , without this restriction the @xmath43-intercept would be very close to zero . the fitting process has been done in an iterative way , rejecting points ( denoted by open circles in fig . [ fig : syniue ] ) located more than @xmath48 away from the linear fit , where @xmath49 and @xmath50 is the number of stars . columns 25 of table [ tab : calib ] give the results of the calibration , its associated error , the rms in magnitudes , and the number of stars left after removing objects as described above . in this work we have presented a set of seventeen mid - uv indices computed from the synthetic stellar seds of the uvblue library . we have explored for the first time their sensitivity in terms of the atmospheric parameters ( @xmath0 , @xmath1 , [ m / h ] ) as well as spectral resolution . we find that theoretical indices display a wide variety of behaviors , however , qualitatively resembling the results from empirical analysis using iue data . this is , all line indices peak at effective temperatures of the order of 6500 k or lower and are higher with increasing gravity . for the purpose of analyzing integrated spectra of evolved populations the detection of features involved in any of the indices manifests the presence of stars older that 1 gyr . bearing in mind that the uv light of stellar aggregates is dominated by stars at the turn off , the effects of chemical composition are more important than those due to surface gravity . the results indicate that in general high metallicity theoretical indices are larger than their metal - poor counterparts ; nevertheless for some of them ( fe ii 2332 , fe ii 2402 and , to some extent , also mg ii 2800 ) we find that the loci of the maxima strongly depend on metallicity . the result of this shift is that for a given lower temperature than the one at the peak , metal poor indices are higher . interestingly we find that the index bl 3096 is even less sentitive to metallicity than fe i 3000 , which was empirically found to marginally respond to changes of this parameter . the effects of instrumental resolution vary from negligible to very important , even if we only consider the range in resolution limited by iue and kurucz low resolution fluxes ( 6 and 10 , respectively ) . the index bl 3096 can change up to 20% in this interval . among the most important results of this work is that , after quantitatively comparing theoretical and empirical indices , we have found that five indices ( although s2850 and s2850l measure the same property on the sed ) agree very well with observations on the entire range of the parameter space delimited by the observed high - gravity stars . the indices cover both line and continuum measurements that show distinctive characteristics in stars . this property ensures their applicability in low as well as intermediate resolution analyses of the mid - uv sed of evolved stellar populations . m.c . and e.b . are pleased to thank financial support from mexican conacyt , via grants 36547-e and sep-2004-c01 - 47904 . a.b . acknowledges partial financial support by the italian miur under grant inaf prin/05 1.06.08.03 . , m.l.m . and c.m . would like to acknowledge funding from italian grants miur cofin-2003028039 and prin - inaf 2005 ( p.i . m. bellazzini ) . thanks are due to alessandro bressan for providing us with the definition of the 2110/2570 , s2850 , s2850l indices in table [ tab : indices ] . lcccl & blue bandpass & index bandpass & red bandpass & main species + index name & ( ) & ( ) & ( ) & + fe ii 2332 & 2285 2325 & 2333 2359 & 2432 2458 & fe ii , fe i , co i , ni i + fe ii 2402 & 2285 2325 & 2382 2422 & 2432 2458 & fe ii , fe i , co i + bl 2538 & 2432 2458 & 2520 2556 & 2562 2588 & fe i , fe ii , mg i , cr i , ni i + fe ii 2609 & 2562 2588 & 2596 2622 & 2647 2673 & fe ii , fe i , mn ii + bl 2720 & 2647 2673 & 2713 2733 & 2762 2782 & fe i , fe ii , cr i + bl 2740 & 2647 2673 & 2736 2762 & 2762 2782 & fe i , fe ii , cr i , cr ii + mg ii 2800 & 2762 2782 & 2784 2814 & 2818 2838 & mg ii , fe i , mn i + mg i 2852 & 2818 2838 & 2839 2865 & 2906 2936 & mg i , fe i , cr ii , fe ii + mg wide & 2470 2670 & 2670 2870 & 2930 3130 & mg i , mg ii , fe i , fe ii , cr i , cr ii + fe i 3000 & 2906 2936 & 2965 3025 & 3031 3051 & fe i , cr i , fe ii , ni i + bl 3096 & 3031 3051 & 3086 3106 & 3115 3155 & fe i , ni i , mg i , al i + 2110/2570 & 2010 2210 & ... ... & 2470 2670 & + 26003000 & 2470 2670 & ... ... & 2930 3130 & + 2609/2660 & 2596 2623 & ... ... & 2647 2673 & + 2828/2921 & 2818 2838 & ... ... & 2906 2936 & + s2850 & 2599 2601 & ... ... & 3099 3101 & + s2850 l & 2590 2610 & ... ... & 3090 3110 & + cccccccccccccc 4000 & 1.0 & -2.0 & 0.44061 & 0.47539 & 0.43790 & 0.87782 & 0.59571 & 0.50926 & 0.76453 & 1.35895 & 1.04094 & 0.35647 & 0.16311 + 4000 & 1.5 & -2.0 & 0.41748 & 0.48331 & 0.59642 & 1.04292 & 0.67021 & 0.58925 & 0.83816 & 1.53405 & 0.99007 & 0.39780 & 0.20965 + 4000 & 2.0 & -2.0 & 0.37282 & 0.53914 & 0.77864 & 1.11860 & 0.76399 & 0.67206 & 0.88320 & 1.70192 & 0.89574 & 0.44023 & 0.29045 + 4000 & 2.5 & -2.0 & 0.29809 & 0.61892 & 0.93896 & 1.08568 & 0.85078 & 0.73388 & 0.85145 & 1.75295 & 0.76781 & 0.47426 & 0.40146 + 4000 & 3.0 & -2.0 & 0.23567 & 0.62265 & 0.98577 & 0.91336 & 0.83173 & 0.69016 & 0.70402 & 1.65113 & 0.62329 & 0.45628 & 0.46000 + 4000 & 3.5 & -2.0 & 0.17223 & 0.55195 & 0.98687 & 0.68457 & 0.77044 & 0.60286 & 0.52936 & 1.50798 & 0.50655 & 0.40967 & 0.46412 + 4000 & 4.0 & -2.0 & 0.09713 & 0.48134 & 0.99168 & 0.50300 & 0.72933 & 0.54143 & 0.38437 & 1.35511 & 0.43448 & 0.37278 & 0.45930 + 4000 & 4.5 & -2.0 & -0.21163 & 0.50405 & 0.88451 & 0.80879 & 1.23938 & 0.77861 & 0.11611 & 0.86800 & 0.48530 & 0.26252 & 0.39315 + 4000 & 5.0 & -2.0 & -0.50491 & 0.36852 & 0.94043 & 0.43989 & 1.21883 & 0.77068 & 0.07321 & 0.85313 & 0.53453 & 0.28996 & 0.47198 + 4500 & 1.5 & -2.0 & 0.31132 & 0.48188 & 0.18393 & 0.41794 & 0.11088 & 0.21140 & 0.59527 & 0.13673 & 0.26370 & 0.05283 & 0.08340 + 4500 & 2.0 & -2.0 & 0.71607 & 1.10758 & 0.53916 & 1.37005 & 0.34172 & 0.51743 & 1.31980 & 0.89645 & 0.48798 & 0.18059 & 0.18137 + 4500 & 2.5 & -2.0 & 0.61902 & 1.07981 & 0.67985 & 1.28936 & 0.41383 & 0.53511 & 1.30672 & 1.17151 & 0.43161 & 0.22482 & 0.21524 + 4500 & 3.0 & -2.0 & 0.51926 & 1.02658 & 0.83930 & 1.22573 & 0.49990 & 0.56285 & 1.25480 & 1.46864 & 0.39972 & 0.28141 & 0.26612 + 4500 & 3.5 & -2.0 & 0.42160 & 0.96628 & 1.01393 & 1.17014 & 0.59235 & 0.59360 & 1.15795 & 1.69445 & 0.38010 & 0.34159 & 0.33579 + index name & b & @xmath51 & rms & # + & & & ( mag ) & + bl 2538 & 1.517 & 0.041 & 0.18 & 62 + bl 2720 & 3.028 & 0.117 & 0.12 & 67 + mg i 2852 & 1.797 & 0.036 & 0.13 & 66 + mg wide & 0.811 & 0.013 & 0.03 & 64 + fe i 3000 & 0.966 & 0.025 & 0.04 & 66 + bl 3096 & 1.260 & 0.046 & 0.05 & 66 + 2110/2570 & 0.934 & 0.027 & 0.24 & 60 + 26003000 & 0.717 & 0.012 & 0.14 & 67 + 2828/2921 & 1.068 & 0.021 & 0.08 & 60 + s2850 & 1.014 & 0.011 & 0.14 & 64 + s2850 l & 0.924 & 0.010 & 0.13 & 67 +	using the uvblue library of synthetic stellar spectra we have computed a set of mid - uv line and continuum spectroscopic indices . we explore their behavior in terms of the leading stellar parameters ( @xmath0 , @xmath1 ) . the overall result is that synthetic indices follow the general trends depicted by those computed from empirical databases . separately we also examine the index sensitivity to changes in chemical composition , an analysis only feasible under a theoretical approach . in this respect , lines indices fe i 3000 , bl 3096 and mg i 2852 and the continuum index 2828/2921 are the least sensitive features , an important characteristic to be taken into account for the analyses of integrated spectra of stellar systems . we also quantify the effects of instrumental resolution on the indices and find that indices display variations up to 0.1 mag in the resolution interval between 610 of fwhm . we discuss the extent to which synthetic indices are compatible with indices measured in spectra collected by the international ultraviolet explorer ( iue ) . five line and continuum indices ( fe i 3000 , 2110/2570 , 2828/2921 , s2850 , and s2850l ) display a remarkable good correlation with observations . the rest of the indices are either underestimated or overestimated , however , two of them , mg wide and bl 3096 , display only marginal discrepancies . for 11 indices we give the coefficients to convert synthetic indices to the iue system . this work represents the first attempt to synthesize mid - uv indices from high resolution theoretical spectra and foresees important applications for the study of the ultraviolet morphology of old stellar aggregates .
besides bound companions or natural satellites , planets may have unbound companions or co - orbitals , i.e. objects trapped in a 1:1 mean motion resonance with the planet . co - orbital bodies are not only interesting curiosities found in the celestial mechanics studies , but also represent temporary reservoirs for certain objects as well as the key to understand the origin of retrograde outer satellites of the giant planets and the accretional processes in the early solar system ( see e.g. namouni , christou & murray 1999 ) . these co - orbital bodies can be primordial , if they were captured in their present resonant state early in the history of the solar system and have remained dynamically stable for billions of years , or transient , if they were captured relatively recently . several thousand minor bodies are known to be currently trapped in the 1:1 mean motion resonance with jupiter and most of them may have remained as such for gyr ( see e.g. milani 1993 ; jewitt , trujillo & luu 2000 ; morbidelli et al . 2005 ; yoshida & nakamura 2005 ; robutel & gabern 2006 ; robutel & bodossian 2009 ; grav et al . 2011 ; nesvorn , vokrouhlick & morbidelli 2013 ) . neptune also hosts a likely large population of co - orbitals both long - term stable ( e.g. kortenkamp , malhotra & michtchenko 2004 ; nesvorn & vokrouhlick 2009 ; sheppard & trujillo 2006 , 2010 ) and transient ( e.g. de la fuente marcos & de la fuente marcos 2012b , c ) . in sharp contrast , the number of asteroids currently engaged in co - orbital motion with uranus appears to be rather small and its nature comparatively ephemeral . uranus has centaurs temporarily trapped in other mean motion resonances as well ( masaki & kinoshita 2003 ; gallardo 2006 , 2014 ) . the theoretical possibility of finding primordial or transient uranian co - orbitals , in particular trojans , has been a subject of study for almost three decades . in general , numerical simulations predict that uranus may have retained a certain amount of its primordial co - orbital minor planet population and also that short - term stable ( for some myr ) co - orbitals may exist . zhang & innanen ( 1988a , b ) and innanen & mikkola ( 1989 ) found that some tadpole orbits may survive for 10 myr . mikkola & innanen ( 1992 ) described tadpole orbits stable for at least 20 myr and with libration amplitudes close to 100 ; horseshoe orbiters were only stable for shorter periods of time ( @xmath610 myr ) . holman & wisdom ( 1993 ) found similar results . gomes ( 1998 ) explained the observational absence of primordial uranus trojans as a side effect of planetary migration . wiegert , innanen & mikkola ( 2000 ) focused on quasi - satellites and found that some test orbits were stable for up to 1 gyr but only at low inclinations ( @xmath62 ) and with eccentricities in the range 0.10.15 . nesvorn & dones ( 2002 ) concluded that any existing uranus primordial trojan population should have now been depleted by a factor of 100 . gacka ( 2003 ) also found short - term stable tadpole orbits . in a very detailed study , marzari , tricarico & scholl ( 2003 ) identified a real absence of trojan stable orbits at low libration amplitudes and also that , among the jovian planets , uranus has the lowest probability of hosting long - term stable trojans . horner & evans ( 2006 ) stated that uranus appears not to be able to efficiently capture objects into the 1:1 commensurability today even for short periods of time . lykawka & horner ( 2010 ) found that uranus should have been able to capture and retain a significant population of trojan objects from the primordial planetesimal disc by the end of planetary migration . these authors concluded that originally the orbits of these objects should have had a wide range of orbital eccentricities and inclinations . predictions from numerical simulations appear to be generally consistent with the available observational evidence that transient uranian co - orbitals do exist but primordial ones may not . so far , uranus has only three known co - orbitals : 83982 crantor ( 2002 go@xmath0 ) ( gallardo 2006 ; de la fuente marcos & de la fuente marcos 2013b ) , 2010 eu@xmath1 ( de la fuente marcos & de la fuente marcos 2013b ) and 2011 qf@xmath2 ( alexandersen et al . 2013a ; alexandersen et al . due to its present short data - arc ( 85 d ) , 2010 eu@xmath1 is better described as a candidate . asteroids crantor and 2010 eu@xmath1 follow horseshoe orbits in a frame of reference rotating with uranus and 2011 qf@xmath2 is an l@xmath7 trojan ; the three of them are only short - term stable and , therefore , they must be captured objects . most published studies focus on what makes a hypothetical primordial population of uranian co - orbitals dynamically unstable , leaving the question of how the transient ones got captured unanswered . in absence of observational bias , the lack of a sizeable present - day population of minor bodies trapped in the 1:1 commensurability with uranus is probably due to persistent perturbations by the other giant planets . uranus trojans are affected by high - order resonances with saturn ( gallardo 2006 ) . in the case of current horseshoe librators , saturn also appears to be the main source of the destabilizing force ( de la fuente marcos & de la fuente marcos 2013b ) . dvorak , bazs & zhou ( 2010 ) studied the stability of hypothetical primordial uranus trojans and concluded that the trojan regions are mostly unstable . for these authors , the orbital inclination is the key parameter to characterize the stability of uranus trojans ; only the inclination intervals ( 0 , 7 ) , ( 9 , 13 ) , ( 31 , 36)and ( 38 , 50)appear to be stable ( regions a , b , c and d , respectively , in dvorak et al . the existence of these islands of stability enables the presence of trojan companions to uranus . asteroid 2011 qf@xmath2 appears to inhabit one of these stable areas in orbital parameter space as its current orbital inclination is nearly 11but it is not a primordial uranus trojan . here , we revisit the subjects of capture and stability of current uranian co - orbitals , providing an independent assessment of the 1:1 commensurability of the newly found uranus trojan , 2011 qf@xmath2 , studying its future orbital evolution and looking into its dynamical past . our numerical investigation is aimed at adding more pieces to the overall puzzle of the apparent scarcity of uranian co - orbitals as we explore the role of multibody ephemeral mean motion resonances . on the other hand , the comparative study of both the dynamics of the few known co - orbitals and candidates , and their discovery circumstances , reveals important additional clues to solve this puzzle . this paper is organized as follows . in section 2 , we briefly discuss the numerical model used in our @xmath3-body simulations . the current status of 2011 qf@xmath2 is reviewed in section 3 . the capture mechanism and stability of 2011 qf@xmath2 are further studied in section 4 . we revisit the cases of crantor and 2010 eu@xmath1 in section 5 . section 6 introduces a few new uranus co - orbital candidates . in section 7 , we discuss our results on the stability of current uranian co - orbitals . we present a practical guide to discover additional objects in section 8 . section 9 summarizes our conclusions . in order to perform a comparative analysis of the orbital evolution of known transient uranian co - orbitals , we solve directly the newtonian equations of motion by means of the hermite integration scheme described by makino ( 1991 ) and implemented by aarseth ( 2003 ) . the adequacy of this integration approach for solar system dynamical studies has been extensively tested by the authors ( de la fuente marcos & de la fuente marcos 2012a , b , c , 2013a , b ) . the standard version of this @xmath3-body sequential code is publicly available from the ioa web sitesverre / web / pages / nbody.htm ] . our calculations include the perturbations by the eight major planets , the moon , the barycentre of the pluto - charon system and the three largest asteroids , ( 1 ) ceres , ( 2 ) pallas and ( 4 ) vesta . for accurate initial positions and velocities , we used the latest heliocentric ecliptic keplerian elements provided by the jet propulsion laboratory ( jpl ) online solar system data service ( giorgini et al . 1996 ) and initial positions and velocities based on the de405 planetary orbital ephemerides ( standish 1998 ) referred to the barycentre of the solar system . the orbits were computed 1 myr forward and backward in time . additional details can be found in de la fuente marcos & de la fuente marcos ( 2012a ) . in addition to the orbital calculations completed using the nominal elements in tables [ elements0 ] and [ elements1 ] , we have performed 50 control simulations with sets of orbital elements obtained from the nominal ones and the quoted uncertainties . these control orbits follow a gaussian distribution in the six - dimensional space of orbital elements and their orbital parameters are compatible with the observations within the 3@xmath8 uncertainties ( see tables [ elements0 ] and [ elements1 ] ) , reflecting the observational incertitude in astrometry . in the integrations , the relative error in the total energy was always as low as @xmath9 or lower after a simulated time of 1 myr . the corresponding error in the total angular momentum is several orders of magnitude smaller . in all the figures , and unless explicitly stated , @xmath10 = 0 coincides with the jd2456800.5 epoch . asteroid 2011 qf@xmath2 was discovered on 2011 august 29 at @xmath11 = 22.6 mag by m. alexandersen observing from mauna kea with the canada france hawaii telescope ( cfht ; alexandersen et al . the observations were made as part of an ongoing survey with cfht , looking for trans - neptunian objects ( tnos ) and minor bodies located between the giant planets . however , they were not submitted to the minor planet center ( mpc ) until march 2013 ( the discovery was announced on 2013 march 18 ; alexandersen et al . the object was identified due to its relatively high proper motion with respect to the background stars . it is comparatively large with @xmath12 = 9.7 mag or a diameter close to 60 km assuming a 5 per cent albedo . its orbit is now reasonably well determined with 29 observations spanning a data - arc of 419 d and it is characterized by @xmath13 = 19.13 au , @xmath14 = 0.18 and @xmath15 = 1081 ( see table [ elements0 ] ) . using these orbital parameters , 2011 qf@xmath2 has been classified as a centaur by both the mpc database and the jpl small - body database . however , its period of revolution around the sun , 83.69 yr at present , is very close to that of uranus , 84.03 yr . that makes 2011 qf@xmath2 a very good candidate to inhabit uranus co - orbital region . 0.3truecm .heliocentric keplerian orbital elements of 2011 qf@xmath2 used in this research . values include the 1@xmath8 uncertainty . the orbit is based on 29 observations spanning a data - arc of 419 d ( epoch = jd2456800.5 , 2014-may-23.0 ; j2000.0 ecliptic and equinox . source : jpl small - body database ) . [ cols="<,^,^ " , ] [ discovery ] ( panel b ) ; as a function of @xmath14 ( panel c ) ; as a function of @xmath16 ( panel d ) ; as a function of @xmath17 ( panel e ) . ] and 12@xmath18 and declination @xmath19(-20 , 20 ) . ] in this paper , we have presented a detailed analysis of the orbital evolution and stability of present - day uranus co - orbitals , focusing on how they got captured in the first place and what makes them dynamically unstable . our calculations show that these objects are submitted to multiple mean motion resonances and exhibit significant secular dynamics characterized by a complex structure that sometimes includes apsidal corotations . the dynamical behaviour analysed here is a typical example of highly nonlinear dynamics , resonances inside a resonance or even resonances inside a resonance inside a resonance . this is best studied using numerical simulations not analytical or semi - analytical work . some of our results verify predictions made by marzari et al . ( 2003 ) after studying the diffusion rate of uranian trojans . we confirm that 2011 qf@xmath2 currently moves inside uranus co - orbital region on a tadpole orbit . the motion of this object is primarily driven by the influence of the sun and uranus , although both jupiter and neptune play a significant role in destabilizing its orbit . the resonant influence of jupiter and neptune was determinant in its capture as uranus co - orbital . the precession of the nodes of 2011 qf@xmath2 , which appears to be controlled by the combined action of saturn and jupiter , marks its evolution and short - term stability . a three - body mean motion resonance is responsible for both its injection into uranus co - orbital region and its ejection from that region . the object will remain as uranus trojan for ( very likely ) less than 1 myr . even if 2011 qf@xmath2 is one of the most stable of the known bodies currently trapped in the 1:1 commensurability with uranus , it is unlikely to be a primordial 1:1 librator . our comparative study of currently known uranus co - orbitals and candidates shows that the candidate 2010 eu@xmath1 is more stable than 2011 qf@xmath2 because of its lower eccentricity ( 0.05 versus 0.18 ) even if 2010 eu@xmath1 has higher orbital inclination ( 148 versus 1080 ) . on the other hand , a new candidate , 2002 vg@xmath4 , is found to exhibit dynamical behaviour very similar to the one discussed for 83982 crantor ( 2002 go@xmath0 ) in de la fuente marcos & de la fuente marcos ( 2013b ) and here . this new candidate is the first identified quasi - satellite of uranus ; crantor will also become a quasi - satellite in the near future . the presence of two objects characterized by similar dynamics and found by chance suggests that others , yet to be discovered , may share their properties . asteroid 1999 hd@xmath5 may signal the edge of uranus co - orbital region . in any case , all these objects have present - day dynamical ages much shorter than that of the solar system ; therefore , they are not members of a hypothetical population of primordial objects trapped in a 1:1 mean motion resonance with uranus since the formation of the solar system . they may be former primordial neptune co - orbitals ( horner & lykawka 2010 ) though . horner & evans ( 2006 ) argued that present - day uranus can not efficiently trap objects in the 1:1 commensurability even for short periods of time . however , the available evidence confirms that , contrary to this view and in spite of the destabilizing role of the other giant planets , uranus still can actively capture temporary co - orbitals , even for millions of years . regarding the issue of stability , both crantor s and 2011 qf@xmath2 s orbital inclinations are within one of the stability islands identified by dvorak et al . ( 2010 ) for the case of trojans . both candidates 2002 vg@xmath4 and 2010 eu@xmath1 move outside the stability islands proposed in that study although they are not trojans . the existence of 1999 hd@xmath5 shows that not only inclination but also eccentricity play an important role in the long - term stability of uranus co - orbitals . a larger sample of uranus co - orbitals is necessary to understand better the complex subject of the stability of these objects although temporary uranian co - orbitals are often submitted to complicated multibody ephemeral mean motion resonances that trigger the switching between the various resonant co - orbital states , making them dynamically unstable . currently available evidence suggests that the small number of known transient uranus co - orbitals may have its origin in observational bias rather than in the strength of the gravitational perturbations by the other giant planets . taking this into account , the number of yet undiscovered transient uranian co - orbitals may likely be as high as that of the neptunian ones . our results can easily be applied to implement improved strategies for discovering additional uranian co - orbitals . we would like to thank the anonymous referee for his / her quick and to - the - point reports , and to s. j. aarseth for providing one of the codes used in this research . this work was partially supported by the spanish ` comunidad de madrid ' under grant cam s2009/esp-1496 . we thank m. j. fernndez - figueroa , m. rego fernndez and the department of astrophysics of the universidad complutense de madrid ( ucm ) for providing computing facilities . most of the calculations and part of the data analysis were completed on the ` servidor central de clculo ' of the ucm and we thank s. cano alsa for his help during this stage . in preparation of this paper , we made use of the nasa astrophysics data system , the astro - ph e - print server and the mpc data server . aarseth s. j. , 2003 , gravitational n - body simulations . cambridge univ . press , cambridge , p. 27 alexandersen m. , kavelaars j. , petit j. , gladman b. , 2013a , mpec circ . , mpec 2013-f19 alexandersen m. , gladman b. , greenstreet s. , kavelaars j. j. , petit j .- m . , gwyn s. , 2013b , science , 341 , 994 beaug c. , ferraz - mello s. , michtchenko t. a. , 2003 , apj , 593 , 1124 bernstein g. , khushalani b. , 2000 , aj , 120 , 3323 de la fuente marcos c. , de la fuente marcos r. , 2012a , mnras , 427 , 728 de la fuente marcos c. , de la fuente marcos r. , 2012b , a&a , 545 , l9 de la fuente marcos c. , de la fuente marcos r. , 2012c , a&a , 547 , l2 de la fuente marcos c. , de la fuente marcos r. , 2013a , mnras , 434 , l1 de la fuente marcos c. , de la fuente marcos r. , 2013b , a&a , 551 , a114 dvorak r. , bazs . , zhou l .- y . , 2010 , celest . astron . , 107 , 51 elliot j. l. et al . , 2005 , aj , 129 , 1117 farago f. , laskar j. , 2010 , mnras , 401 , 1189 fraser w. c. , 2009 , apj , 706 , 119 gacka i. , 2003 , ap&ss , 284 , 1069 gallardo t. , 2006 , icarus , 184 , 29 gallardo t. , 2014 , icarus , 231 , 273 giorgini j. d. et al . , 1996 , baas , 28 , 1158 gomes r. s. , 1998 , aj , 116 , 2590 grav t. et al . , 2011 , apj , 742 , 40 holman m. j. , wisdom j. , 1993 , aj , 2015 , 1987 horner j. , evans n. w. , 2006 , mnras , 367 , l20 horner j. , lykawka p. s. , 2010 , mnras , 402 , 13 innanen k. a. , mikkola s. , 1989 , aj , 97 , 900 jewitt d. c. , 2009 , aj , 137 , 4296 jewitt d. c. , trujillo c. a. , luu j. x. , 2000 , aj , 120 , 1140 kortenkamp s. s. , malhotra r. , michtchenko t. , 2004 , icarus , 167 , 347 kozai y. , 1962 , aj , 67 , 591 laskar j. , 1990 , icarus , 88 , 266 lee m. h. , peale s. j. , 2002 , apj , 567 , 596 lematre a. , henrard j. , 1990 , icarus , 83 , 391 lykawka p. s. , horner j. , 2010 , mnras , 405 , 1375 makino j. , 1991 , apj , 369 , 200 marzari f. , tricarico p. , scholl h. , 2003 , a&a , 410 , 725 masaki y. , kinoshita h. , 2003 , a&a , 403 , 769 meech k. j. et al . , 2002 , mpec circ . , mpec 2002-x26 michel p. , thomas f. , 1996 , a&a , 307 , 310 mikkola s. , innanen k. , 1992 , aj , 104 , 1641 mikkola s. , innanen k. , wiegert p. , connors m. , brasser r. , 2006 , mnras , 369 , 15 milani a. , 1993 , celest . astron . , 57 , 59 milani a. , nobili a. m. , 1985 , celest . mech . , 35 , 269 millis r. l. et al . , 1999 , mpec circ . , mpec 1999-k18 millis r. l. , buie m. w. , wasserman l. h. , elliot j. l. , kern s. d. , wagner r. m. , 2002 , aj , 123 , 2083 morbidelli a. , 2002 , modern celestial mechanics : aspects of solar system dynamics . taylor & francis , london morbidelli a. , levison h. f. , tsiganis k. , gomes r. , 2005 , nature , 435 , 462 murray c. d. , dermott s. f. , 1999 , solar system dynamics . cambridge univ . press , cambridge , p. 97 murray n. , holman m. , 2001 , nature , 410 , 773 murray n. , holman m. , potter m. , 1998 , aj , 116 , 2583 namouni f. , murray c. d. , 2000 , celest . astron . , 76 , 131 namouni f. , christou a. a. , murray c. d. , 1999 , phys . lett . , 83 , 2506 nesvorn d. , dones l. , 2002 , icarus , 160 , 271 nesvorn d. , morbidelli a. , 1998 , aj , 116 , 3029 nesvorn d. , vokrouhlick d. , 2009 , aj , 137 , 5003 nesvorn d. , vokrouhlick d. , morbidelli a. , 2013 , apj , 768 , 45 robutel p. , bodossian j. , 2009 , mnras , 399 , 69 robutel p. , gabern f. , 2006 , mnras , 372 , 1463 shankman c. j. , 2012 , m.sc . thesis , the university of british columbia sheppard s. s. , trujillo c. a. , 2006 , science , 313 , 511 sheppard s. s. , trujillo c. a. , 2010 , apj , 723 , l233 standish e. m. , 1998 , jpl planetary and lunar ephemerides , de405/le405 , interoffice memo . 312.f-98 - 048 , jet propulsion laboratory , pasadena , california stockwell j. n. , 1873 , smithsonian contributions to knowledge , vol . xviii , art . 3 , p. xiv verrier p. e. , evans n. w. , 2008 , mnras , 390 , 1377 verrier p. e. , evans n. w. , 2009 , mnras , 394 , 1721 wiegert p. , innanen k. , mikkola s. , 2000 , aj , 119 , 1978 winter o. c. , murray c. d. , 1997 , a&a , 328 , 399 yoshida f. , nakamura t. , 2005 , aj , 130 , 2900 zhang s .- p . , innanen k. a. , 1988a , aj , 96 , 1989 zhang s .- p . , innanen k. a. , 1988b , aj , 96 , 1995	uranus has three known co - orbitals : 83982 crantor ( 2002 go@xmath0 ) , 2010 eu@xmath1 and 2011 qf@xmath2 . all of them were captured in their current resonant state relatively recently . here , we perform a comparative analysis of the orbital evolution of these transient co - orbitals to understand better how they got captured in the first place and what makes them dynamically unstable . we also look for additional temporary uranian co - orbital candidates among known objects . our @xmath3-body simulations show that the long - term stability of 2011 qf@xmath2 is controlled by jupiter and neptune ; it briefly enters the 1:7 mean motion resonance with jupiter and the 2:1 with neptune before becoming a trojan and prior to leaving its tadpole orbit . during these ephemeral two - body mean motion resonance episodes , apsidal corotation resonances are also observed . for known co - orbitals , saturn is the current source of the main destabilizing force but this is not enough to eject a minor body from the 1:1 commensurability with uranus . these objects must enter mean motion resonances with jupiter and neptune in order to be captured or become passing centaurs . asteroid 2010 eu@xmath1 , a probable visitor from the oort cloud , may have been stable for several myr due to its comparatively low eccentricity . additionally , we propose 2002 vg@xmath4 as the first transient quasi - satellite candidate of uranus . asteroid 1999 hd@xmath5 may signal the edge of uranus co - orbital region . transient uranian co - orbitals are often submitted to complex multibody ephemeral mean motion resonances that trigger the switching between resonant co - orbital states , making them dynamically unstable . in addition , we show that the orbital properties and discovery circumstances of known objects can be used to outline a practical strategy by which additional uranus co - orbitals may be found . celestial mechanics minor planets , asteroids : individual : 1999 hd@xmath5 minor planets , asteroids : individual : 83982 crantor ( 2002 go@xmath0 ) minor planets , asteroids : individual : 2002 vg@xmath4 minor planets , asteroids : individual : 2010 eu@xmath1 minor planets , asteroids : individual : 2011 qf@xmath2 planets and satellites : individual : uranus .
we consider the problem of searching an unstructured database of @xmath0 records for exactly one record which has been specifically marked . this can be rephrased in mathematical terms as an oracle problem as follows : label the records of the database with the integers @xmath1 and denote the label of the unknown marked record by @xmath2 . we are given an oracle which computes the @xmath3 bit binary function @xmath4 defined by @xmath5{cl}1 & \text{if } x = x_{0}\\ & \\ 0 & \text{otherwise}\end{array } \right.\ ] ] we remind the readers that , as a standard oracle idealization , we have no access to the internal workings of the function @xmath6 . it operates simply as a blackbox function , which we can query as many times as we like . but with each such a query comes an associated computational cost . * search problem for an unstructured database . * _ find the record labeled as _ @xmath2 _ _ with the minimum amount of computational work , i.e. , with the minimum number of queries of the oracle _ _ @xmath6__. _ _ from probability theory , we know that if we examine @xmath7 records , i.e. , if we compute the oracle @xmath6 for @xmath7 randomly chosen records , then the probability of finding the record labeled as @xmath2 is @xmath8 . hence , on a classical computer it takes @xmath9 queries to find the record labeled @xmath2 . however , as lov grover so astutely observed , on a quantum computer the search of an unstructured database can be accomplished in @xmath10 steps , or more precisely , with the application of @xmath11 sufficiently local unitary transformations . although this is not exponentially faster , it is a significant speedup . let @xmath12 be a 2 dimensional hilbert space with orthonormal basis @xmath13 and let @xmath14 denote the induced orthonormal basis of the hilbert space @xmath15 from the quantum mechanical perspective , the oracle function @xmath6 is given as a blackbox unitary transformation @xmath16 , i.e. , by @xmath17{ccc}\mathcal{h}\otimes\mathcal{h}_{2 } & \overset{u_{f}}{\longrightarrow } & \mathcal{h}\otimes\mathcal{h}_{2}\\ & & \\ \left\| x\right\rangle \otimes\left\| y\right\rangle & \longmapsto & \left\| x\right\rangle \otimes\left\| f(x)\oplus y\right\rangle \end{array}\ ] ] where ` @xmath18 ' denotes exclusive ` or ' , i.e. , addition modulo 2 . , as defined in sections 10.3 and 10.4 of @xcite . ] instead of @xmath16 , we will use the computationally equivalent unitary transformation @xmath19{cl}-\left\| x_{0}\right\rangle & \text{if \ } x = x_{0}\\ & \\ \left\| x\right\rangle & \text{otherwise}\end{array } \right.\ ] ] that @xmath20 is computationally equivalent to @xmath21 follows from the easily verifiable fact that @xmath22 and also from the fact that @xmath16 can be constructed from a controlled-@xmath20 and two one qubit hadamard transforms . ( for details , please refer to @xcite , @xcite . ) the unitary transformation @xmath20 is actually an * inversion * @xcite in @xmath23 about the hyperplane perpendicular to @xmath24 . this becomes evident when @xmath20 is rewritten in the form @xmath25 where ` @xmath26 ' denotes the identity transformation . more generally , for any unit length ket @xmath27 , the unitary transformation @xmath28 is an inversion in @xmath23 about the hyperplane orthogonal to @xmath29 . we digress for a moment to discuss the properties of the unitary transformation @xmath30 . to do so , we need the following definition . let @xmath27 and @xmath31 be two kets in @xmath23 for which the bracket product @xmath32 is a real number . we define @xmath33 as the sub - hilbert space of @xmath23 spanned by @xmath34 and @xmath31 . we associate with the hilbert space @xmath35 a real inner product space lying in @xmath35 defined by @xmath36 where the inner product on @xmath37 is that induced by the bracket product on @xmath23 . if @xmath27 and @xmath31 are also linearly independent , then @xmath37 is a 2 dimensional real inner product space ( i.e. , the 2 dimensional euclidean plane ) lying inside of the complex 2 dimensional space @xmath35 . let @xmath27 and @xmath31 be two linearly independent unit length kets in @xmath23 with real bracket product ; and let @xmath38 and @xmath39 . then * both @xmath35 and @xmath37 are invariant under the transformations @xmath30 , @xmath40 , and hence @xmath41 , i.e. , @xmath42{lrl}i_{\left\| \psi\right\rangle } \left ( \mathcal{s}_{\mathbb{c}}\right ) = \mathcal{s}_{\mathbb{c } } & \ \text{and\ } & i_{\left\| \psi\right\rangle } \left ( \mathcal{s}_{\mathbb{r}}\right ) = \mathcal{s}_{\mathbb{r}}\\ & & \\ i_{\left\| \chi\right\rangle } \left ( \mathcal{s}_{\mathbb{c}}\right ) = \mathcal{s}_{\mathbb{c } } & \ \text{and\ } & i_{\left\| \chi\right\rangle } \left ( \mathcal{s}_{\mathbb{r}}\right ) = \mathcal{s}_{\mathbb{r}}\\ & & \\ i_{\left\| \psi\right\rangle } i_{\left\| \chi\right\rangle } \left ( \mathcal{s}_{\mathbb{c}}\right ) = \mathcal{s}_{\mathbb{c } } & \ \text{and\ } & i_{\left\| \psi\right\rangle } i_{\left\| \chi\right\rangle } \left ( \mathcal{s}_{\mathbb{r}}\right ) = \mathcal{s}_{\mathbb{r}}\end{array } $ } \ ] ] * if @xmath43 is the line in the plane @xmath37 which passes through the origin and which is perpendicular to @xmath27 , then @xmath44 restricted to @xmath37 is a reflection in ( i.e. , a mbius inversion @xcite about ) the line @xmath45 . a similar statement can be made in regard to @xmath31 . * if @xmath46 is a unit length vector in @xmath37 perpendicular to @xmath27 , then @xmath47 ( hence , @xmath48 is real . ) finally we note that , since @xmath49 , it follows that if @xmath27 is a unit length ket in @xmath23 , and if @xmath50 is a unitary transformation on @xmath23 , then @xmath51 let @xmath52 be the hadamard transform , i.e. , @xmath53 where @xmath54{rr}1 & 1\\ 1 & -1 \end{array } \right)\ ] ] with respect to the basis @xmath55 , @xmath56 . we begin by using the hadamard transform @xmath57 to construct a state @xmath58 which is an equal superposition of all the standard basis states @xmath55 , @xmath56,@xmath59,@xmath60 ( including the unknown state @xmath61 ) , i.e. , @xmath62 both @xmath63 and the unknown state @xmath61 lie in the euclidean plane @xmath64 . our strategy is to rotate within the plane @xmath37 the state @xmath63 about the origin until it is as close as possible to @xmath24 . then a measurement with respect to the standard basis of the state resulting from rotating @xmath63 , will produce @xmath61 with high probability . to achieve this objective , we use the oracle @xmath20 to construct the unitary transformation @xmath65 which by proposition 2 above , can be reexpressed as @xmath66 let @xmath67 and @xmath68 denote unit length vectors in @xmath37 perpendicular to @xmath24 and @xmath69 , respectively . there are two possible choices for each of @xmath67 and @xmath68 respectively . to remove this minor , but nonetheless annoying , ambiguity , we select @xmath67 and @xmath70 so that the orientation of the plane @xmath37 induced by the ordered spanning vectors @xmath58 , @xmath24 is the same orientation as that induced by each of the ordered bases @xmath71 , @xmath24 and @xmath69 , @xmath70 . ( please refer to figure 2 . ) the removal of the above ambiguities is really not essential . however , it does simplify the exposition given below . + figure 2 . the linear transformation @xmath72 is reflection in the line @xmath73 followed by reflection in the line @xmath74 which is the same as rotation by the angle @xmath75 . thus , @xmath72 rotates @xmath58 by the angle @xmath75 toward @xmath24 . we proceed by noting that , by the above proposition 1 , the plane @xmath37 lying in @xmath23 is invariant under the linear transformation @xmath76 , and that , when @xmath76 is restricted to the plane @xmath37 , it can be written as the composition of two inversions , i.e. , @xmath77 in particular , @xmath72 is the composition of two inversions in @xmath37 , the first in the line @xmath78 in @xmath37 passing through the origin having @xmath24 as normal , the second in the line @xmath79 through the origin having @xmath70 as normal . is the intersection of the plane @xmath37 with the hyperplane in @xmath23 orthogonal to @xmath24 . a similar statement can be made in regard to @xmath79 . ] we can now apply the following theorem from plane geometry : if @xmath80 and @xmath81 are lines in the euclidean plane @xmath82 intersecting at a point @xmath83 ; and if @xmath84 is the angle in the plane from @xmath80 to @xmath81 , then the operation of reflection in @xmath80 followed by reflection in @xmath81 is just rotation by angle @xmath75 about the point @xmath83 . let @xmath84 denote the angle in @xmath85 from @xmath78 to @xmath79 , which by plane geometry is the same as the angle from @xmath71 to @xmath63 , which in turn is the same as the angle from @xmath24 to @xmath86 . then by the above theorem @xmath87 is a rotation about the origin by the angle @xmath75 . the key idea in grover s algorithm is to move @xmath63 toward the unknown state @xmath24 by successively applying the rotation @xmath76 to @xmath63 to rotate it around to @xmath24 . this process is called * amplitude amplification. once this process is completed , the measurement of the resulting state ( with respect to the standard basis ) will , with high probability , yield the unknown state @xmath24 . this is the essence of grover s algorithm . but how many times @xmath88 should we apply the rotation @xmath76 to @xmath69 ? if we applied @xmath76 too many or too few times , we would over- or undershoot our target state @xmath24 . we determine the integer @xmath88 as follows : since @xmath89 the state resulting after @xmath7 applications of @xmath76 is @xmath90 \left\| x_{0}\right\rangle + \cos\left [ \left ( 2k+1\right ) \beta\right ] \left\| x_{0}^{\perp}\right\rangle \text { .}\ ] ] thus , we seek to find the smallest positive integer @xmath91 such that @xmath92\ ] ] is as close as possible to @xmath93 . in other words , we seek to find the smallest positive integer @xmath91 such that @xmath94 is as close as possible to @xmath95 . it follows that function instead of the @xmath96 function . ] @xmath97 where `` @xmath96 '' is the function that rounds to the nearest integer . we can determine the angle @xmath84 by noting that the angle @xmath98 from @xmath63 and @xmath24 is complementary to @xmath84 , i.e. , @xmath99 and hence , @xmath100 thus , the angle @xmath84 is given by @xmath101 and hence , @xmath102 in summary , we provide the following outline of grover s algorithm : we complete our summary with the following theorem : with a probability of error function rather than the @xmath96 function , then probability of error becomes @xmath103 . ] @xmath104 grover s algorithm finds the unknown state @xmath24 at a computational cost of @xmath105 the probability of error @xmath106 of finding the hidden state @xmath24 is given by @xmath107 \text { , } \ ] ] where @xmath108{rrl}\beta & = & \sin^{-1}\left ( \frac{1}{\sqrt{n}}\right ) \\ & & \\ k & = & round\left ( \frac{\pi}{4\beta}-\frac{1}{2}\right ) \end{array } \right . \text{,}\ ] ] where `` @xmath96 '' is the function that rounds to the nearest integer . hence , @xmath17{rrl}\frac{\pi}{4\beta}-1\leq k\leq\frac{\pi}{4\beta } & \longrightarrow & \frac { \pi}{2}-\beta\leq\left ( 2k+1\right ) \beta\leq\frac{\pi}{2}+\beta\\ & & \\ & \longrightarrow & \sin\beta=\cos\left ( \frac{\pi}{2}-\beta\right ) \geq \cos\left [ \left ( 2k+1\right ) \beta\right ] \geq\cos\left ( \frac{\pi}{2}+\beta\right ) = -\sin\beta \end{array}\ ] ] thus , @xmath107 \leq\sin^{2}\beta=\sin^{2}\left ( \sin^{-1}\left ( \frac{1}{\sqrt{n}}\right ) \right ) = \frac{1}{n}\ ] ] * the computational cost of the hadamard transform @xmath109 is @xmath110 single qubit operations . the transformations @xmath111 and @xmath20 each carry a computational cost of @xmath112 . + @xmath113 1 is the computationally dominant step . in @xmath113 1 there are @xmath114 iterations . in each iteration , the hadamard transform is applied twice . the transformations @xmath115 and @xmath20 are each applied once . hence , each iteration comes with a computational cost of @xmath116 , and so the total cost of @xmath113 1 is @xmath11 . as an example , we search a database consisting of @xmath117 records for an unknown record with the unknown label @xmath118 . the calculations for this example were made with openquacks , which is an open source quantum simulator maple package developed at umbc and publically available . we are given a blackbox computing device @xmath119{l}$i_{\left\| ? \right\rangle } $ \end{tabular } } } \rightarrow\text{out}\ ] ] that implements as an oracle the unknown unitary transformation @xmath120{rrrrrrrrr}1 & 0 & 0 & 0 & & 0 & 0 & 0 & 0\\ 0 & 1 & 0 & 0 & & 0 & 0 & 0 & 0\\ 0 & 0 & 1 & 0 & & 0 & 0 & 0 & 0\\ 0 & 0 & 0 & 1 & & 0 & 0 & 0 & 0\\ & & & & & & & & \\ 0 & 0 & 0 & 0 & & -1 & 0 & 0 & 0\\ 0 & 0 & 0 & 0 & & 0 & 1 & 0 & 0\\ 0 & 0 & 0 & 0 & & 0 & 0 & 1 & 0\\ 0 & 0 & 0 & 0 & & 0 & 0 & 0 & 1 \end{array } \right)\ ] ] we can not open up the blackbox @xmath121{l}$i_{\left\| ? \right\rangle } $ \end{tabular } } $ } \rightarrow$ ] to see what is inside . so we do not know what @xmath122 and @xmath2 are . the only way that we can glean some information about @xmath2 is to apply some chosen state @xmath29 as input , and then make use of the resulting output . using of the blackbox @xmath123{l}$i_{\left\| ? \right\rangle } $ \end{tabular } } } \rightarrow$ ] as a component device , we construct a computing device @xmath123{l}$-hi_{\left\| 0\right\rangle } hi_{\left\| ? \right\rangle } $ \end{tabular } } } \rightarrow$ ] which implements the unitary operator @xmath124{rrrrrrrrr}-3 & 1 & 1 & 1 & & -1 & 1 & 1 & 1\\ 1 & -3 & 1 & 1 & & -1 & 1 & 1 & 1\\ 1 & 1 & -3 & 1 & & -1 & 1 & 1 & 1\\ 1 & 1 & 1 & -3 & & -1 & 1 & 1 & 1\\ & & & & & & & & \\ 1 & 1 & 1 & 1 & & 3 & 1 & 1 & 1\\ 1 & 1 & 1 & 1 & & -1 & -3 & 1 & 1\\ 1 & 1 & 1 & 1 & & -1 & 1 & -3 & 1\\ 1 & 1 & 1 & 1 & & -1 & 1 & 1 & -3 \end{array } \right)\ ] ] we do not know what unitary transformation @xmath76 is implemented by the device @xmath123{l}$-hi_{\left\| 0\right\rangle } hi_{\left\| ? \right\rangle } $ \end{tabular } } } \rightarrow$ ] because the blackbox @xmath123{l}$i_{\left\| ? \right\rangle } $ \end{tabular } } } \rightarrow$ ] is one of its essential components . * we proceed to loop @xmath126 times in @xmath113 1 . * * on the first iteration , we obtain the unknown state @xmath127 * * on the second iteration , we obtain the unknown state @xmath128 and branch to @xmath113 2 . lomonaco , samuel j. , jr . , * a rosetta stone for quantum mechanics with an introduction to quantum computation : lecture notes for the ams short course on quantum computation , washington , dc , january 2000 , * in * `` quantum computation , '' * edited by s.j . lomonaco , jr . , ams psapm series . ( to appear ) shor , peter w. , * introduction to quantum algorithms * , * lecture notes for the ams short course on quantum computation , washington , dc , january 2000 * , to appear in `` * * quantum computation * * , '' edited by s.j . lomonaco , ams psapm series . ( to appear ) ( quant - ph/0005003 ) vazirani , umesh , * quantum complexity theory * , * lecture notes for the ams short course on quantum computation , washington , dc , january 2000 * , to appear in `` * * quantum computation * * , '' edited by s.j . lomonaco , ams psapm series . ( to appear )	this paper ia a written version of a one hour lecture given on lov grover s quantum database search algorithm . it is based on @xcite , @xcite , and @xcite .
color confinement can be interpreted using the dual version of the superconductivity@xcite-@xcite . in this picture , condensation of color magnetic monopoles is the key concept and leads to squeezing of the color - electric flux between quarks through the dual meissner effect . qcd is reduced into an abelian gauge theory with monopoles in the t hooft abelian gauge@xcite-@xcite . as for the maximally abelian ( ma ) gauge , recent lattice studies suggest that only abelian gauge fields with monopoles are essential for description of nonperturbative phenomena in the infrared region@xcite-@xcite . we study the role of monopoles for confinement using a monopole action . in the confinement phase , only short - range interactions remain among monopoles@xcite due to the _ screening effect _ @xcite by the dual higgs mechanism , and therefore the infrared monopole - current dynamics would be described by a simple action@xcite , @xmath1@xmath2@xmath3@xmath4@xmath5 in the lattice formalism . here , @xmath6@xmath7@xmath8z and @xmath3@xmath7@xmath9@xmath10@xmath11 denote the magnetic current and its self - energy on the lattice , respectively . the monopole current system is generated with the current conservation condition , @xmath12 . here , we update the _ link _ of the monopole current using the metropolis method . we show in fig.1 the monopole density @xmath13@xmath7@xmath14 with the lattice volume @xmath15@xmath2@xmath16 and the clustering parameter @xmath17@xmath18 , where @xmath19 denotes the loop length of the @xmath20-th monopole cluster . with increasing @xmath21 , the monopole density @xmath13 is reduced gradually for @xmath22@xmath23@xmath24 . the clustering parameter is drastically changed at @xmath24 . quantitatively , the critical value of monopole self - energy , @xmath24@xmath25@xmath26 , is close to @xmath27@xmath25@xmath28 , which is derived from the entropy factor @xmath29 in the 4-dimensional self - avoiding random walk@xcite . such a transition is quite similar to the kosterlitz - thouless type transition in 1 + 2-dimensional superconductors . vortex condensation at high temperature in the superconductor corresponds to monopole condensation in the strong coupling region of qcd . in this section , we study confinement properties in the monopole current system with @xmath30@xmath31@xmath32 and @xmath33@xmath2@xmath32 , which is the dual version of qed . to estimate the abelian wilson loop , we introduce the dual gauge field @xmath34 , since the gauge field @xmath35 inevitably includes the singularity as the dirac string in the presence of magnetic monopoles . using the relations @xmath36 and @xmath37@xmath2@xmath38@xmath39@xmath40 the dual gauge field @xmath41 can be described from the monopole current @xmath30 , @xmath42 in the dual landau gauge @xmath43 . the wilson loop is obtained as @xmath44 in estimating the integral in eq.([eq : dual ] ) numerically , we should take _ a finer mesh than original lattice spacing @xmath45 _ to extract @xmath41 correctly . otherwise , the unexpected electric currents appear as the lattice artifact . however , too finer mesh is not necessary , because the original current @xmath46 includes the error in the order of @xmath45 . numerical analyses show that the calculation with @xmath47 is good enough for the estimation of @xmath41 , and hence we take @xmath48 hereafter . on lattices , the dual gauge field @xmath49 is defined in the dual landau gauge ; @xmath50 using the lattice coulomb propagator @xmath51 . the dual version of the abelian field strength @xmath52 is defined by @xmath53 where @xmath54 denotes the backward derivative . as shown in fig.2 , the wilson loop @xmath55 exhibits the area law , which leads to the linear confinement potential . in fig.3 , we show the string tension @xmath56 in the lattice unit . for @xmath21@xmath57@xmath24 , the string tension is finite due to the long monopole currents , the current simulations for @xmath21@xmath58@xmath24 indicate that @xmath59@xmath25@xmath32 in the system without monopole currents . on the other hand , lattice qcd simulations show that the long monopole currents cover the whole system and string tension is finite in the confinement phase even for the large @xmath60 as @xmath61@xmath62@xmath63 on @xmath64 lattices . therefore , the monopole self - energy @xmath21 extracted from qcd must be smaller than @xmath24@xmath7 ln(@xmath65 ) = ln7 in the confinement phase . here , to extract the monopole action , one should use the monopole part in qcd , where only monopole currents exist . since the electric current and the short - range coulomb force can not be expressed only by the monopole currents , it is dangerous to compare the monopole action with the full su(2 ) qcd ( or the abelian part ) ; the former includes monopole currents only , while the latter includes also electric currents and coulomb force . in any case , _ the monopole action is to be extracted from the monopole part in qcd and the monopole self - energy @xmath21 should be smaller than @xmath24=ln7 in the confinement phase . _ the aim of our study is to analyze the theoretical structure of qcd in terms of the monopole degrees of freedom . our studies indicate that the multi - monopole system has a similar structure to qcd in the ma gauge . however , the monopole in abelian - projected qcd has the underlying structure reflecting qcd . for instance , qcd - monopoles would be as _ extended objects _ like hadrons@xcite , because it is not an elementary particle in qcd , but a _ collective mode composed by gluons_. we compare the monopole current system with qcd @xcite especially in terms of the self - energy and the size of monopoles@xcite . in the static frame of the monopole with an intrinsic radius @xmath66 , the spherical magnetic field is created around the monopole as @xmath67 for @xmath68 and @xmath69 for @xmath70@xmath71@xmath66 , with qcd running gauge coupling @xmath72@xcite . for @xmath45@xmath73@xmath66 , the electromagnetic energy observed on the lattice around a monopole is roughly estimated as @xmath74@xmath25@xmath75@xmath25@xmath76 which largely depends on the lattice mesh @xmath45 . hence , the monopole contribution to the lattice action reads @xmath77@xmath2@xmath78@xmath10@xmath79 , where @xmath79 denotes the monopole current length measured in the lattice unit @xmath45 . therefore , @xmath80 is related to the monopole - current coupling @xmath3 and @xmath60@xmath2@xmath81 , @xmath82@xmath25@xmath78@xmath25@xmath83@xmath2@xmath84@xmath10@xmath85 by the screening effect , this relation becomes @xmath82@xmath25@xmath86@xmath10@xmath87@xcite , which is consistent with the numerical result as shown in fig.3 . thus , as long as the mesh is large as @xmath45@xmath73@xmath66 , the lattice monopole action would not need modification by the monopole size effect , and @xmath3 is proportional to @xmath60 . on the other hand , there exists a critical coupling @xmath24@xmath25@xmath887 in the current dynamics . above @xmath24 , the lattice current action provides no monopole condensation and no confinement , while @xmath60@xmath89@xmath90 can be taken in the original qcd keeping the confinement property shown in fig.3 . such a discrepancy between @xmath60 and @xmath21 can be naturally interpreted by introducing the monopole size effect , which should lead the nonlinear correspondence between @xmath21 and @xmath60 . for @xmath21@xmath91@xmath66 , the extended structure of the monopole can be observed on the lattice , and the monopole creates the electromagnetic energy @xmath74 @xmath25@xmath92@xmath2@xmath93 on the lattice , which is almost @xmath45-independent . accordingly , the lattice action should be changed as @xmath1@xmath2@xmath94@xmath10@xmath95 because of the self - avoidness originating from the monopole size @xmath66 . thus , the monopole - current theory becomes nonlocal in the uv region . the monopole size @xmath66 provides a critical scale for the description of qcd in terms of the dual higgs mechanism . in the infrared region as @xmath45@xmath73@xmath66 , qcd can be approximated as a local monopole - current action@xcite , and the qcd vacuum can be regarded as the dual superconductor . on the other hand , in the ultraviolet region as @xmath45<@xmath66 , the monopole theory becomes nonlocal and complicated due to the monopole size effect , and the perturbative qcd would be applicable instead . y. nambu , phys . * d10 * ( 1974 ) 4262 . g. t hooft , nucl . * b190 * ( 1981 ) 455 . z. f. ezawa and a. iwazaki , phys . * d25 * ( 1982 ) 2681 ; * d26 * ( 1982 ) 631 . h. suganuma , s. sasaki and h. toki , nucl . * b435 * ( 1995 ) 207 . h. ichie , h. suganuma and h. toki , phys . + rev.d52 ( 1995 ) 2994 ; * d54 * ( 1996 ) 3382 . m. i. polikarpov , nucl . * b * ( proc.suppl . ) * 53 * ( 1997 ) 134 and references therein . o. miyamura , phys . lett . * b353 * ( 1995 ) 91 . h. suganuma , a. tanaka , s. sasaki and o. miyamura nucl . phys.b ( proc . suppl . ) * 47 * ( 1996 ) 302 . h. suganuma _ et al . _ , this proceedings . j.d.stack , nucl . phys . * b * ( ps ) * 53 * ( 97 ) 524 . h. ichie , h. suganuma and a. tanaka , hep - lat/9709111 , nucl . phys . * a * in press . h. ichie and h. suganuma , proc . of insam symp . 96 , insam report , hep - lat/9709109 .	we study the role of the monopole for color confinement by using the monopole current system . for the self - energy of the monopole current less than ln@xmath0 , long and complicated monopole world - lines appear and the wilson loop obeys the area law , and therefore the monopole current system almost reproduces essential features of confinement properties in the long - distance physics . in the short - distance physics , however , the monopole - current theory would become nonlocal due to the monopole size effect . this monopole size would provide a critical scale of qcd in terms of the dual higgs mechanism .
after over 20 years extensive studies , it has become clear that superconductivity in doped cuprates results when charge carriers pair up into cooper pairs @xcite as in the conventional superconductors @xcite , then these charge carrier cooper pairs condensation reveals the superconducting ( sc ) ground - state . however , as a natural consequence of the unconventional sc mechanism that is responsible for the high sc transition temperatures @xcite , the charge carrier cooper pairs in cuprate superconductors have a dominated d - wave symmetry @xcite . in particular , this d - wave sc state remains one of the cornerstones of our understanding of the physics in cuprate superconductors . however , in spite of the unconventional sc mechanism , the angle - resolved photoemission spectroscopy ( arpes ) experimental results have unambiguously established the bogoliubov - quasiparticle nature of the sharp sc quasiparticle peak in cuprate superconductors @xcite , then the sc coherence of the low energy quasiparticle peak is well described by the simple bardeen - cooper - schrieffer ( bcs ) formalism @xcite with the d - wave sc gap function @xmath0 . in this d - wave case , the characteristic feature is the existence of four nodal points @xmath1 $ ] ( in units of inverse lattice constant ) in the brillouin zone @xcite , where the sc gap function vanishes , then the most physical properties of cuprate superconductors in the sc state are controlled by the quasiparticle excitations around the nodes @xcite . in particular , the key signature of the nodal quasiparticle transport appears in the microwave conductivity @xmath2 @xcite , which is essentially electromagnetic absorption by the quasiparticles excited out of the condensation ( either thermally excited quasiparticles or excitations created by the impurity scattering ) . understanding the role of impurities in cuprate superconductors has taken many years of great effort @xcite . this follows from that the physical properties of cuprate superconductors in the sc state are extreme sensitivity to the impurity effect than the conventional superconductors due to the finite angular - momentum charge carrier cooper pairing @xcite . experimentally , by virtue of systematic studies using the microwave conductivity measurements , some essential features of the evolution of the quasiparticle transport of cuprate superconductors with energy and temperature in the sc state have been established @xcite : ( 1 ) at low temperatures , the experimental results show the existence of the very long - live excitation deep in the sc state , as evidenced by the sharp cusplike energy dependent microwave conductivity spectrum , where the width of the sharp peak is nearly temperature independent , and main behaviors of the microwave conductivity are governed by thermally excited quasiparticles being scattered by impurities or other defects @xcite ; ( 2 ) at low energies , the temperature dependent microwave conductivity increases linearly with increasing temperatures at low temperatures , and reaches a maximum ( a large broad peak ) around intermediate temperature , then decreases with increasing temperatures at high temperatures @xcite . in particular , this broad peak shifts to higher temperatures as the energy is increased . theoretically , an agreement has emerged that the bcs formalism with the d - wave sc gap function is useful in the phenomenological description of the quasiparticle transport of cuprate superconductors in the sc state @xcite , although the sc pairing mechanism is beyond the conventional electron - phonon mechanism . in this case , the microwave conductivity of cuprate superconductors has been phenomenologically discussed in the zero temperature and energy by including the contributions of the vertex corrections @xcite . recently , these discussions have been generalized to study the temperature and energy dependence of the quasiparticle transport of cuprate superconductors in the sc state @xcite . to the best of our knowledge , the microwave conductivity of cuprate superconductors has not been treated starting from a microscopic sc theory , and no explicit calculations of the doping dependence of the microwave conductivity has been made so far . in this paper , we start from the kinetic energy driven sc mechanism @xcite , and study the effect of the extended impurity scatterers on the microwave conductivity of cuprate superconductors . we evaluate explicitly the microwave conductivity of cuprate superconductors within the nodal approximation of the quasiparticle excitations and scattering processes , and qualitatively reproduced some main features of the microwave conductivity measurements on cuprate superconductors in the sc state @xcite . it is shown that there is a cusplike shape of the energy dependent microwave conductivity spectrum . at low temperatures , the microwave conductivity increases linearly with increasing temperatures , and reaches a maximum at intermediate temperature , then decreases with increasing temperatures at high temperatures . in contrast with the dome shape of the doping dependent sc gap parameter , the minimum microwave conductivity occurs around the optimal doping , and then increases in both underdoped and overdoped regimes . this paper is organized as follows . we present the basic formalism in sec . ii , and then discuss the energy , temperature , and doping dependence of the quasiparticle transport of cuprate superconductors in the sc state in sec . iii , where we show that the quasiparticle transport of cuprate superconductors in the sc state can be qualitatively understood within the framework of the kinetic energy driven sc mechanism by considering the effect of the extended impurity scatterers . finally , we give a summary in sec . iv . in cuprate superconductors , the single common feature is the presence of the two - dimensional cuo@xmath3 plane , and it is believed that the unconventional physical properties of cuprate superconductors is closely related to the doped cuo@xmath3 planes @xcite . it has been argued that the @xmath4-@xmath5 model on a square lattice captures the essential physics of the doped cuo@xmath3 plane @xcite , @xmath6 where @xmath7 , @xmath8 is spin operator with @xmath9 as pauli matrices , the constrained electron operator @xmath10 with @xmath11 , and @xmath12 is the chemical potential . in the constrained electron operator , the operators @xmath13 and @xmath14 are to be thought of as operating within the full hilbert space , while the constrained electron operator @xmath15 ( @xmath16 ) does not create ( destroy ) any doubly occupied sites , and therefore represents physical creation ( annihilation ) operator acting in the restricted hilbert space without double electron occupancy @xcite . the strong electron correlation in the @xmath4-@xmath5 model manifests itself by the restriction of the motions of the electrons in the restricted hilbert space without double electron occupancy @xcite , which can be treated properly in analytical calculations within the charge - spin separation ( css ) fermion - spin theory @xcite , where the constrained electron operators are decoupled as @xmath17 and @xmath18 , with the spinful fermion operator @xmath19 describes the charge degree of freedom together with some effects of spin configuration rearrangements due to the presence of the doped hole itself ( dressed holon ) , while the spin operator @xmath20 describes the spin degree of freedom ( spin ) , then the motions of electrons are restricted in the restricted hilbert space without double electron occupancy in analytical calculations . in particular , it has been shown that under the decoupling scheme , this css fermion - spin representation is a natural representation of the constrained electron defined in the restricted hilbert space without double electron occupancy @xcite . moreover , these dressed holon and spin are gauge invariant @xcite , and in this sense , they are real and can be interpreted as the physical excitations @xcite . in this css fermion - spin representation , the low - energy behavior of the @xmath4-@xmath5 model ( 1 ) can be expressed as , @xmath21 with @xmath22 , and @xmath23 is the hole doping concentration . as an important consequence , the kinetic energy term in the @xmath4-@xmath5 model has been transferred as the dressed holon - spin interaction , which reflects that even the kinetic energy term in the @xmath4-@xmath5 hamiltonian has strong coulombic contribution due to the restriction of the motions of electrons in the restricted hilbert space without double electron occupancy . recently , we have developed a kinetic energy driven sc mechanism @xcite based on the css fermion - spin theory @xcite , where the dressed holon - spin interaction from the kinetic energy term in the @xmath4-@xmath5 model ( 2 ) induces the dressed holon pairing state with the d - wave symmetry by exchanging spin excitations , then the electron cooper pairs originating from the dressed holon pairing state are due to the charge - spin recombination , and their condensation reveals the d - wave sc ground - state . moreover , this d - wave sc state is controlled by both sc gap function and quasiparticle coherence , then the maximal sc transition temperature occurs around the optimal doping , and decreases in both underdoped and overdoped regimes . in particular , we have shown that this sc state is the conventional bcs like with the d - wave symmetry @xcite , so that the basic bcs formalism with the d - wave sc gap function is still valid in quantitatively reproducing all main low energy features of the arpes experimental measurements on cuprate superconductors , although the pairing mechanism is driven by the kinetic energy by exchanging spin excitations , and other exotic magnetic scattering @xcite is beyond the bcs formalism . following our previous discussions @xcite , the full dressed holon green s function in the sc state can be obtained in the nambu representation as , @xmath24 where @xmath25 is the unit matrix , the renormalized dressed holon excitation spectrum @xmath26 , with the mean - field ( mf ) dressed holon excitation spectrum @xmath27 , the spin correlation function @xmath28 , @xmath29 , @xmath30 is the number of the nearest neighbor sites , the renormalized dressed holon d - wave pair gap function @xmath31 , where the effective dressed holon d - wave pair gap function @xmath32 with @xmath33 , and the dressed holon quasiparticle spectrum @xmath34 , while the dressed holon quasiparticle coherent weight @xmath35 and effective dressed holon gap parameters @xmath36 are determined by the following two equations @xcite , @xmath37 respectively , where @xmath38 $ ] , @xmath39 $ ] , @xmath40 , @xmath41 , the dressed holon s particle - hole parameter @xmath42 , the spin correlation function @xmath43 , @xmath44 [ 1 - 2n_{f}(e_{h{\bf k}})]+e_{h{\bf k}}[n_{b } ( \omega_{{\bf p ' } } ) n_{b}(-\omega_{{\bf p}})+n_{b } ( \omega_{{\bf p } } ) n_{b}(-\omega_{{\bf p'}})]$ ] , @xmath45[1 - 2n_{f } ( e_{h{\bf k}})]+e_{h{\bf k}}[n_{b } ( \omega_{{\bf p'}})n_{b}(\omega_{{\bf p}})+n_{b}(-\omega_{{\bf p ' } } ) n_{b}(-\omega_{{\bf p}})]$ ] , @xmath46- n_{b}(\omega_{{\bf p}})n_{b}(-\omega_{{\bf p ' } } ) $ ] , @xmath47-n_{b}(\omega_{{\bf p ' } } ) n_{b } ( -\omega_{{\bf p}})$ ] , @xmath48 + n_{b } ( \omega_{{\bf p}})n_{b}(\omega_{{\bf p'}})$ ] , @xmath49+n_{b}(-\omega_{{\bf p}})n_{b}(-\omega_{{\bf p ' } } ) $ ] , @xmath50 and @xmath51 are the boson and fermion distribution functions , respectively , and the mf spin excitation spectrum , @xmath52,\end{aligned}\ ] ] where @xmath53 , @xmath54 , and the spin correlation functions @xmath55 and @xmath56 . in order to satisfy the sum rule of the correlation function @xmath57 in the case without the antiferromagnetic long - range - order , an important decoupling parameter @xmath58 has been introduced in the above mf calculation @xcite , which can be regarded as the vertex correction . these two equations in eqs . ( 4a ) and ( 4b ) must be solved simultaneously with other self - consistent equations , then all order parameters , decoupling parameter @xmath58 , and chemical potential @xmath12 are determined by the self - consistent calculation . in the css fermion - spin theory @xcite , the electron green s function is a convolution of the spin green s function and dressed holon green s function . following our previous discussions @xcite , we can obtain the electron diagonal and off - diagonal green s functions in the sc state as , @xmath59 \left({u^{2}_{h{\bf p+k}}\over\omega+e_{h{\bf p+k}}-\omega_{{\bf p } } } + { u^{2}_{h{\bf p+k}}\over\omega+e_{h{\bf p+k}}+\omega_{{\bf p } } } \right . \right .\nonumber \\ & + & \left . { v^{2}_{h{\bf p+k}}\over\omega - e_{h{\bf p+k}}+ \omega_{{\bf p}}}+{v^{2}_{h{\bf p+k}}\over\omega - e_{h{\bf p+k } } -\omega_{{\bf p}}}\right)+{\rm tanh}[{1\over 2}\beta e_{h{\bf p+k } } ] \left({u^{2}_{h{\bf p+k}}\over\omega+e_{h{\bf p+k}}+\omega_{{\bf p } } } \right . \nonumber \\ & -&{u^{2}_{h{\bf p+k}}\over\omega+e_{h{\bf p+k}}-\omega_{{\bf p } } } + \left . \left . { v^{2}_{h{\bf p+k}}\over\omega - e_{h{\bf p+k}}- \omega_{{\bf p}}}-{v^{2}_{h{\bf p+k}}\over\omega - e_{h{\bf p+k } } + \omega_{{\bf p } } } \right ) \right \ } , \end{aligned}\ ] ] @xmath60\left({1\over\omega - e_{h{\bf p+k}}-\omega_{{\bf p } } } + { 1\over \omega - e_{h{\bf p+k}}+\omega_{{\bf p } } } \right .\right .\nonumber \\ & - & \left . { 1\over \omega + e_{h{\bf p+k}}+\omega_{{\bf p}}}- { 1\over \omega + e_{h{\bf p+k}}-\omega_{{\bf p } } } \right ) + { \rm tanh}[{1\over 2}\beta e_{h{\bf p+k}}]\left ( { 1\over \omega - e_{h{\bf p+k } } -\omega_{{\bf p } } } \right .\nonumber \\ & -&\left .\left . { 1\over \omega - e_{h{\bf p+k}}+\omega_{{\bf p } } } -{1\over \omega + e_{h{\bf p+k}}+\omega_{{\bf p}}}+ { 1\over \omega + e_{h{\bf p+k}}-\omega_{{\bf p } } } \right ) \right \},\end{aligned}\ ] ] respectively , where the dressed holon quasiparticle coherence factors @xmath61 and @xmath62 . these convolutions of the spin green s function and dressed holon diagonal and off - diagonal green s functions reflect the charge - spin recombination @xcite . since the spins center around the @xmath63 $ ] point in the mf level @xcite , then the main contributions for the spins comes from the @xmath63 $ ] point . in this case , the electron diagonal and off - diagonal green s functions in eqs . ( 6a ) and ( 6b ) can be approximately reduced as the bcs formalism with the d - wave sc gap function in terms of @xmath64}\sim 0 $ ] and the equation @xcite @xmath65 , @xmath66 where the electron quasiparticle coherent weight @xmath67 , the electron quasiparticle coherence factors @xmath68 and @xmath69 , with @xmath70 , @xmath71 , and @xmath72 $ ] , @xmath73 , and electron quasiparticle spectrum @xmath74 , i.e. , the hole - like dressed holon quasiparticle coherence factors @xmath75 and @xmath76 and hole - like dressed holon quasiparticle spectrum @xmath77 have been transferred into the electron quasiparticle coherence factors @xmath78 and @xmath79 and electron quasiparticle spectrum @xmath80 , respectively , by the convolutions of the spin green s function and dressed holon green s functions due to the charge - spin recombination . this means that within the kinetic energy driven sc mechanism , the dressed holon pairs condense with the d - wave symmetry in a wide range of the hole doping concentration , then the electron cooper pairs originating from the dressed holon pairing state are due to the charge - spin recombination , and their condensation automatically gives the electron quasiparticle character . for the convenience in the following discussions , these electron green s functions in eq . ( 7 ) in the sc state can be expressed in the nambu representation as , @xmath81 with the helps of this bcs formalism under kinetic energy driven sc mechanism , now we can discuss the effect of the extended impurity scatterers on the quasiparticle transport in cuprate superconductors . in the presence of impurities , the unperturbed electron green s function in eq . ( 8) is dressed via the impurity scattering @xcite , @xmath82 with the self - energy function @xmath83 . it has been shown that all but the scalar component of the self - energy function can be neglected or absorbed into @xmath84 @xcite . in this case , the dressed electron green s function in eq . ( 9 ) can be explicitly rewritten as , @xmath85\tau_{0}+\bar{\delta}_{z}({\bf k})\tau_{1}+ [ \bar{\varepsilon}_{{\bf k}}+\sigma_{3}({\bf k},\omega)]\tau_{3 } \over [ \omega-\sigma_{0}({\bf k},\omega)]^{2 } -\bar{\varepsilon}^{2}_{{\bf k}}-\bar{\delta}^{2}_{z}({\bf k})}.\end{aligned}\ ] ] based on the phenomenological d - wave bcs - type electron green s function @xcite , the energy and temperature dependence of the microwave conductivity of cuprate superconductors has been fitted @xcite , where the electron self - energy functions @xmath86 and @xmath87 have been treated within the framework of the t - matrix approximation . following their discussions @xcite , the electron self - energy function @xmath88 can be obtained approximately as , @xmath89 with @xmath90 is the impurity concentration , and @xmath91 is the diagonal element of the t - matrix , @xmath92 where @xmath93 is the impurity scattering potential . as mentioned in sec . i , there is no gap to the quasiparticle excitations at the four nodes for the d - wave sc state of cuprate superconductors , therefore the quasiparticles are generated only around these four nodes . it has been shown @xcite that this characteristic feature is very useful when considering the impurity scattering , since the initial and final momenta of a scattering event must always be approximately equal to the @xmath94-space location of one of the four nodes in the zero temperature and zero energy , while the impurity scattering potential @xmath95 varies slowly over the area of a node . in this case , a general scattering potential @xmath93 need only be evaluated in three possible case : the intranode impurity scattering @xmath96 ( @xmath97 and @xmath98 at the same node ) , the adjacent - node impurity scattering @xmath99 ( @xmath100 and @xmath98 at the adjacent nodes ) , and the opposite - node impurity scattering @xmath101 ( @xmath97 and @xmath102 at the opposite nodes ) , then the impurity scattering potential @xmath93 in the t - matrix can be effectively reduced as @xcite , @xmath103 at the zero temperature and zero energy , these nodes reduce to points . in this case , this nodal approximation for the impurity potential can reproduce any impurity potential . however , at finite temperatures and energies , there is a limitation on the forward scattering character of the impurity potential because this nodal approximation assumes the brillouin zone quadrant around a particular node @xcite . it has been shown @xcite that although the strict forward scattering limit can therefore not be reached at finite temperatures and energies , this nodal approximation is still appropriate to treat the intermediate range scatters . therefore in the following discussions , we employ the simplified impurity scattering potential in eq . ( 13 ) to study the impurity scattering effect on the quasiparticle transport of cuprate superconductors . substituting eq . ( 13 ) into eq . ( 12 ) , the t - matrix can be obtained as a @xmath104-matrix around the nodal points , @xmath105 where @xmath106 is the integral of the electron green s function , and can be obtained as , @xmath107 with @xmath108 and @xmath109 are given by , @xmath110 and the self - energy functions @xmath111 and @xmath111 are evaluated as , @xmath112^{2}- [ \tilde{g}_{i0}(\omega)v_{13}]^{2}}+{\tilde{g}_{i0}(\omega ) ( v^{-}_{123})^{2}\over [ 1-\tilde{g}_{i3}(\omega)v^{-}_{123}]^{2}- [ \tilde{g}_{i0}(\omega)v^{-}_{123}]^{2}}\right . \nonumber \\ & + & \left . { \tilde{g}_{i0}(\omega)(v^{+}_{123})^{2}\over [ 1- \tilde{g}_{i3}(\omega)v^{+}_{123}]^{2}-[\tilde{g}_{i0}(\omega ) v^{+}_{123}]^{2}}\right ) , \\ \sigma_{3}(\omega)&=&{\rho_{i}\over 4}\left ( { 2v_{13}[1- \tilde{g}_{i3}(\omega ) v_{13}]\over [ 1- \tilde{g}_{i3}(\omega)v_{13}]^{2}- [ \tilde{g}_{i0}(\omega ) v_{13}]^{2}}+{v^{-}_{123}[1-\tilde{g}_{i3 } ( \omega)v^{-}_{123}]\over [ 1-\tilde{g}_{i3}(\omega)v^{-}_{123}]^{2}- [ \tilde{g}_{i0}(\omega)v^{-}_{123}]^{2}}\right . \nonumber\\ & + & \left . { v^{+}_{123}[1-\tilde{g}_{i0}(\omega)v^{+}_{123}]\over [ 1-\tilde{g}_{i3}(\omega)v^{+}_{123}]^{2}- [ \tilde{g}_{i0}(\omega ) v^{+}_{123}]^{2 } } \right ) , \end{aligned}\ ] ] where @xmath113 , @xmath114 , and @xmath115 . in the framework of the linear response theory , the microwave conductivity of cuprate superconductors can be calculated by means of the kubo formula as @xcite , @xmath116 with @xmath117 is the electron current - current correlation function in the sc state , and can be expressed as , @xmath118 in the css fermion - spin representation @xcite , the electron polarization operator can be evaluated as , @xmath119 then within the @xmath4-@xmath5 model ( 2 ) , the current density of electrons is obtained by the time derivation of this polarization operator using the heisenberg s equation of motion as , @xmath120=i{1\over 2}et\sum_{i\hat{\eta}}\hat{\eta } ( h^{\dagger}_{i+\hat{\eta}\uparrow } h_{i\uparrow}s_{i}^{+ } s^{-}_{i+\hat{\eta}}+ h^{\dagger}_{i+\hat{\eta}\downarrow } h_{i\downarrow}s^{\dagger}_{i}s^{-}_{i+\eta})=-i{1\over 2}et \sum_{i\eta\sigma}\hat{\eta } \hat{c}^{\dagger}_{i\sigma } \hat{c}_{i+\hat{\eta}\sigma}\approx { ev_{f}\over \sqrt{2 } } \sum_{{\bf k}\sigma}{\bf k } \hat{c}^{\dagger}_{{\bf k}\sigma } \hat{c}_{{\bf k}\sigma},\end{aligned}\ ] ] with @xmath121 is the electron velocity at the nodal points . according to this current density ( 21 ) , the current - current correlation function in eq . ( 19 ) can be obtained as , @xmath122,\end{aligned}\ ] ] with @xmath123 is the vertex function , and can be evaluated explicitly as the sum of ladder diagrams , @xmath124 where the kernel function @xmath125 is expressed as , @xmath126\over [ 1-(l_{a}i_{0}^{(0)}+l_{b } i^{(0)}_{3})][1-(l_{a}i_{3}^{(3)}+l_{b}i^{(3)}_{0})]-[l_{a } i_{0}^{(3)}+l_{b}i^{(3)}_{3}][l_{a}i_{3}^{(0)}+l_{b}i^{(0)}_{0}]},\end{aligned}\ ] ] with the functions , @xmath127,\\ l_{b}(\omega,\omega')&=&[t_{11}^{(0)}(\omega)t_{11}^{(3)}(\omega+ \omega')+t_{11}^{(3)}(\omega)t_{11}^{(0)}(\omega+\omega ' ) -t_{13}^{(0)}(\omega)t_{13}^{(3)}(\omega+\omega ' ) -t_{13}^{(3)}(\omega)t_{13}^{(0)}(\omega+\omega')],\end{aligned}\ ] ] while the functions @xmath128 and @xmath129 are give by , @xmath130 substituting eq . ( 23 ) into eq . ( 18 ) , the microwave conductivity of cuprate superconductors is obtained explicitly as , @xmath131.\end{aligned}\ ] ] we emphasize that based on the nodal approximation of the quasiparticle excitations and scattering processes @xcite , this microwave conductivity of cuprate superconductors in eq . ( 27 ) is obtained within the kinetic energy driven sc mechanism , although its expression is similar to that obtained within the phenomenological bcs formalism with the d - wave sc gap function @xcite . in cuprate superconductors , although the values of @xmath5 and @xmath4 is believed to vary somewhat from compound to compound , however , as a qualitative discussion , the commonly used parameters in this paper are chosen as @xmath132 , with an reasonably estimative value of @xmath133k . we are now ready to discuss the doping , energy , and temperature dependence of the quasiparticle transport of cuprate superconductors in the sc state with extended impurities . we have performed a calculation for the microwave conductivity @xmath2 in eq . ( 27 ) , and the results of @xmath2 as a function of energy with temperature @xmath134k ( solid line ) , @xmath135k ( dashed line ) , @xmath136k ( dash - dotted line ) , and @xmath137k ( dotted line ) under the slightly strong impurity scattering potential with @xmath138 , @xmath139 , and @xmath140 at the impurity concentration @xmath141 for the doping concentration @xmath142 are plotted in fig . 1 in comparison with the corresponding experimental results of cuprate superconductors in the sc state @xcite ( inset ) . obviously , the energy evolution of the microwave conductivity of cuprate superconductor @xcite is qualitatively reproduced . in particular , a low temperature cusplike shape of the microwave conductivity is obtained for cuprate superconductors in the presence of the impurity scattering . at the low energy regime ( @xmath143 ) , this low temperature microwave conductivity @xmath2 rises rapidly from the universal zero - temperature limit to a much larger microwave conductivity . however , this low temperature microwave conductivity @xmath2 becomes smaller and varies from weakly energy dependence at the intermediate energy regime ( @xmath144 ) , to the almost energy independence at the high energy regime ( @xmath145 ) . k ( solid line ) , @xmath135k ( dashed line ) , @xmath146k ( dash - dotted line ) , and @xmath137k ( dotted line ) at @xmath141 for @xmath132 , @xmath138 , @xmath139 , and @xmath140 in @xmath142 . inset : the corresponding experimental result of cuprate superconductors in the sc state taken from ref . [ 14 ] . ] k and @xmath147ghz at @xmath141 for @xmath132 , @xmath138 , @xmath139 , and @xmath140 . the dashed line is the corresponding result of the superconducting gap parameter . ] for a better understanding of the physical properties of the microwave conductivity @xmath2 in cuprate superconductors , we have studied the doping evolution of the microwave conductivity , and the result of @xmath2 as a function of doping with temperature @xmath134k and energy @xmath147ghz under the slightly strong impurity scattering potential with @xmath138 , @xmath139 , and @xmath140 at the impurity concentration @xmath141 is plotted in fig . 2 ( solid line ) . for comparison , the corresponding result of the sc gap parameter of cuprate superconductors is also shown in the same figure ( dashed line ) . our result shows that in contrast to the dome shape of the doping dependent sc gap parameter @xcite , the microwave conductivity @xmath2 decreases with increasing doping in the underdoped regime , and reaches a minimum in the optimal doping , then increases in the overdoped regime . this doping dependent behavior of the low energy microwave conductivity @xmath2 at low temperatures is also qualitatively consistent with the universal microwave conductivity limit @xcite @xmath148 at low energy as temperature @xmath149 , if this sc gap parameter @xmath150 in the phenomenological bcs formalism @xcite has the similar dome shape doping dependence . ghz ( solid line ) , @xmath151ghz ( long dashed line ) , @xmath152ghz ( dash - dotted line ) , @xmath153ghz ( dotted line ) , and @xmath154ghz ( short dashed line ) at @xmath141 for @xmath132 , @xmath138 , @xmath139 , and @xmath140 in @xmath142 . inset : the corresponding experimental result of cuprate superconductors in the sc state taken from ref . [ 14 ] . ] in the above discussions , we mainly study the effect of the extended impurity scatterers on the quasiparticle transport at low temperatures ( @xmath155 ) and low energies ( @xmath156 ) . now we discuss the temperature dependence of the quasiparticle transport of cuprate superconductors , where @xmath157 may approach to @xmath158 from low temperature side . in this case , it has been shown @xcite that the inelastic scattering process should be considered at higher temperatures , such as the quasiparticle - quasiparticle scattering . this is followed from the fact that the inelastic quasiparticle - quasiparticle scattering process is suppressed at low temperatures due to the large sc gap parameter in the quasiparticle excitation spectrum . however , the contribution from this inelastic quasiparticle - quasiparticle scattering process is increased rapidly when @xmath157 approaches to @xmath158 from low temperature side , since there is a small sc gap parameter near @xmath158 . in particular , it has been pointed out @xcite that the contribution from the quasiparticle - quasiparticle scattering process to the transport lifetime is exponentially suppressed at low temperatures , and therefore the effect of this inelastic quasiparticle - quasiparticle scattering can be considered by adding the inverse transport lifetime @xcite @xmath159 to the imaginary part of the self - energy function @xmath111 in eq . ( 17a ) , then the total self - energy function @xmath160 can be expressed as @xcite , @xmath161^{-1},\end{aligned}\ ] ] with @xmath162 has been chosen as @xmath163^{-1}=91.35(t-0.005)^{4}j$ ] . using this total self - energy function @xmath164 to replace @xmath111 in eq . ( 27 ) , we have performed a calculation for the microwave conductivity @xmath2 , and the results of @xmath2 as a function of temperature @xmath157 with energy @xmath165ghz ( solid line ) , @xmath166ghz ( long dashed line ) , @xmath167ghz ( dash - dotted line ) , @xmath168ghz ( dotted line ) , and @xmath169ghz ( short dashed line ) under the slightly strong impurity scattering potential with @xmath138 , @xmath139 , and @xmath140 at the impurity concentration @xmath141 for the doping concentration @xmath142 are plotted in fig . 3 in comparison with the corresponding experimental results of cuprate superconductors in the sc state @xcite ( inset ) . our results show that the temperature dependent microwave conductivity @xmath2 increases rapidly with increasing temperatures to a large broad peak around temperature @xmath170k for energy @xmath165ghz , and then falls roughly linearly . however , this broad peak in the microwave conductivity spectrum resulting from the temperature dependent impurity scattering rate in eq . ( 28 ) is energy dependent , and moves to higher temperatures with increasing energies , in qualitative agreement with the experimental data @xcite . within the framework of the kinetic energy driven d - wave cuprate superconductivity @xcite , our present results of the energy and temperature dependence of the microwave conductivity by considering the effect of the extended impurity scatterers are qualitatively similar to the earlier attempts to fit the experimental data by using a phenomenological d - wave bcs formalism @xcite . establishing this agreement is important to confirming the nature of the sc phase of cuprate superconductors as the d - wave bcs - like sc state within the kinetic energy driven sc mechanism . it has been shown @xcite that there are some subtle differences for different families of cuprate superconductors , and these subtle differences may be induced by the other effects except the impurity scattering . however , we in this paper are primarily interested in exploring the general notion of the effect of the extended impurity scatterers on the kinetic energy driven cuprate superconductors in the sc state . the qualitative agreement between the present theoretical results and experimental data also show that the presence of impurities has a crucial effect on the microwave conductivity of cuprate superconductors . in conclusion we have shown very clearly in this paper that if the effect of the extended impurity scatterers is taken into account in the framework of the kinetic energy driven d - wave superconductivity , the microwave conductivity of the @xmath4-@xmath5 model calculated based on the nodal approximation of the quasiparticle excitations and scattering processes per se can correctly reproduce some main features found in the microwave conductivity measurements on cuprate superconductor in the sc state @xcite , including the energy and temperature dependence of the microwave conductivity spectrum . the theory also predicts a v - shaped doping dependent microwave conductivity , which is in contrast with the dome shape of the doping dependent sc gap parameter , and therefore should be verified by further experiments . this work was supported by the national natural science foundation of china under grant no . 10774015 , and the funds from the ministry of science and technology of china under grant nos . 2006cb601002 and 2006cb921300 . p. j. turner , r. harris , saeid kamal , m. e. hayden , d. m. broun , d. c. morgan , a. hosseini , p. dosanjh , g. k. mullins , j. s. preston , ruixing liang , d. a. bonn , and w. n. hardy , phys . lett . * 90 * , 237005 ( 2003 ) . d. a. bonn , r. liang , t. m. riseman , d. j. baar , d. c. morgan , k. zhang , p. dosanjh , t. l. duty , a. macfarlane , g. d. morris , j. h. brewer , w. n. hardy , c. kallin , and a. j. berlinsky , phys . b * 47 * , 11314 ( 1993 ) . r. harris , p. j. turner , saeid kamal , a. r. hosseini , p. dosanjh , g. k. mullins , j. s. bobowski , c. p. bidinosti , d. m. broun , ruixing liang , w. n. hardy , and d. a. bonn , phys . b * 74 * , 104508 ( 2006 ) . shiping feng , phys . rev . b * 68 * , 184501 ( 2003 ) ; shiping feng , tianxing ma , and huaiming guo , physica c * 436 * , 14 ( 2006 ) ; shiping feng and tianxing ma , in _ superconductivity research horizons _ , edited by eugene h. peterson ( nova science publishers , new york , 2007 ) chapter 5 , p. 129 , and references therein . m. arai , t. nishijima , y. endoh , t. egami , s. tajima , k. tomimoto , y. shiohara , m. takahashi , a. garret , and s. m. bennington , phys . 83 * , 608 ( 1999 ) ; s. m. hayden , h. a. mook , p. dai , t. g. perring , and f. dogan , nature * 429 * , 531 ( 2004 ) ; c. stock , w. j. buyers , r. a. cowley , p. s. clegg , r. coldea , c. d. frost , r. liang , d. peets , d. bonn , w. n. hardy , and r. j. birgeneau , phys . b * 71 * , 024522 ( 2005 ) .	within the framework of the kinetic energy driven superconducting mechanism , the effect of the extended impurity scatterers on the quasiparticle transport of cuprate superconductors in the superconducting state is studied based on the nodal approximation of the quasiparticle excitations and scattering processes . it is shown that there is a cusplike shape of the energy dependent microwave conductivity spectrum . at low temperatures , the microwave conductivity increases linearly with increasing temperatures , and reaches a maximum at intermediate temperature , then decreases with increasing temperatures at high temperatures . in contrast with the dome shape of the doping dependent superconducting gap parameter , the minimum microwave conductivity occurs around the optimal doping , and then increases in both underdoped and overdoped regimes .
0.1 truecm the symmetrical and topological properties of graphene layers and their stacks have opened up new fields of physical research in the past decade . according to the hexagonal lattice symmetry , a number of graphene layers can be so stacked that any adjacent two are shifted from each other in either direction along the armchair orientation . among others , abc - stacking configuration is especially noticeable in that all the layers are shifted in a unique direction . a non - trivial topological phase is accommodated therein,@xcite giving rise to surface states that are localized at the outermost layers and coupled to the rest , so - called bulk states.@xcite this is distinct from ab - stacking configuration,@xcite whose trivial topology does not render any surface states.@xcite in general , these surface states are characterized by flat band dispersion about the zero energy . it is well known that the flat surface subbands in abc - stacked @xmath0-layer graphene can be obtained from a chiral effective hamiltonian with chirality @xmath1.@xcite corresponding quantum hall effects ( qhes ) have been predicted@xcite and observed at least for @xmath2.@xcite apart from the surface subbands , the bulk subbands are gapped and seem to be intricate . they evolve to be even more complex with increasing @xmath0,@xcite while the surface subbands are more and more flat.@xcite so far , the bulk subbands remain unsolved , probably because no attractive properties are expected from them . there is , however , something remarkable in their evolution to the three - dimensional ( @xmath3d ) realm as is highlighted in recent research . in the @xmath3d counterpart of abc - stacked @xmath0-layer graphene , namely , rhombohedral graphite ( rg ) , the bulk atomic lattice is usually modelled by infinitely many layers . there is a crossover of lattice symmetry when @xmath0 goes from being finite to being infinite , with a two - atom rhombohedral lattice appearing in rg.@xcite the resulting @xmath4 chiral hamiltonian leads to a @xmath3d dirac cone structure,@xcite where the dirac points ( dps ) form continuous lines in a spiraling fashion in association with a sausage - like fermi surface.@xcite such gapless bulk subbands in rg can be used to explain the @xmath3d qhe observed in graphite,@xcite in which the transport takes place in each layer with @xmath5 charge carriers and the quantization of the conductivity is the same as in monolayer graphene.@xcite this is a general attribute of graphite as a weakly coupled layered system.@xcite the interlayer hopping brings about the described @xmath3d features of the bulk subbands in rg , whereas its much smaller ratio to the intralayer hopping is also responsible for the @xmath6d transport character revealed in the @xmath3d qhe . indeed , the dimensionality of graphite can be effectively determined by the interplay between the interlayer hopping and the interlayer coulomb interaction,@xcite the latter being absent from the single - particle hamiltonian . in this regard , the @xmath3d nature of the dirac cone structure in rg has been verified by the existence of optical magnetoplasmons in the presence of that interplay.@xcite the non - trivial topology uniquely endowed in the abc stack of graphene layers leads to a bulk - surface correspondence in rg,@xcite just as the bulk - edge correspondence in monolayer graphene with zigzag edges.@xcite the topological stability against perturbative disturbance such as impurities and crystal deformations is ensured by the dp spirals , which are nodal lines just as the dps are nodal points in monolayer graphene.@xcite consequently , rg possesses metallic flat surface subbands while being semimetallic in the bulk with a 3d dirac cone structure . as such , rg can be compared to certain topological insulators , e.g. , bi@xmath7se@xmath8 and bi@xmath7te@xmath8 , which are also abc - stacked layered materials.@xcite since the physics arising from abc - stacking configuration is so rich and deep , it is of interest to investigate the band structure of abc - stacked @xmath0-layer graphene ranging from @xmath6d to @xmath3d realm . recently , first - principle calculations simultaneously taking into account the surface and the bulk states have been performed with a number of layers up to @xmath9 , showing that the bulk subbands are still gapped.@xcite a recursive analysis has been achieved for infinite @xmath0,@xcite and the resulting density of states ( dos ) and landau level spectrum are the same as in monolayer graphene in agreement with the @xmath3d dirac cone structure.@xcite it is therefore reasonable to envisage a dimensional crossover from the gapped bulk subbands to the @xmath3d dirac cone structure . nevertheless , to our knowledge the evolution has not been ever explicitly studied in spite that similar studies of ab stacks have been conducted.@xcite in this paper , we analytically calculate the bulk subbands in abc - stacked @xmath0-layer graphene by deriving and using an expression for arbitrary @xmath0 . we investigate their evolution in the increase of @xmath0 and dimensional crossover in the infinite limit of @xmath0 . in sec . we describe the model and construct a @xmath6d non - perturbative effective hamiltonian closed in the bulk subspace so as to treat the coupling to the surface subbands . the obtained bulk subbands are certain @xmath6d projections from the @xmath3d dirac cone structure in rg . in sec . , we describe how to resolve a second - order recursion involved in the eigenvalue problem of the effective hamiltonian . then we derive the expression of the eigenenergies and bulk dos . in sec . , at first we analyze the bulk subbands and demonstrate their evolution together with the well understood surface subbands . then we elucidate the dimensional crossover . finally , we give in sec . a summary with an outlook . 0.6 truecm 0.1 truecm the lattice of abc - stacked @xmath0-layer graphene is schematically shown in fig . 1(a ) , where graphene layers infinitely extended in the @xmath10 plane and stacked along the @xmath11 direction are labeled by @xmath12 , @xmath6 , ... , @xmath0 , with the lattice constant of a single layer @xmath13 nm and interlayer distance @xmath14 nm . carbon atoms in this configuration are classified into to two sets . one set contains the sublattices of surface atoms ( @xmath15 and @xmath16 ) ; the other contains the sublattices of the bulk atoms . each bulk atom is vertically bonded with another one that is located in either the adjacent upper or the adjacent lower layer . in the two outermost layers the surface atoms are free from such bonding . a solid view in fig . 1(b ) illustrates that infinitely many abc - stacked graphene layers are taken to model the bulk of rg , which has a two - atom rhombohedral primitive unit cell . also shown in fig . 1(b ) , for reference , is an alternatively set hexagonal non - primitive unit cell of triple volume , which is also the primitive unit cell in abc - stacked trilayer graphene . from abc - stacked graphene@xcite to rg,@xcite there always exist coupled surface and bulk subbands , respectively localized at the surface and the bulk atomic sublattices . the conventional tight - binding ( tb ) model is set forth in fig . 1(a),@xcite where those nearest - neighbor electron hoppings are shown . the bulk subbands have been known to correspond to the @xmath6d projection of a dp spiral surrounds the @xmath17 point of the @xmath6d hexagonal brillouin zone ( bz),@xcite as shown in fig . the projection is close to the annular valleys that are present in the bulk subbands in abc - stacked graphene.@xcite a @xmath3d rhombohedral - to-@xmath6d hexagonal zone - folding relation could be in principle figured out to obtain the @xmath6d bulk subbands . however , one can not do a straight folding along the stacking direction ( @xmath18 ) here as one does for ab - stacked graphite.@xcite . this fact can be understood since the vertical line through @xmath17 is no longer a symmetry line in the @xmath3d rhombohedral bz.@xcite we shall derive a @xmath6d effective hamiltonian in the bulk subspace , which is used together with the well realized effective hamiltonian in the surface subspace.@xcite the evolution and dimensional crossover can be clarified once the eigenenergies are solved out for arbitrary @xmath0 . such methodology using subspace effective hamiltonians has been applied to study ab - stacked @xmath0-layer graphene,@xcite for which the subband indices are directly related to the projections of the @xmath3d band structure at allowed @xmath18 , in consistency with the vertical zone - folding scheme.@xcite although a direct relation between the bulk subband index and @xmath18 is not accessible for abc - stacked @xmath0-layer graphene , it is rational to relate the indices to certain @xmath6d projections from the @xmath3d dirac cone structure at all allowed @xmath19 in the @xmath3d rhombohedral bz . the hexagonal bz of the stack of @xmath0 layers becomes extremely thinned as @xmath0 goes infinitely large , so that we have an approximately zero extent in the @xmath18 dimension . hence , the continuum of projections forming a bulk zone in the @xmath6d hexagonal bz is acquired . the model is described as follows . referring to fig . 1(a ) , the nearest interlayer hopping , @xmath20 , takes place between the vertically bonded bulk atoms in adjacent layers . those interlayer hoppings other than @xmath20 alter the surface subbands but they only perturb the bulk subbands.@xcite the minimal model including only @xmath20(@xmath21 ev ) and the nearest intralayer hopping @xmath22(@xmath23 ev ) suffices for our purpose . the hamiltonian @xmath24 , based on the tb bloch functions @xmath25 with respect to the @xmath26 sublattices , is represented as @xmath27 with @xmath28 and @xmath29 , where @xmath30 is the momentum measured from @xmath17 and @xmath31 are the valley index for @xmath32 and @xmath33 , respectively . the two dp spirals with respect to @xmath32 and @xmath33 have clockwise and counterclockwise spiraling senses with increasing @xmath18,@xcite as shown in fig . 1(c ) . without loss of generality , we take @xmath34 for @xmath32 below . it is noted that within the minimal model , the dp spiral is laid on a cylinder surface around the verticle line through @xmath17 and therefore its @xmath6d projection becomes a circle around @xmath17.@xcite for the scale of @xmath35 , we remark the radius of the projected dp spiral@xcite @xmath36 this delimits the boundary of the non - trivial topology for the bulk - surface correspondence.@xcite the hamiltonian @xmath24 in eq . ( 1 ) is partitioned as follows . the upper left @xmath4 block is denoted by @xmath37 with respect to the surface subspace spanned by @xmath38 , and the lower right @xmath39 block is @xmath40 for the bulk subspace spanned by the bulk sublattices . the coupling between @xmath37 and @xmath40 is given by the off - diagonal blocks @xmath41 and @xmath42 $ ] . it is easy to identify the surface subbands as being lower in energy than the bulk subbands . the secular equation reduces to @xmath43 at @xmath32 , where the coupling is absent since @xmath41 and @xmath44 are zero matrices with @xmath45 . the lowest eigenenergy @xmath46 is associated with two degenerate eigenstates in the surface subspace , while the eigenenergies @xmath47 are each associated with @xmath48 degenerate eigenstates in the bulk subspace . our goal is to construct an effective hamiltonian in the bulk subspace and solve the eigenvalues for arbitrary @xmath0 . that is , a block diagonalization for the full hamiltonian @xmath24 is required : @xmath49 where @xmath50 and @xmath51 are the effective hamiltonians in the surface and the bulk subspaces , respectively . considering the full hamiltonian [ see eq . ( 1 ) ] with @xmath52 in general , we expand its eigenvectors as @xmath53 in terms of the eigenvectors @xmath54 of the uncoupled hamiltonian ( @xmath45 ) . hence , the schrdinger equation is reformulated to be @xmath55 where @xmath56 and @xmath57 are , respectively , @xmath58 and @xmath59 partitions of the coefficient matrix @xmath60 . the block diagonalization in eq . ( 3 ) is then conceptually performed by a similarity transformation of the basis from the set of @xmath61 to the unsolved set of @xmath62 . from eq . ( 4 ) the effective hamiltonian @xmath50 in the surface subspace is given by @xmath63 and the effective hamiltonian @xmath51 in the bulk subspace is @xmath64 the derivation of eq.s ( 5 ) and ( 6 ) for the effective hamiltonians is consistent with the use of the green s function @xmath65 such that we have @xmath66 , for which the partitions @xmath67 and @xmath68 of the green s function @xmath69 can be explicitly obtained.@xcite the effective hamiltonians shown in eq.s ( 5 ) and ( 6 ) depend on the energy @xmath70 due to the coupling between the surface and the bulk subbands . this might be deemed the price for the reduction in dimension . an effective hamiltonian can even be non - hermitian in general , but this is not the case here . at first sight , a perturbation calculation is expected as usual . unfortunately , divergence of the perturbation series would be encountered if the coupling is strong compared with a diminishing bulk energy gap as envisaged here for infinitely large @xmath0 . we shall take a non - perturbative approach below , which is accessible owing to the specific form of eq.s ( 5 ) and ( 6 ) . it is noted that the similarity transformation for the block diagonalization is not unique in general , while different effective hamiltonians obtained through different transformations yield the same eigenvalues . the surface subbands have been well realized by the chiral effective hamiltonian @xmath71 for abc - stacked trilayer graphene.@xcite this can be obtained in the present framework by removing the @xmath70-dependence due to the powers of @xmath72 in the coupling term of @xmath50 in eq . the chiral effective hamiltonian for arbitrary @xmath0 can be deduced to be@xcite @xmath73 it should be noted that the use of @xmath74 fails outside the interval between the @xmath32 point and the projection of the dp spiral , @xmath75 , because it is derived by the power expansion of @xmath76 . inside the interval , the surface subbands are more and more flat and approach the zero energy.@xcite the effective hamiltonian @xmath77 in the bulk space is deduced from eq . ( 6 ) without any perturbation or power expansion . the effect of the surface on the bulk subbands are thus completely retained . in so doing we can reach the region around @xmath78 , onto which the dp spiral is projected , even though the energy gap is known to be nearly closed up for infinitely large @xmath0 . because @xmath37 is zero in eq . ( 1 ) , the coupling term @xmath79 in eq . ( 6 ) has non - vanishing elements at the two diagonal corners only , given by @xmath80 . therefore , the coupling is almost absent from the vicinity of @xmath32 with @xmath81 and @xmath82 , where the effective hamiltonian is identical to the block with respect to the bulk subspace : @xmath83 in general , the non - perturbative effective hamiltonian @xmath77 is given by @xmath84 in comparison with eq . ( 7 ) for @xmath50 , where the effect of the bulk subbands on the surface subbands is significant , eq . ( 9 ) shows that the coupling effect on the bulk subbands is relatively weak . it is observed that @xmath77 has a nested structure as well as @xmath40 does , a characteristic of abc - stacking configuration.@xcite for the bulk subbands in abc - stacked @xmath0-layer graphene , the eigenenergies of the effective hamiltonian @xmath77 described in eq . ( 9 ) are solved from the secular equation @xmath85 . owing to the nested structure of @xmath77 , a second - order recursion is involved . the secular equation is separated as @xmath86 where @xmath87 $ ] and @xmath88 are given by @xmath89 @xmath90 note that @xmath88 comes from the coupling to the surface subbands . the nested structure in eq.s ( 11 ) and ( 12 ) brings about @xmath91 which reduces to a second - order recursive equation @xmath92 where @xmath93 from eq . ( 11 ) , the initial conditions are given by @xmath94 the recursion for @xmath88 has the same form as for @xmath95 in eq . ( 14 ) , with more tedious initial conditions . we rewrite eq . ( 14 ) as @xmath96 in order to resolve @xmath95 with the initial conditions given in eq . ( 16 ) by two steps . at first , we get @xmath97 . then we achieve the resolution and express @xmath98 for arbitrary @xmath0 . a concise form of resolved @xmath95 is written as @xmath99,\;\;\;\;\ ; n\ge3.\ ] ] with eq . ( 17 ) , the resolved expression of @xmath88 can be obtained from eq.s ( 13 ) . in proceeding to solve the eigenvalue problem of @xmath100 from eq . ( 10 ) , now the resolved terms @xmath95 and @xmath88 obtained from eq.s ( 13 ) and ( 17 ) are manipulated further . the variables @xmath101 and @xmath102 defined in eq . ( 15 ) form a complex conjugate pair if @xmath103[\varepsilon-(\beta_{1}-vp)]% [ \varepsilon+(\beta_{1}+vp)][\varepsilon+(\beta_{1}-vp)]<0.\ ] ] we take eq . ( 18 ) as _ a priori _ true , which leads to self - consistent results as shown in the following . thus , all the eigenvalues are definitely enveloped by the branches @xmath104 and @xmath105 . by changing variables as @xmath106 and @xmath107 , eq . ( 15 ) is equivalent to @xmath108 which provides an easy - to - handle connection between the eigenvalue @xmath70 and variable @xmath109 . accordingly , eq . ( 17 ) is transformed to be @xmath110 with the aid of the trigonometric identity @xmath111 . using eq.s ( 13 ) with eq . ( 17 ) again , we also have @xmath112 consequently , the secular eq . ( 10 ) is transformed to be @xmath113 where eq . ( 19 ) is used . there are @xmath48 roots to eq . ( 22 ) in correspondence to the @xmath114 eigenvalues of @xmath100 . once those roots @xmath115 , @xmath116 , as functions of @xmath117 are determined , the whole eigenenergy spectrum of @xmath100 can be acquired according to eq . that is , @xmath118^{1/2},\;\;\;j=1,2,\ldots , n-1,\ ] ] where @xmath119 is the subband index and @xmath120 denote , respectively , the conduction and the valence bulk subbands . the quantum number @xmath119 is related to the allowed @xmath6d projections from the @xmath3d dirac cone structure . to find out @xmath115 we firstly survey the region where the term @xmath121 in eq . ( 22 ) can be approximately dropped . then we go away from @xmath122 at @xmath32 and reach @xmath123 around the @xmath6d projection @xmath124 of the dp spiral [ see eq . ( 2 ) ] . with @xmath122 , the roots @xmath115 are simply specified by @xmath125 , @xmath126 , given by @xmath127 with @xmath128 , eq . ( 22 ) becomes @xmath129 , @xmath126 . for @xmath130 , the roots @xmath115 are given by @xmath131 in eq.s ( 24 ) and ( 25 ) , the numerals @xmath119 are so specified as to exclude the case of @xmath132 and to omit double counting of the value of @xmath133 . the two sets of @xmath134 at the two ends of @xmath121 are somewhat different , but with increasing @xmath0 they approach and ultimately coincide with each other . between the two ends , we directly calculate @xmath115 from eq . ( 22 ) for few - layer graphene . for large @xmath0 , the use of either eq . ( 24 ) or eq . ( 25 ) is convenient . it is conventionally known that the local dos in a space is related to a green s function . we firstly establish this relationship in the bulk subspace . if @xmath135 is made , the related green s function reads @xmath136.@xcite referring to eq . ( 23 ) , the subband index @xmath137 and band index @xmath138 are lumped as @xmath139 in the following . in the eigenspace of @xmath77 , where the eigenvector and eigenvalue are respectively denoted by @xmath140 and @xmath141 , the green s function represented as @xmath142 reduces to @xmath143 for which the identity @xmath144 for an operator @xmath145 is used . based on the @xmath114 tb bloch functions @xmath146 , the imaginary part of the diagonal element @xmath147 of @xmath148 in eq . ( 26 ) is given by @xmath149 with @xmath150 being the component of @xmath140 at the @xmath151th bulk sublattice , where @xmath152 , @xmath153 , is defined for the delta function . the local dos has a manifestation in eq . ( 27 ) since the eigenvector @xmath140 contributes a probability density @xmath154 at the @xmath151th bulk sublattice . in the infinitely extended @xmath10 plane , the number of states is obtained by counting the allowed wave vectors @xmath155 . hence , the local dos is given by @xmath156 where the integration turns out to run along the circular isoenergetic path with respect to each subband @xmath157 given in eq . ( 23 ) . since all the bulk atoms are indistinguishable in the infinite limit of @xmath0 , we are going to calculate the bulk dos for arbitrary @xmath0 and observe the evolution . the bulk dos is obtained from the local dos in eq . ( 28 ) by summing @xmath158 over all the @xmath114 bulk sublattices and is thus given by @xmath159 . this leads to @xmath160 where @xmath161 for normalized @xmath140 . note that the superscript @xmath0 of @xmath162 is implied by the up bound of @xmath119 in @xmath163 . in performing the calculation with eq . ( 29 ) , we use the lorentzian @xmath164^{-1}$ ] with a tiny width @xmath165 to approximate the delta function @xmath166 . 0.6 truecm 0.1 truecm the eigenenergies @xmath167 of @xmath77 for the bulk subbands are calculated from eq . ( 23 ) for various numbers ( @xmath0 ) of abc - stacked graphene layers . the results are presented and discussed below , together with the energies of the surface subbands @xmath168 which is obtained from the chiral effective hamiltonian @xmath74 given in eq . ( 7 ) . in fig . 2 , the evolution of the band structure of abc - stacked graphene is clear in an overview from a few layers ( @xmath169 ) to a lot of layers ( @xmath170 ) . certainly , the surface subbands are rapidly flatten inside the interval @xmath171 . it is noted that the use of @xmath74 fails dramatically outside this interval , whereas actually the surface subbands should be suppressed due to band repulsion . this failure can be circumvented using the full hamiltonian consisting of coupled surface and bulk states , whether around the @xmath32 point@xcite or over the whole bz.@xcite by contrast , our results from the non - perturbative @xmath100 for the bulk subbands agree with those obtained using the full hamiltonian over a wide region.@xcite the intricate bulk subbands can be unraveled by analyzing the expression of @xmath167 given in eq . obviously , the eigenenergy spectrum has isotropic and electron - hole symmetries , as a consequence of the minimal model . all the @xmath48 conduction bulk subbands @xmath172 , as well as all the @xmath48 valence bulk subbands @xmath173 , are degenerate at the @xmath32 point . to be explicit , those @xmath134 given in eq . ( 24 ) , which is derived about @xmath32 , are paired according to @xmath174 ; in particular , @xmath175 stands solely if @xmath0 is even . therefore , near @xmath32 the conduction and the valence bulk subbands comprise pairs of @xmath176 and pairs of @xmath177 , respectively , with @xmath178 , where @xmath179 is the gaussian symbol . besides , for even @xmath0 they also comprise quadratic @xmath180 and @xmath181 , respectively . in each of the @xmath182 pairs there is one conduction ( valence ) bulk subband extending downward ( upward ) from @xmath32 towards the zero energy . it is just the subband @xmath183 [ @xmath184 . remarkably , it turns away around @xmath124 and forms a valley as shown in fig . such valleys are annular since all the eigenenergies are isotropic . according to eq . ( 23 ) , the turning is due to the nonlinear terms in powers of @xmath121 in the expansion of @xmath167 . there is an energy gap between the valley edges of the lowest downward conduction and the highest upward valence bulk subbands @xmath185 , which are the extensions from the aforementioned @xmath186 and @xmath187 , respectively . by differentiating them in eq . ( 23 ) with respect to @xmath188 near @xmath189 , the edge momentum in abc - stacked @xmath0-layer graphene is approximately given by @xmath190 the edge energies of @xmath185 are then obtained and so is the energy gap , the latter being @xmath191 equation ( 31 ) indicates that @xmath192 increases with increasing @xmath0 . in the course of the evolution , @xmath193 diminishes as described by eq . the results displayed in fig . 2 show that @xmath192 approaches @xmath124 rapidly but as predicted,@xcite @xmath193 approaches zero relatively slowly . a comparison in this aspect can be made with the case for ab - stacked graphene , whose band structure exhibits similar @xmath3d features to ab - stacked graphite with just several layers ( @xmath194).@xcite taking the advantage of the explicit expression of @xmath167 given in eq . ( 23 ) , our calculation can be achieved for arbitrary @xmath0 . in fig . 2 , the energy gap at @xmath170 has become narrow , though still opening . for further realization of the evolution , the energy dispersion should also be investigated . the bulk subbands @xmath167 tend to disperse linearly in the increase of @xmath0 as observed from fig . this tendency can be understood by analyzing the functionality of @xmath195 in eq . ( 23 ) . the bulk doss @xmath162 ranging from a few to dozens of layers and up to @xmath170 are also calculated and are shown in fig . 3 . there are as many peaks displayed in @xmath162 as there are the bulk subbands . each bulk subband yields a peak at its own valley edge . the most prominent two peaks are due to @xmath185 , separated by the aforementioned diminishing gap @xmath193 . besides , a dip is present at energy @xmath20 ( @xmath196 ) in the conduction ( valence ) bulk subband spectrum . this reflects the cusp at the @xmath32 point , where all the subbands are degenerate . all these features are gradually smeared in the evolution . the dimensional crossover from the gapped bulk subbands in abc - stacked graphene to the @xmath3d dirac cone structure in rg is elucidated with respect to eq . ( 23 ) . as @xmath197 , there exists an infinitely large subset of the @xmath182 downward ( upward ) conduction ( valence ) bulk subbands whose elements all approach @xmath198 where the values of respective @xmath199 all approach @xmath200 according to eq.s ( 24 ) or ( 25 ) . the two limits given in eq . ( 33 ) are two branches of the envelope described in eq . ( 18 ) , a necessary condition for the existence of the eigenvalues of @xmath77 . they constitute the boundary of a zone containing the whole bulk subband spectrum for infinite @xmath0 , which forms the continuum of the @xmath6d projections from the @xmath3d dirac cone structure . they intersect at the infinite limit of the edge momentum @xmath201 determined by eq . ( 31 ) , at which the energy gap closes up with @xmath202 [ see eq . consequently , the bulk zone boundary forms a linear annular cone apexed along a circle at a distance @xmath203 from the @xmath32 point . the circle is known to be the @xmath6d projection of the dp spiral as described in eq . ( 2).@xcite using eq . ( 23 ) , the bulk subbands can be calculated for a huge number ( @xmath0 ) of abc - stacked graphene layers as desired . the results for @xmath204 are plotted in fig . 4(a ) , in which the completely flat surface subbands are not plotted for brevity . as shown , the infinite limit has been practically attained . a gapless bulk zone full of bulk subbands are displayed . the zone boundary is , indeed , a linear annular cone with the apex located along a circle of radius @xmath124 . the characteristics of the projected dirac cones are also exhibited in the infinite limit of the bulk dos @xmath205 . the calculation results from eq . ( 29 ) for @xmath204 are plotted in fig . 4(b ) , shown to be consistent with the subband energies plotted in fig . those features that are present for finite @xmath0 , as shown in fig . 3 , are completely smeared in the infinite limit . in agreement with previous research,@xcite the results of @xmath205 acquired here is the same as the dos in monolayer graphene , reflecting the characteristics of the dirac cone . in particular , the vanishing bulk dos at the zero energy , viz . , @xmath206 , characterizes rg as a semimetal with half - filled dp spirals in the bulk.@xcite 0.6 truecm 0.1 truecm the band structure of abc - stacked @xmath0-layer graphene comprises flat surface and gapped bulk subbands as a consequence of the unique stacking configuration , where the non - trivial topology leads to a bulk - surface correspondence.@xcite so far , the intricate bulk subbands have not yet been unravelled , although the surface subbands are well understood . in this paper , for the first time ever the bulk subbands can be explicitly obtained for arbitrary @xmath0 . hence , their evolution with increasing @xmath0 can be demonstrated together with the surface subbands . the bulk subbands are certain @xmath6d projections from the @xmath3d dirac cone structure at allowed wave vectors in the bulk of rg . towards the infinite limit of the evolution , a bulk zone of the continuum of the @xmath6d projections shows up and a dimensional crossover occurs from the gapped bulk subbands to the gapless @xmath3d dirac cone structure . correspondingly , the surface subbands become completely flat within a region confined by the @xmath6d projection of the dp spiral . the physics with respect to abc - stacking configuration for layered systems is rich and deep . as compared with the famous abc - stacked layered @xmath3d topological insulators such as bi@xmath7se@xmath8 and bi@xmath7te@xmath8 , the @xmath3d system rg here we focus on behaves as a topological semimetal . it possesses metallic flat surface subbands while being semimetallic in the bulk with a 3d dirac cone structure . a generalization of the present study to other abc - stacked systems is expected , regarding the evolution and dimensional crossover of the bulk subbands from @xmath6d to @xmath3d realm . we point out the convenience of using an effective hamiltonian closed in the bulk subspace . the resolution of the recursion involved in the eigenvalue problem can be achieved utilizing the nested structure of the effective hamiltonian . an explicit expression of the bulk subbands for arbitrary @xmath0 can be thus obtained . moreover , it is appropriate to construct an effective hamiltonian that is non - perturbation in case the coupling between the surface and the bulk subbands is strong compared with a vanishingly small bulk energy gap for infinitely large @xmath0 . * ( color online ) ( a ) stacking configuration of abc - stacked @xmath0-layer graphene . @xmath15 ( black filled circles ) and @xmath16 ( black unfilled circles ) : surface sublattices ; the rest ( red circles ) : bulk sublattices . nearest - neighbor tb hoppings ( @xmath207 ) are shown ; only @xmath22 and @xmath20 are used in the minimal model . ( b ) solid view of the bulk lattice in rg , composed of infinitely many abc - stacked graphene layers , with the biparticle rhombohedral primitive unit cell ( red ) . an alternatively set hexagonal unit cell of triple volume ( blue ) is also plotted for reference . ( c ) illustration of several @xmath6d projections ( red dots ) of the dp spiral around @xmath32 . the inset displays clockwise and counterclockwise dp spiraling senses with increasing @xmath18 around @xmath32 and @xmath33 , respectively . * ( color online ) band structure of abc - stacked @xmath0-layer graphene . flat pair about the zero energy ( grey ) : surface subbands ; the rest ( red ) : bulk subbands . the direction of the momentum is taken arbitrarily . * ( color online ) bulk doss in abc - stacked @xmath0-layer graphene , in the unit of number of states per @xmath20 per atom . * ( color online ) ( a ) energies of the bulk subbands in abc - stacked @xmath208-layer graphene , pratically showing the dimensional crossover with the bulk zone full of bulk subbands . ( b ) bulk dos associated with ( a ) , in the unit of number of states per @xmath20 per atom .	the band structure of abc - stacked @xmath0-layer graphene comprises topologically corresponding flat surface and gapped bulk subbands , as a consequence of the unique stacking configuration . in this paper , the bulk subbands are for the first times ever obtained for arbitrary @xmath0 . a non - perturbative effective hamiltonian closed in the bulk subspace is derived and used . the gapped bulk subbands are shown to evolve towards the zero energy with increasing @xmath0 and in the infinite limit , they touch linearly along a circle . this outcome is a manifestation of the dimensional crossover to a three - dimensional dirac cone structure known to exist in the bulk of rhombohedral graphite . the dirac points , forming continuous nodal lines in a spiraling fashion , are projected onto the circle , within which the surface subbands are confined and flatten .
the origins of power - law distributions as well as their conceptual implications have been an active topic of research in recent years . power laws are intrinsically related to the emergence of macroscopic features which are scale invariant within some bounds and distinct from the microscopic elementary degrees of freedom . often , these features are insensitive to the details of the microscopic structures . well known examples of power law distributions include the energy distribution between scales in turbulence @xcite , the distribution of earthquake magnitudes@xcite , the diameter distribution of craters and asteroids@xcite , the distribution of city populations @xcite , the distributions of income and of wealth @xcite , the size - distribution of business firms @xcite and the distribution of the frequency of appearance of words in texts@xcite . the fact that multiplicative dynamics tends to generate power - law distributions was intuitively invoked long ago @xcite but the limitations in computer simulation power kept the models under the constraints imposed by the applicability of analytical treatment . more recently , a broader class of models has been studied combining computer simulations with theoretical analysis within the microscopic representation paradigm proposed in ref . in particular , it was shown @xcite that power laws appear in a variety of dynamical processes and are maintained even under highly non - stationary conditions . in this paper we consider a generic model of stochastic dynamics with many degrees of freedom @xmath19 , @xmath1 . the time evolution of the @xmath0 s is described by an asynchronous update mechanism in which at each time step one variable is chosen randomly and is multiplied by a factor @xmath9 taken from a predefined distribution . in addition , there is a global coupling constraint which does not allow the @xmath0 s to fall below the lower cutoff given by @xmath3 , where @xmath4 is the momentary average of the @xmath0 s and @xmath5 is a constant . the dynamic variables @xmath0 are found to exhibit a power - law distribution of the form @xmath6 . the exponent @xmath18 is found to be insensitive to the distribution @xmath8 of the random factor @xmath9 . however , @xmath18 is non - universal , and increases monotonically as a function of @xmath10 . in the limit @xmath20 ( where the @xmath0 s become decoupled ) @xmath21 for any _ finite _ @xmath14 . however , in the `` thermodynamic '' limit @xmath22 , @xmath23 for any _ positive _ @xmath10 . thus the two limits do not commute . this is important for applications since typically in empirical systems @xmath23 @xcite unlike the case of the free multiplicative random walk which predicts a log - normal distribution @xcite corresponding to @xmath21 @xcite . the time evolution of @xmath24 exhibits intermittent fluctuations parametrized by a truncated lvy - stable distribution with the same index @xmath18 . this intricate relation between the distribution of the @xmath0 s at a given time and the temporal fluctuations of their average is examined and its relevance to empirical systems is discussed . our model indicates that in certain cases the scaling exponent may be insensitive to the distribution of the multiplicative ( random ) factor @xmath9 and depends only on the `` lower bound '' features which control the smallest values of the elementary variables . the relation between the limiting conditions and the power law exponent is to be applied in each particular case and it constitutes a strong instrument in identifying and validating the relevant degrees of freedom responsible for the emergence of scaling . the present paper proposes to consolidate by numerical simulations the control one has on a specific model and help in this way its further application to additional systems . the paper is organized as follows . in sec . ii we present the model . simulations and results are reported in sec . iii , followed by a discussion in sec . iv and a summary in sec . v. the model@xcite describes the evolution in discrete time of @xmath14 dynamic variables @xmath25 , @xmath1 . at each time step @xmath26 , an integer @xmath27 is chosen randomly in the range @xmath28 , which is the index of the dynamic variable @xmath0 to be updated at that time step . a random multiplicative factor @xmath29 is then drawn from a given distribution @xmath8 , which is independent of @xmath27 and @xmath26 and satisfies @xmath30 . this can be , for example , a uniform distribution in the range @xmath31 , where @xmath32 and @xmath33 are predefined limits . the system is then updated according to the following stochastic time evolution equation @xmath34 this is an asynchronous update mechanism . the average value of the system components at time t is given by @xmath35 the term on the right hand side of eq . ( [ eq : mult ] ) describes the effect of auto - catalysis at the individual level . in addition to the update rule of eq . ( [ eq : mult ] ) , the value of the updated variable @xmath36 is constrained to be larger or equal to some lower bound which is proportional to the momentary average value of the @xmath0 s according to @xmath37 where @xmath38 is a constant factor . this constraint is imposed immediately after step ( [ eq : mult ] ) by setting @xmath39 where @xmath24 , evaluated just before the application of eq . ( [ eq : mult ] ) , is used . this constraint describes the effect of auto - catalysis at the community level . our model is characterized by a fixed ( conserved ) number of dynamic variables @xmath14 , while the sum of their values is not conserved . the conservation of the number of dynamic variables , which is enforced through the lower cutoff constraint is essential since otherwise the system dwindles over time . the non - conservation of the sum of the values of the dynamic variables is important as well . it allows to perform the multiplicative updating on a single variable at a time with no explicit binary interactions since a gain in @xmath0 does not require a corresponding immediate loss by other @xmath40 s . in fact , the interactions between the dynamic variables are implied only in the step of eq . ( [ eq : subsidize ] ) in which the lower cutoff is imposed . the dynamic rule ( [ eq : mult ] ) can be described by a master equation for the probability distribution @xmath41 of the form @xmath42 , \label{eq : master}\ ] ] where the @xmath43 factor takes into account the fact that only one of the @xmath0 s is updated in each time step . this description applies for the bulk of the distribution of the @xmath0 s but not in the vicinity of the lower cutoff where the step of eq . ( [ eq : subsidize ] ) which is not taken into account by eq . ( [ eq : master ] ) may be dominant . for the following analysis it is convenient to normalize the @xmath40 s according to @xmath44 as a result , the new average @xmath24 is normalized to @xmath45 while @xmath46 . performing this normalization step after each iteration removes the non stationary part of the distribution and amounts statistically to an overall multiplicative factor . this ( time dependent ) factor which represents a global inflation rate can be recorded at each step . it is convenient to represent the dynamics ( [ eq : master ] ) on the logarithmic scale . in terms of the new variables @xmath47 eq . ( [ eq : mult ] ) defines a random walk with steps of random size @xmath48 : @xmath49 the corresponding probability distribution @xmath50 becomes @xmath51 in terms of @xmath52 and @xmath53 , the master equation ( [ eq : master ] ) becomes : @xmath54 . \label{eq : addmast}\end{aligned}\ ] ] the asymptotic stationary solution , is found to be @xcite @xmath55 in terms of the original variable @xmath0 , we get according to eq . ( [ eq : transf ] ) a power law distribution : @xmath56 the value of the exponent @xmath18 is determined by the normalization condition [ eq . ( [ eq : mean ] ) ] divided by the probability normalization condition @xmath57 ( in order to eliminate the constant factor @xmath58 ) , which yields : @xmath59 . \label{eq : calc}\ ] ] the exponent @xmath18 is given implicitly as a function of @xmath10 and @xmath14 by eq . ( [ eq : calc ] ) . we identify two regimes within @xmath38 in which eq . ( [ eq : calc ] ) can be simplified and @xmath18 can be obtained explicitly . for a given @xmath14 and values of @xmath10 in the range @xmath60 one obtains @xmath61 as well as @xmath62 . consequently , in this range , one can neglect the @xmath63 terms in eq . ( [ eq : calc ] ) to obtain to a good approximation @xmath64 . \label{eq : calc1}\ ] ] which gives the explicit , @xmath14-independent solution @xmath65 this relation is exact in the `` thermodynamic '' limit @xmath22 . the relation ( [ eq : temp ] ) has two remarkable properties : ( a ) it does not depend on the distribution @xmath66 ; ( b ) it gives rise to @xmath18 values in the experimentally realistic range @xmath23 . for finite @xmath14 and values of @xmath10 lower than @xmath67 the approximation eq . ( [ eq : temp ] ) breaks down and values @xmath68 become possible . however , for any finite n , another approximation holds in the range @xmath69 . in this range @xmath70 and therefore one can neglect @xmath71 in the numerator of eq . ( [ eq : calc ] ) and @xmath72 in the denominator to obtain : @xmath73 . \label{eq : calc2}\ ] ] by taking the logarithm on both sides and neglecting terms of order @xmath74 we obtain @xmath75 note that even for systems in which the lower bound ( which is due to some microscopic discretization ) given by @xmath10 , is orders of magnitude smaller than @xmath43 , the resulting @xmath18 may differ significantly from the free multiplicative random walk result @xmath21 . since @xmath10 enters in the formula ( [ eq : c->0 ] ) for @xmath18 through its logarithm , the system gives away information on its microscopic scale cut - off @xmath10 through the exponent @xmath18 of its macroscopic power law behavior . one should emphasize that in the region where @xmath76 the average @xmath4 of the distribution @xmath41 in eq . ( [ eq : mean ] ) is not well defined and in fact one expects in the actual runs very wide macroscopic fluctuations of this mean . these fluctuations are however never infinite because according to the formulae above , as one increases the size of the system @xmath14 , the region along the @xmath10 axis where @xmath77 shrinks to 0 . for @xmath78 it is only the standard deviation of the distribution @xmath79 which is formally divergent . this gives rise in the actual computer simulations to wide fluctuations of the individual values of @xmath0 . however , this divergence is kept in check too by the fact that no @xmath80 can possibly exceed @xmath81 , namely @xmath82 . this amounts to a _ truncation from above _ of the power law eq . ( [ eq : power ] ) . numerical simulations of the stochastic multiplicative process described by eqs . ( [ eq : mult ] ) and ( [ eq : subsidize ] ) , confirm the validity of eq . ( [ eq : power ] ) for a wide range of lower bounds @xmath10 . it appears that the exponent @xmath18 is largely independent of the shape of the probability distribution @xmath8 . [ fig : pareto ] shows the distribution of @xmath0 , @xmath1 , obtained for @xmath83 , @xmath84 , and @xmath9 uniformly distributed in the range @xmath85 . a power law distribution is found for a range of three decades between @xmath86 and @xmath87 . the slope of the best linear fit within this range is given by @xmath88 , in agreement with eqs . ( [ eq : calc ] ) and ( [ eq : temp ] ) . on the horizontal axis of this graph the sum of all @xmath0 s is normalized to 1 and therefore @xmath89 . the exponent @xmath18 as a function of the lower cutoff @xmath10 is shown in fig . [ fig : alpha ] . numerical results are presented for @xmath90 ( empty dots ) , @xmath91 ( full dots ) and @xmath92 ( squares ) . the prediction of eq . ( [ eq : calc ] ) is shown for @xmath83 ( solid line ) , which is in good agreement with the numerical results for all values of @xmath10 . the approximate expression eq . ( [ eq : temp ] ) is also shown ( dashed line ) . it is observed that for @xmath93 this approximation gradually starts to hold as @xmath10 is increased beyond @xmath94 , in agreement with the theoretical analysis . in general , for a given @xmath14 , @xmath18 is monotonically increasing as a function of @xmath10 , starting from @xmath21 ( which corresponds to @xmath95 distribution ) at @xmath96 , where the @xmath0 s are uncoupled . it is also observed that as @xmath14 is raised , the value of @xmath18 which corresponds to a given @xmath10 increases monotonically . as a result , the range of validity @xmath97 of the approximation eq . ( [ eq : temp ] ) is extended and the knee adjacent to @xmath96 sharpens and becomes a discontinuity for @xmath11 . the range @xmath98 in which the approximation of eq . ( [ eq : c->0 ] ) is valid , shrinks correspondingly . let us turn now to the dynamics of the system as a whole . the dynamics of the system involves , according to eq . ( [ eq : mult ] ) , a generalized random walk with step sizes distributed according to eq . ( [ eq : power ] ) . therefore , the stochastic fluctuations of @xmath99 after @xmath100 time steps : @xmath101 are governed @xcite by a truncated lvy distribution @xmath102 . in fig . [ fig : levy ] we show the distribution of the stochastic fluctuations @xmath103 for @xmath104 , which is given by a ( truncated ) lvy distribution @xmath102 . according to ref . @xcite , the peak of the ( truncated ) lvy - stable distribution scales with @xmath100 as @xmath105 where @xmath18 is the index of the lvy distribution . in fig . [ fig : levyzero ] we show the height of the peak p@xmath106 of fig . [ fig : levy ] as a function of @xmath100 . it is found that the slope of the fit in fig . [ fig : levyzero ] is @xmath107 , which following the scaling relation ( [ eq : fluct ] ) means that the index of the lvy distribution in fig . [ fig : levy ] is @xmath108 . these results were obtained for the same parameters which gave rise to the power law distribution with @xmath88 in fig . [ fig : pareto ] . thus , the prediction that the fluctuations of @xmath4 in fig . [ fig : levy ] follow a ( truncated ) lvy - stable distribution with an index @xmath18 which equals the exponent @xmath18 of the power - law distribution in fig . [ fig : pareto ] , is confirmed . the model considered in this paper may be relevant to a variety of empirical systems in the physical , biological and social sciences which can be described by a set of interacting dynamic variables which follow a stochastic multiplicative dynamics . such dynamical processes may play a role in the formation of the mass distribution in the universe where clusters of galaxies accumulate and eventually form super - clusters . in a different context , the growth of cities is basically a multiplicative process governed by the reproduction rate of the local population in addition to mobility between cities . enhanced diffusion processes , which can be described by the lvy - stable distribution have been observed in a variety of nonlinear dynamical systems @xcite . unlike the stochastic model studied here , these systems are governed by deterministic rules . they exhibit intermittent chaotic motion which gives rise to enhanced diffusion . in population dynamics , the number of individuals in each specie varies stochastically from one season to the next with a multiplicative factor which depends on the local conditions . the lower bound may represent the minimal number of individuals required for the species to survive in the given environment . in this case the number of species may not be strictly a constant , but species that are wiped out may be replaced by others which invade their area . in this context it was found that the number of species of a given size often follows a decreasing power - law distribution as a function of their size ( see e.g. ref . @xcite ) . in the economic context of a stock - market system the dynamic variables @xmath0 , @xmath1 may represent the wealth of individual investors . in this case the dynamics represents the increase ( or decrease ) by a random factor @xmath29 of the wealth @xmath0 of the investor @xmath27 between times @xmath26 and @xmath109 . the lower bound may represent a minimal wealth required in order to participate in stock market trading . in a more general economic model , this lower bound may be related to a basket of basic publicly funded services which every individual receives . in another possible interpretation , the @xmath0 s represent the capitalization ( total market value ) of the firm @xmath27 , which may increase ( or decrease ) by a factor @xmath29 at each time step . in this case the lower bound may represent the minimal requirements for a company stock to be publicly traded . studies of the distribution of wealth in the general population revealed a power - law behavior ( see e.g. ref . more recently it was shown @xcite that the distribution of individual wealth of the 400 richest people in the united states ( forbes 400 ) corresponds to a power law with @xmath110 [ more precisely @xmath111 where @xmath112 is the wealth of the @xmath113-th richest person on the list ] . recent analysis of stock market returns , measured over many years found a truncated lvy distribution @xmath102 with the index @xmath114 for an extended ( but finite ) range of returns @xmath115 @xcite . these results indicate that the property observed in our model , namely that the same value of the index @xmath18 appears both in the power law distribution and in the lvy - stable distribution of the fluctuations may be of relevance in the economic context . to further explore this possibility it would be interesting to examine whether the distribution of total market values of companies in the stock market exhibits a power law behavior of the form ( [ eq : power ] ) with @xmath114 . we have studied a generic model of stochastic auto - catalytic dynamics of many degrees of freedom using computer simulations . the model consists of dynamic variables @xmath0 , @xmath1 which are updated randomly one at a time through an autocatalytic process at the individual level . in addition , the variables are coupled through a lower bound constraint which enhances the variables which fall below a fraction of the global average . the model may describe a large variety of systems such as stock markets and city populations . the distribution @xmath116 of the system components @xmath0 turns out to fulfill a power law distribution of the form @xmath117 . in the limit @xmath118 , @xmath12 one obtains the case often encountered in nature : @xmath119 . the average @xmath120 exhibits intermittent fluctuations following a lvy - stable distribution with the same index @xmath18 . this relation between the distribution of system components and the temporal fluctuations of their average may be relevant to a variety of empirical systems . for example , it may provide a connection between the distribution of wealth / capitalization in a stock market and the distribution of the index fluctuations .	a generic model of stochastic autocatalytic dynamics with many degrees of freedom @xmath0 , @xmath1 is studied using computer simulations . the time evolution of the @xmath0 s combines a random multiplicative dynamics @xmath2 at the individual level with a global coupling through a constraint which does not allow the @xmath0 s to fall below a lower cutoff given by @xmath3 , where @xmath4 is their momentary average and @xmath5 is a constant . the dynamic variables @xmath0 are found to exhibit a power - law distribution of the form @xmath6 . the exponent @xmath7 is quite insensitive to the distribution @xmath8 of the random factor @xmath9 , but it is non - universal , and increases monotonically as a function of @xmath10 . the `` thermodynamic '' limit @xmath11 and the limit of decoupled free multiplicative random walks @xmath12 do not commute : @xmath13 for any finite @xmath14 while @xmath15 ( which is the common range in empirical systems ) for any positive @xmath10 . the time evolution of @xmath16 exhibits intermittent fluctuations parametrized by a ( truncated ) lvy - stable distribution @xmath17 with the same index @xmath18 . this non - trivial relation between the distribution of the @xmath0 s at a given time and the temporal fluctuations of their average is examined and its relevance to empirical systems is discussed .
recently , there has been a great progress in observational studies of structures of the high - redshift universe . @xcite found a proto - cluster at @xmath7 with a large population of lyman break galaxies ( lbgs ) . a significant excess of ly@xmath2 emitters ( laes ) was found by @xcite around a radio galaxy at @xmath8 , and a subsequent study concluded that this excess is indeed a proto - cluster with a size of @xmath9 mpc @xcite . more recently , @xcite reported clustering of 6 laes near a radio galaxy at @xmath10 . for structures beyond cluster scales , wide - field narrow - band surveys revealed very inhomogeneous distribution of laes @xcite at @xmath11 . @xcite report an elongated overdense region of laes at @xmath12 on the sky of 20 mpc in width and 50 mpc in length . however , on the same sky , @xcite found no large - scale structure at @xmath13 which is closer to us by 39 mpc , indicating a large cosmic variance over their surveyed volumes . although a few segments of high-@xmath14 structures have been found to date , the whole picture of high-@xmath14 universe is veiled . moreover , we do not know when and how the structures of galaxies seen at present formed from the initial matter density fluctuations . we started a systematic survey to map the high-@xmath14 universe with laes and lbgs at @xmath15 in the 1 deg@xmath1 sky of the subaru / xmm - newton deep field ( sxdf : r.a.@xmath16 , decl.@xmath17[j2000 ] ; @xcite ) . we obtained cosmic maps on scales of much larger than present - day lsss to cover a network of lsss at several redshifts between @xmath18 and @xmath19 . in this letter , we report the initial results of our survey about structures made of laes at @xmath0 . throughout this paper , we adopt @xmath20 km s@xmath21 mpc@xmath21 and @xmath22=[0.3,0.7,1.0,0.9]$ ] . we carried out extensive deep narrow - band imaging on 2003 september 27 - 30 and october 22 with subaru / suprime - cam @xcite . we used the narrow - band filter @xmath23 whose central wavelength and fwhm are @xmath24 and 120 , respectively @xcite . this filter enables us to identify laes , galaxies with a strong ly@xmath2 emission line , at @xmath25 with a small fraction of foreground contaminants ( 20 - 30% ; e.g. @xcite ) . our 20-hour exposure imaging covered an area of 1.04 deg@xmath1 with 5 suprime - cam pointings . the sky was clear and the observational conditions were stable , with typical seeing sizes of @xmath26 . the data were reduced with a software package sdfred @xcite . the sky noise of the reduced image is @xmath27 ( ab ) at the @xmath28 level with a @xmath29-diameter circular aperture . we detected 305,012 objects down to @xmath30 with sextractor @xcite and measured their narrow - band excess color ( @xmath31 ) and continuum color ( @xmath32 ) by combining existing very deep @xmath33- and @xmath34-band images of the sxdf ( see @xcite ) . figure [ fig : cc_nb816_modelobj ] shows the two color diagram of @xmath31 and @xmath32 for the @xmath23 detected objects , together with colors of model galaxies and galactic stars . in figure [ fig : cc_nb816_modelobj ] , model laes at @xmath0 have very red colors both in @xmath31 and @xmath32 because of a strong ly@xmath2 emission line and a continuum trough at @xmath35 , which are distinguished from colors of foreground objects . we defined the criteria for @xmath25 laes as @xmath36 and @xmath37 , and selected 515 laes from all the detected objects . these selection criteria isolate emission line objects with an observed - frame equivalent width of @xmath38 and a line flux of @xmath39 erg s@xmath21 @xmath40 , if a flat continuum spectrum is assumed . the contamination rate of our lae sample is estimated to be about 30% based on our spectroscopic follow - up observations described in section 3.2 . the surface density and number density of these 515 laes are estimated to be @xmath41 arcmin@xmath42 and @xmath43 mpc@xmath44 . we also calculate the surface density down to the same magnitude limit as @xcite ( @xmath45 ) and obtain @xmath46 arcmin@xmath42 which is consistent with that of @xcite . figure [ fig : density_cont_nb816_2 ] is the cosmic map showing the distribution of 515 laes at @xmath0 in the sxdf . the surveyed volume has a transverse dimension of 180 mpc @xmath3 180 mpc and a depth of @xmath4 mpc in comoving units . this is the first cosmic map , ever obtained , covering a @xmath47 mpc square area of the universe at any high redshifts ( @xmath48 ) . in figure [ fig : density_cont_nb816_2 ] , the laes have a very clumpy distribution , forming concentrations with a typical size of a few mpc , comparable in size to the proto - cluster found by @xcite . these concentrations are not isolated , but connected with one another by filamentary overdense regions . the elongated overdense region of laes found by @xcite at @xmath12 may be a segment of a descendant of such a filamentary structure . there are also found several voids of ellipsoidal shapes with sizes of 10 - 40 mpc in which almost no galaxy exists . the characteristic sizes of filaments and voids seen in our map are comparable to those of the present - day universe . thus , this map marks the discovery of primeval lsss at @xmath0 . we quantify the large - scale clumpiness of the galaxy distribution by estimating @xmath49 , the rms fluctuation of galaxy overdensity within a sphere of 20 mpc ( comoving ) radius . considering that the radial depth of the surveyed volume is about 40 mpc , we estimate the fluctuation by @xmath50/ \bar{\sigma}_{20}^2 $ ] , where @xmath51 is the rms surface overdensity within circles of 20 mpc radius , and @xmath52 and @xmath53 are the observed number and the mean number of laes in a circle @xcite . we obtain @xmath54 . this @xmath49 is comparable to or at least a half the value for the present - day lsss , i.e. @xmath55 , obtained from galaxies with @xmath56 @xcite . the characteristic shapes and the rms fluctuations indicate that these primeval lsss at @xmath57 are similar to the present - day lsss . we will present results of more detailed analyses for these primeval lsss , such as counts - in - cell , in our forthcoming paper . we find that among the dense concentrations seen in figure [ fig : density_cont_nb816_2 ] , the one at @xmath58 , @xmath59[j2000 ] has the highest density contrast , @xmath60 , with the @xmath61 significance level , where @xmath62 is the surface overdensity for a circle of 8 mpc radius . we refer to this concentration as region a. region b , a neighboring concentration at @xmath63,@xmath64[j2000 ] , also has a high density contrast of @xmath65 ( @xmath66 ) . in order to obtain three - dimensional distributions of laes in regions a and b , we carried out spectroscopy for laes with subaru / focas @xcite on 2003 december 20 , 23 , and 25 . we placed a slit mask on each region . another mask was put on a control field well separated from regions a and b. we made a 2-hour exposure for each mask , and obtained 22 spectra in total , whose spectral range and resolution are @xmath67=4900 - 9400 and @xmath68 , respectively . among the 22 objects , we classify one as an [ oiii ] emitter at @xmath69 , two as [ oii ] emitters at @xmath70 , and eight as securely identified laes with a characteristic asymmetric feature in ly@xmath2 line whose skewness exceeds 1.5 with the @xmath71 level . we show spectra of these emitters in figure [ fig : spec_disp_paper ] . since we have eight secure laes among the 11 ( = 8 + 2 + 1 ) identified objects , we estimate the contamination rate of our lae sample to be about 30% , which is consistent with those of other lae samples @xcite . the other 11 objects out of the 22 are emitters with an unresolved - single line at @xmath72 . since most unresolved - single line emitters have been found to be laes with a moderate velocity dispersion ( @xmath73 km s@xmath21 ; @xcite ) , we also regard these 11 objects as laes . thus we have 19 laes with spectroscopic redshifts as summarized in table [ tab : lae_spec ] . we plot in figure [ fig : threedmap_sxdss_plushist ] the three - dimensional distribution of laes for regions a and b. the histogram of figure [ fig : threedmap_sxdss_plushist ] shows that these regions have two clumps made of 10 laes with a significant excess . each of the clumps has a diameter of about 1 mpc in physical units ( 7 mpc in comoving units ) . for these clumps , the 1-dimensional velocity dispersion of galaxies is fairly small , @xmath74 km s@xmath21 ( clump a ) and @xmath75 km s@xmath21 ( b ) . although these clumps may not be collapsed , the formal virial masses can be calculated from velocity dispersion , @xmath76 , and radius , @xmath77 , with @xmath78 , where @xmath79 is the gravitational constant . we obtain @xmath80 ( clump a ) and @xmath81 ( b ) . the three - dimensional density contrast of laes , @xmath82 , for the average of these clumps is @xmath6 , which is comparable to those of present - day clusters , @xmath83 . these clumps have the highest density in our surveyed volume of @xmath84 mpc@xmath85 in comoving units . in the present - day universe , this volume typically contains two massive clusters with mass of @xmath86 @xcite . thus , the discovered clumps are likely proto - clusters which are ancestors of today s such clusters . a particularly interesting feature is that these clumps are very high concentrations of star - forming galaxies with a star - formation rate of @xmath87 yr@xmath21 . the star - formation rate density ( sfrd ) in these clumps inferred from ly@xmath2 fluxes is about 130 times higher than in the mean of the whole 1 deg@xmath1 field , mainly due to a large overdensity , @xmath88 , of laes . this very high sfrd implies that these clumps would be just producing a number of galaxies in a short period by a burst of galaxy formation . in contrast , present - day massive clusters are dominated by old early - type galaxies , and show a deficit of young star - forming galaxies @xcite . since the formation epochs of early - type galaxies are estimated to be @xmath89 or earlier ( e.g. @xcite ) , galaxies residing in these clumps are likely progenitors of the old early types seen in the core of present - day clusters . therefore , we are probably witnessing forming clusters where a number of present - day early - type galaxies are just emerging as star - forming galaxies by a burst of galaxy formation . the cold dark matter ( cdm ) model , the standard hierarchical scenario , predicts that very small initial fluctuations of mass density grow up gradually with time to evolve into galaxies and lsss . however , the observed distribution of laes appears much more inhomogeneous than the dark - matter distribution , since the amplitude of the matter fluctuations at @xmath0 is predicted to be only 1/5 of the current value . if one wants to accept the cdm model , a solution to this discrepancy is to assume that laes are biased tracers of mass in a way to enhance the mass fluctuations . previous observational studies of high-@xmath14 galaxies discuss the clustering bias @xcite . theoretically , such biasing can be produced , if galaxies are formed at rare , highest peaks of yet small fluctuations of matter density ( e.g. @xcite ) . from @xmath49 , we estimate that the bias parameter , @xmath90 , the ratio of galaxy overdensity to matter overdensity , is @xmath91 on scales of lsss ( 20 mpc ) , which is comparable to the prediction at @xmath92 from simulations of galaxy formation based on the cdm model @xcite . thus , the cdm model may be consistent with the early formation of lsss that we discovered , though the model needs to reproduce the characteristic shapes of observed lsss , long filaments and large voids . on the other hand , we estimate the @xmath90 value , in the cdm , which reproduces the observed number of forming clusters ( two in @xmath84 mpc@xmath85 ) with the size and the overdensity , @xmath88 , following the prescription of @xcite . we obtain @xmath93 for the two forming clusters , which is about ten times higher than the value for lss scales , although this estimation of @xmath90 is based on small statistics and might include a large error due to cosmic variance . this variety of @xmath90 could be due to the dependence of @xmath90 on the clustering scale or galaxy density @xcite . however , it is not yet clear how such a large difference in @xmath90 may be produced within the framework of the cdm model . our cosmic map may raise a question about how and where galaxies are formed in the early universe . adelberger , k. l. , steidel , c. c. , giavalisco , m. , dickinson , m. , pettini , m. , & kellogg , m. 1998 , , 505 , 18 ajiki , m. , et al . 2003 , , 126 , 2091 baugh , c. m. , benson , a. j. , cole , s. , frenk , c. s. , & lacey , c. g. 1999 , , 305 , l21 benson , a. j. , frenk , c. s. , baugh , c. m. , cole , s. , & lacey , c. g. 2001 , , 327 , 1041 bertin , e. & arnouts , s. 1996 , , 117 , 393 bruzual , g. & charlot , s. 2003 , , 344 , 1000 butcher , h. & oemler , a. 1984 , , 285 , 426 coleman , g. d. , wu , c .- c . , & weedman , d. w. 1980 , , 43 , 393 dressler , a. 1980 , , 236 , 351 giavalisco , m. & dickinson , m. 2001 , , 550 , 177 gunn , j. e. & stryker , l. l. 1983 , , 52 , 121 hayashino , t. , et al . 2003 , publications of the national astronomical observatory of japan , 7 , 33 hu , e. m. , cowie , l. l. , capak , p. , mcmahon , r. g. , hayashino , t. , & komiyama , y. 2004 , , 127 , 563 kashikawa , n. et al . 2002 , , 54 , 819 kauffmann , g. , colberg , j. m. , diaferio , a. , & white , s. d. m. 1999 , , 307 , 529 kinney , a. l. , calzetti , d. , bohlin , r. c. , mcquade , k. , storchi - bergmann , t. , & schmitt , h. r. 1996 , , 467 , 38 kodama , t. , arimoto , n. , barger , a. j. , & aragon - salamanca , a. 1998 , , 334 , 99 kodama , t. , et al . 2004 , , 350 , 1005 madau , p. 1995 , , 441 , 18 miley , g. k. , et al . 2004 , , 427 , 47 miyazaki , s. et al . 2002 , , 54 , 833 mo , h. j. & white , s. d. m. 2002 , , 336 , 112 rhoads , j. e. , et al . 2003 , , 125 , 1006 ouchi , m. et al . 2001 , , 558 , l83 ouchi , m. et al . 2003 , , 582 , 60 ouchi , m. , et al . 2004a , , 611 , 660 ouchi , m. , et al . 2004b , , 611 , 685 peebles , p. j. e. 1980 , the large - scale structure of the universe ( princeton : princeton univ . press ) reiprich , t. h. & b " ohringer , h. 2002 , , 567 , 716 seaborne , m. d. , et al . 1999 , , 309 , 89 sekiguchi , k. et al . 2004 , astrophysics and space science library , 301 , 169 shimasaku , k. et al . 2003 , , 586 , l111 shimasaku , k. , et al . 2004 , , 605 , l93 steidel , c. c. , adelberger , k. l. , dickinson , m. , giavalisco , m. , pettini , m. , & kellogg , m. 1998 , , 492 , 428 steidel , c. c. , adelberger , k. l. , shapley , a. e. , pettini , m. , dickinson , m. , & giavalisco , m. 2000 , , 532 , 170 venemans , b. p. et al . 2002 , , 569 , l11 venemans , b. p. , et al . 2004 , , 424 , l17 yagi , m. , kashikawa , n. , sekiguchi , m. , doi , m. , yasuda , n. , shimasaku , k. , & okamura , s. 2002 , , 123 , 66 ccclcc 02:17:45.03 & -05:28:42.5 & 26.8 & 5.749 & 2.0 & 5.9 + 02:20:13.36 & -04:51:09.3 & 26.7 & 5.742 & 1.3 & 3.7 + 02:20:26.14 & -04:52:34.3 & 25.7 & 5.717 & 0.8 & 2.4 + 02:20:19.79 & -04:52:29.9 & 26.4 & 5.706 & 1.0 & 2.9 + 02:17:49.13 & -05:28:54.2 & 26.2 & 5.694 & 5.0 & 14.5 + 02:17:50.00 & -05:27:08.2 & 27.4 & 5.693 & 0.7 & 1.9 + 02:17:47.32 & -05:26:48.0 & 25.5 & 5.692 & 0.7 & 2.1 + 02:17:48.48 & -05:31:27.0 & 26.6 & 5.687 & 1.4 & 3.9 + 02:17:45.27 & -05:29:36.1 & 26.6 & 5.687 & 4.1 & 11.7 + 02:17:43.35 & -05:28:07.1 & 26.2 & 5.683 & 7.0 & 20.0 + 02:18:23.82 & -05:32:05.3 & 27.1 & 5.678 & 1.0 & 2.8 + 02:18:19.13 & -05:33:11.7 & 27.1 & 5.674 & 2.0 & 5.7 + 02:18:20.88 & -05:33:21.9 & 26.5 & 5.670 & 0.2 & 0.7 + 02:18:14.43 & -05:32:49.1 & 26.5 & 5.670 & 1.0 & 2.9 + 02:20:21.52 & -04:53:15.0 & 26.7 & 5.669 & 1.8 & 5.1 + 02:18:22.61 & -05:33:37.8 & 27.6 & 5.648 & 1.5 & 4.4 + 02:18:17.35 & -05:32:22.8 & 26.6 & 5.642 & 2.9 & 8.2 + 02:17:43.31 & -05:31:35.0 & 26.5 & 5.624 & 2.5 & 7.1 + 02:17:45.88 & -05:27:14.5 & 25.6 & 5.600 & 0.9 & 2.4 + 02:17:51.14 & -05:29:35.3 & 27.2 & 1.188 & 1.7 & + 02:20:12.15 & -04:49:50.7 & 27.4 & 1.179 & 1.5 & + 02:20:12.84 & -04:50:22.5 & 24.5 & 0.623 & 11.9 &	we report the discovery of primeval large - scale structures ( lsss ) including two proto - clusters in a forming phase at @xmath0 . we carried out extensive deep narrow - band imaging in the 1 deg@xmath1 sky of the subaru / xmm - newton deep field , and obtained a cosmic map of 515 ly@xmath2 emitters ( laes ) in a volume with a transverse dimension of 180 mpc @xmath3 180 mpc and a depth of @xmath4 mpc in comoving units . this cosmic map shows filamentary lsss , including clusters and surrounding 10 - 40 mpc scale voids , similar to the present - day lsss . our spectroscopic follow - up observations identify overdense regions in which two dense clumps of laes with a sphere of @xmath5-mpc diameter in physical units are included . these clumps show about 130 times higher star - formation rate density , mainly due to a large overdensity , @xmath6 , of laes . these clumps would be clusters in a formation phase involving a burst of galaxy formation .
multiphoton interference is a nonclassical effect widely utilized in optical realizations of quantum - enhanced technologies @xcite and testing foundations of quantum mechanics . high visibility of multiphoton interference depends critically on the absence of distinguishing information between the interfering photons @xcite . while early experiments relied on spatial and spectral filtering to fulfill this requirement , a great deal of effort is currently being expended on the development of sources that guarantee suitable characteristics of the collected photons already at the production stage . such sources can offer substantially higher brightness , compatibility with integrated optics circuits , and strong photon number correlations , the last feature needed for example in device - independent quantum cryptography and randomness generation @xcite . a promising route to photon sources with well - defined , controllable characteristics is based on spontaneous parametric down - conversion in @xmath3 nonlinear waveguides @xcite . compared to bulk crystals , the phase matching conditions that define the nonlinear process in a waveguide assume a different form owing to the discreteness of transverse spatial modes propagating through the structure @xcite . this opens up new possibilities to engineer properties of the produced nonclassical radiation @xcite . in particular , generation of spatially pure photon pairs in a multimode waveguide has been recently reported in @xcite , with a high degree of spatial coherence verified via a heralded photon counting measurement of the beam quality factors based on free - space diffraction . single spatial modes for the generated photons were selected by exploiting the effects of intermodal dispersion in the down - conversion process . this technique overcomes waveguide manufacturing limitations for shorter wavelengths . prospectively , it can also be used to produce spatial photonic entanglement @xcite . in this paper , we study experimentally two - photon interference using a periodically poled potassium titanyl phosphate ( ppktp ) waveguide source of spatially pure photon pairs that relies on the technique exploiting intermodal dispersion @xcite . measurements performed in two setups : a hong - ou - mandel interferometer @xcite in the common - path configuration @xcite and a shih - alley source of polarization entanglement @xcite yield visibilities consistently exceeding @xmath0 . remarkably , these values have been obtained without any spatial filtering of the generated photons , and only a coarse selection of their spectral range . these results provide a compelling verification of the high - quality modal characteristics of the generated photons , paving way towards employing them in more complex multiphoton interference experiments . this paper is organized as follows . first , in sec . [ sec : method ] we briefly review parametric down - conversion in a nonlinear multimode waveguide and discuss operating conditions that ensure generation of indistinguishable photon pairs . next , sec . [ sec : setup ] descibes the experimental setup and alignment procedures . results of two - photon interference measurements are presented and discussed in sec . [ sec : interference ] . finally , sec . [ sec : conclusions ] concludes the paper . we begin by reviewing the technique to control the spatial characteristics of photon pairs generated in a multimode nonlinear waveguide and discussing the operation of the source that ensures high - visibility two - photon interference . the waveguide is chosen to realize a type - ii down - conversion process in which a pump photon @xmath4 is converted into a pair of orthogonally polarized photons @xmath5 and @xmath6 . the basic object in our analysis is the phase matching function , which in the quantum picture of parametric down - conversion is interpreted as the probability amplitude that a pair of photons with wavelengths @xmath7 and @xmath8 will be generated from a pump photon with the wavelength defined by energy conservation . while in a bulk medium the phase matching function depends also on continuous spatial degrees of freedom of the photons taking part in the process , parameterized for example with transverse wave vectors @xcite , in the case of a multimode waveguide the function depends on the specific triplet of spatial modes of the photons @xmath4 , @xmath5 , and @xmath6 involved in the process . mm long ppktp waveguide analogous to the structure used in the experiment , with the pump field prepared in the fundamental spatial mode of the waveguide . separate phase matching bands correspond to different combinations of the spatial modes of @xmath5 and @xmath6 photons . the energy conservation condition is depicted with a solid white line . ( b ) the joint and marginal spectra of the generated photon pairs . application of coarse spectral filtering represented by dashed lines allows one to select the spectral region where both the photons are generated in the fundamental waveguide modes @xmath9 and @xmath10 . [ fig : phmatch],width=453 ] as illustrated in fig . [ fig : phmatch](a ) , intermodal dispersion between waveguide modes makes the phase matching condition satisfied in general at different regions of the plane spanned by the wavelengths @xmath7 and @xmath8 , resulting in a series of bands whose width is inversely proportional to the waveguide length . in particular , if the pump is prepared in the fundamental waveguide mode @xmath11 , then the phase matching function for generating down - converted photons in fundamental modes @xmath9 and @xmath10 is separated from processes involving higher - order @xmath5 and/or @xmath6 modes , provided that the waveguide is long enough to sufficiently narrow the relevant bands . this condition is satisfied for the 1 mm long structure in our experiment , which was simulated in fig . [ fig : phmatch](a ) . it needs to be stressed that this scheme relies critically on the pump beam prepared in the fundamental spatial mode : coupling the pump into modes other than @xmath11 generates contributions from phase matching bands involving higher - order spatial modes for @xmath5 and @xmath6 photons that overlap spectrally with the desired @xmath12 band . if a cw pump is employed to induce the down - conversion process , the wavelengths of the down - converted photons must satisfy the constraint of constant total energy , which is symmetric with respect to swapping @xmath7 and @xmath8 and for the wavelength range depicted in fig . [ fig : phmatch](a ) becomes a nearly straight line tilted at 45@xmath13 with respect to the graph axes . in contrast , the slope of the phase matching bands is noticeably different , as the @xmath5 and @xmath6 photons are generated in the type - ii process and they propagate as extraordinary and ordinary rays . consequently , pairs of photons are produced for combinations of wavelengths that form a set of islands , shown in fig . [ fig : phmatch](b ) , located at crossings of the energy conservation line with phase matching bands . the island corresponding to the process @xmath12 can be separated from other processes by coarse spectral filtering of one or both of the generated photons . it is worth noting that overlap may occur between islands corresponding to generation of photon pairs in higher - order spatial modes owing to approximate mode degeneracy with respect to their propagation constants . while the above scheme ensures generation of photons in fundamental spatial modes , two - photon interference requires also spectral indistinguishability within the produced pairs . this condition is not satisfied automatically , as the separation between the spectra of the @xmath5 and @xmath6 photons will vary with the wavelength of the pump photons , which shifts the energy conservation line in the diagonal direction of fig . [ fig : phmatch](a ) . however , the graph indicates that for a suitable pump wavelength the regime of spectral degeneracy should be achievable . because our numerical simulation of the waveguide is sensitive to details of the actual refractive index profile , this regime of operation needs to be identified by experimental means . we found that a suitable procedure , described in detail in sec . [ sec : setup ] , was to measure the individual photon spectra in the heralded regime and to carefully tune the pump wavelength until the spectral profiles of @xmath5 and @xmath6 photons were matched within the resolution of the measuring apparatus . the waveguide source of photon pairs used in our experiments is shown schematically in fig . [ fig : setup ] . its heart was a 1 mm long ppktp structure ( advr inc . ) temperature stabilized at @xmath14 using a thermoelectric cooler . a series of waveguides localized just beneath the surface had lateral transverse dimensions of approx . @xmath15 m , effective depths of approx . @xmath16 m , and poling periods designed for efficient type - ii second harmonic generation in the 800 nm wavelength region . at these wavelengths the waveguides supported at least 8 transverse spatial modes ( 4 for each polarization ) . the structure was placed between two infinity corrected @xmath17 microscope objectives with numerical apertures @xmath18 for incoupling and @xmath19 for outcoupling . a magnified image of the waveguide illuminated with incoherent white light along with measured spatial profiles of the fundamental pump , @xmath5 , and @xmath6 modes at their respective wavelengths is shown in fig . [ fig : modes ] . , half - wave plate ; pol , glan - taylor polarizer ; if , interference filter ; dm , dichroic mirror ; pbs , polarizing beam splitter ; mmf , multimode fiber ; smf , single mode fiber ; spcm , single photon counting module ; bsc , babinet - soleil compensator ; bs , non - polarizing beam splitter , ndf , neutral density filter ; cf , color filter . , width=396 ] the down - conversion process was pumped by a narrowband ( linewidth @xmath20 ) cw diode laser ( toptica bluetune ) beam that could be tuned around the central wavelength of 400 nm . the pump beam was delivered to the setup with a single - mode fiber . the pump power and polarization were controlled using a glan - taylor polarizer placed between two half - wave plates . typical pump power incident on the waveguide input facet was @xmath21w , whereas the pump power coupled into the waveguide is estimated to be at least @xmath22w . the estimate is based on the power measured at the exit of the waveguide , assuming no propagation losses . this corresponds to coupling efficiency of at least @xmath23 , which is similar to the values reported in other experiments @xcite . this figure could be improved by approx . @xmath24 by using a waveguide sample with anti - reflection coated facets . as preparing the pump field in the fundamental transverse waveguide mode was crucial for the spatial purity of the generated photons , after the outcoupling objective the remainder of the pump beam was directed with a dichroic mirror to a ccd camera . its purpose was to monitor the quality of the pump spatial mode through a comparison with a reference profile generated beforehand in the sum - frequency process @xcite . an exemplary image of the pump beam mode is shown in fig . [ fig : modes](b ) . although the waveguide cross section is elongated and asymmetric in the vertical direction , the fundamental pump beam mode , being well - confined within the structure , bears only a minor signature of this asymmetry . this enabled us to excite selectively the fundamental pump mode by direct coupling using a laser beam spatially filtered by a single mode fiber and focused to an appropriate diameter with the microscope objective . an auxiliary infrared beam from a modelocked ti : sapphire laser , coupled predominantly into the fundamental spatial mode of the waveguide at the down - conversion wavelength , was used to identify the waveguide with a suitable poling period and to align optical elements following the source . the profiles of @xmath5 and @xmath6 fundamental modes presented in fig . [ fig : modes](c , d ) were recorded with the help of this beam . mode excited with the pump laser ( b ) as well as the fundamental @xmath5 mode ( c ) and @xmath6 mode ( d ) excited with the auxiliary ti : sapphire laser laser beam . the intensity distributions were obtained by imaging the output facet of the waveguide onto a ccd camera using the outcoupling objective and a 200 mm focal length lens.,width=453 ] in order to verify the spectral indistinguishability of the generated photon pairs , in the first step we measured individual spectra of heralded photons using the setup depicted in fig . [ fig : setup](a ) . photon pairs were sent through a color filter ( cut - off wavelength @xmath25 nm ) and separated on a polarizing beam splitter . the output paths for the photons could be swapped with the help of a half - wave plate placed before the polarizing beam splitter . at the output , one photon was used as a herald , while the second one was transmitted through a @xmath26 nm full width at half maximum ( fwhm ) interference filter mounted on a motorized rotation stage . the photons were subsequently coupled using @xmath27 mm focal length aspheric lenses into @xmath28 m core diameter , @xmath29 na multimode fibers connected to single photon counting modules ( perkin elmer spcm - aqrh-14-fc ) . coincidence events were counted within a 3 ns window . the rotation angle of the interference filter was calibrated in terms of the transmitted central wavelength using a ti : sapphire beam and a spectrometer . ppktp phase matching properties make the spectral characteristics of the generated photons strongly dependent on the pump wavelength , as can be inferred from fig . [ fig : phmatch](a ) . using the overlap of the single photon spectra as the optimization criterion , we fine - tuned the wavelength of the pump laser , obtaining the best match at @xmath30 nm , shown in fig . [ fig : singlephotonspectra ] . in the same graph , we also depict power transmission profiles of two interference filters used in measurements of two - photon interference described in sec . [ sec : interference ] . the fwhm widths of the filters are approximately 11 nm and 3 nm . the broader filter encompasses the entire spectra of photons generated in fundamental spatial modes . as illustrated in fig . [ fig : phmatch](b ) , its principal role in the setup is to cut off down - conversion processes involving higher spatial modes that occur in distinct frequency regions @xcite . nm fwhm bandpass filter . coincidence counts ( points , left scale ) collected over 5 s intervals are fitted with gaussian functions ( solid lines ) assuming poissonian errors . experimentally obtained spectral profiles of the interference filters used in measurements of two - photon interference : 11 nm , dashed - dotted line ; 3 nm , dotted line ( right scale ) . , width=340 ] we have tested nonclassical interference between photons generated in the waveguide using two setups . the first one was the hong - ou - mandel interferometer @xcite in the common - path configuration @xcite depicted in fig . [ fig : setup](b ) . in this case , a pair of orthogonally polarized photons was rotated with a half - wave plate by @xmath31 and sent to a polarizing beam splitter . the time delay between the photons was adjusted using a babinet - soleil compensator . zero delay was found by measuring spectral fringes with help of the ti : sapphire beam . photons leaving the polarizing beam splitter were coupled into multimode fibers and counted using the same configuration as before . in fig . [ fig : homdip ] we show measured coincidence rates as functions of the time delay . the depth of the hong - ou - mandel dip with respect to the reference level of fully distinguishable particles , determined from gaussian fits , is @xmath32 for 11 nm filter and @xmath33 for 3 nm filter . these figures confirm that the photons are highly indistinguishable in both the spectral and the spatial degrees of freedom . it is seen that the coincidence count rate is minimized for a non - zero delay of @xmath34 ps , which compensates the temporal walk - off within photon pairs due to waveguide birefringence . the second test of two - photon interference was carried out through preparation of polarization entanglement in the shih - alley configuration @xcite . in this setup , shown in fig . [ fig : setup](c ) , two photons in orthogonal horizontal ( @xmath5 ) and vertical ( @xmath6 ) polarizations impinge on a nonpolarizing beam splitter at a nearly normal incidence . their polarizations are analyzed at the outputs using half - wave plates and polarizers . when the two photons emerge at different ports of the beam splitter , their postselected polarization state takes the maximally entangled form @xmath35 . the presence of entanglement can be verified by detecting photons in linear polarization bases . specifically , if one of the photons is detected at @xmath31 , the coincidence rate with the second detector measuring photons at an angle @xmath36 is proportional to @xmath37 where the real parameter @xmath38 characterizes fringe visibility . it can be shown @xcite that for temporally compensated photons this parameter is theoretically equal to the depth of the hong - ou - mandel dip , hence the use of the same symbol for both quantities . of the collected linear polarization , when conjugate photons were detected in horizontal ( @xmath5 ) , vertical ( @xmath6 ) , diagonal ( @xmath39 ) , and antidiagonal ( @xmath40 ) polarizations , shown for ( a ) 11 nm and ( b ) 3 nm interference filters placed at the setup entrance.,width=491 ] before measuring polarization fringes , the babinet - soleil compensator placed before the non - polarizing beam splitter was set to the position that minimized the coincidence rate in the hong - ou - mandel interferometer . in fig . [ fig : polarizationfringes ] we present coincidence count rates between one photon projected onto horizontal ( @xmath5 ) , vertical ( @xmath6 ) , diagonal ( @xmath39 ) , or antidiagonal ( @xmath40 ) polarization and the second photon detected in linear polarization at an angle @xmath36 . visibilities determined from sinusoidal fits to interference fringes are collected in tab . [ tab : visibilities ] . fringe visibilities for diagonal and antidiagonal polarizations match within their uncertainties the depths of hong - ou - mandel dips . .visibilities of polarization fringes obtained from measurements for multimode ( mmf ) and single mode ( smf ) fibers in the setup for polarization entanglement preparation . absolute uncertainties of presented values are approximately 1% . the last column specifies the source brightness for each combination of the interference filters and coupling fibers . [ cols="^,^,>,>,>,>,^ " , ] in order to estimate the effects of nonideal spatial overlap of the interfering photons , we repeated the measurements of polarization correlations using single mode fibers to deliver photons to detectors . fringe visibilities determined from these data are also presented in tab . [ tab : visibilities ] . it is seen that for measurements in the diagonal basis , fringe visibilities increased by approximately by 5% . this can be mainly attributed to transverse walk - off of orthogonally polarized photons in the babinet - soleil compensator , a minor discrepancy between fundamental waveguide mode profiles for orthogonal polarizations that can be noticed in fig . [ fig : modes](c , d ) , and contributions from down - conversion processes involving residually excited higher - order pump modes . the brightness of the source defined as the ratio of detected photon pairs to the pump power incident on the waveguide is specified in the last column of tab . [ tab : visibilities ] . typical ratio of coincidence to single count rates in our experiments was 8.9% for the 11 nm filter . although standard optical elements were antireflection - coated , contributions to non - unit transmission came from the waveguide - air interface ( @xmath41 ) , outcoupling objective ( @xmath42 ) , soleil - babinet compensator ( @xmath43 ) , interference filter ( @xmath44 ) , coupling into multimode fibers ( @xmath45 ) are further combined with the non - unit detector efficiency ( @xmath46 ) . other relevant effects could include intra - waveguide losses and additional parasitic down - conversion processes . it is worthwhile to note substantial improvement of indistinguishability compared to pioneering experiments with the ppktp nonlinear medium @xcite . recently , visibilities exceeding 95% have been reported for two - photon interference experiments based on waveguide sources at 1.3 @xmath2 m and 1.5 @xmath2 m telecom bands @xcite , where difficulties to manufacture single - mode structures are less severe . in the 800 nm wavelength region , the brightness @xmath47 pairs / s / mw and the average visibility above 97% for photons filtered through a single - mode fiber have been achieved in ref . @xcite , where a pair of two crossed 20 mm long bulk ppktp crystals was used to realize a type-0 down - conversion process . we studied experimentally distinguishability of photons generated via spontaneous parametric down - conversion in a multimode nonlinear ppktp waveguide . measurements taken in two different setups implementing two - photon interference yielded visibilities robustly above 90% without resorting to spatial filtering . this directly demonstrates that photon sources based on exploiting intermodal dispersion in multimode structures , a key technique used in our setup , are suitable for multiphoton interference experiments . the benefits of waveguide sources in photonic quantum technologies , such as high brightness and integrability , can be therefore extended also to spectral regions where single - mode structures are not readily available . we thank r. demkowicz - dobrzaski , p. michelberger , k. thyagarajan , and w. wasilewski for insightful discussions . this work was supported by the polish ncbir under the era - net chist - era project quasar and the foundation for polish science team project cofinanced by the eu european regional development fund .	we report measurements of two - photon interference using a cw - pumped type - ii spontaneous parametric down - conversion source based on a multimode perodically poled potassium titanyl phosphate ( ppktp ) waveguide . we have used the recently demonstrated technique of controlling the spatial characteristics of the down - conversion process via intermodal dispersion to generate photon pairs in fundamental transverse modes , thus ensuring their spatial indistinguishability . good overlap of photon modes within the pairs has been verified using the hong - ou - mandel interferometer and the preparation of polarization entanglement in the shih - alley configuration , yielding visibilities consistently above @xmath0 . 99 j. l. obrien , a. furusawa , and j. vukovi , `` photonic quantum technologies , '' nature photon . * 3 , * 687695 ( 2009 ) . m. ukowski , a. zeilinger , m. a. horne , and a. k. ekert , `` ` event - ready - detectors ' bell experiment via entanglement swapping , '' , 4287 - 4290 ( 1993 ) . b. g. christensen , k. t. mccusker , j. b. altepeter , b. calkins , t. gerrits , a. e. lita , a. miller , l. k. shalm , y. zhang , s. w. nam , n. brunner , c. c. w. lim , n. gisin , and p. g. kwiat , `` detection - loophole - free test of quantum nonlocality , and applications , '' , 130406 ( 2013 ) . m. d. c. pereira , f. e. becerra , b. l. glebov , j. fan , s. w. nam , a. migdall , `` demonstrating highly symmetric single - mode , single - photon heralding efficiency in spontaneous parametric downconversion , '' , 1609 - 1611 ( 2013 ) . k. banaszek , a. b. uren , i. a. walmsley `` generation of correlated photons in controlled spatial modes by downconversion in nonlinear waveguides '' , 13671369 ( 2001 ) . a. b. uren , c. silberhorn , k. banaszek , and i. a. walmsley , `` efficient conditional preparation of high - fidelity single photon states for fiber - optic quantum networks , '' , 093601 ( 2004 ) . m. fiorentino , s. m. spillane , r. g. beausoleil , t. d. roberts , p. battle and m. w. munro `` spontaneus parametric down - conversion in periodically poled ktp waveguides and bulk crystals , '' , 74797488 ( 2007 ) . a. martin , v. cristofori , p. aboussouan , h. herrmann , w. sohler , d.b . ostrowsky , o. alibart , and s. tanzilli , `` integrated optical source of polarization entangled photons at 1310 nm , '' , 10331041 ( 2009 ) . t. zhong , f. n. c. wong , t. d. roberts , and p. battle , `` high performance photon - pair - source based on fiber - coupled periodically poled ktiopo@xmath1 waveguide , '' , 1201912030 ( 2009 ) . g. harder , v. ansari , b. brecht , t. dirmeier , c. marquardt and c. silberhorn , `` an optimized photon pair source for quantum circuits , '' , 1397513985 ( 2013 ) . m. karpiski , c. radzewicz , and k. banaszek , `` experimental characterization of three - wave mixing in a multimode nonlinear ktiopo@xmath1 waveguide , '' appl . . lett . * 94 * , 181105 ( 2009 ) . r.machulka , j.svozilk , j.soubusta , j.peina jr . , o.haderka , `` spatial and spectral properties of fields generated by pulsed second - harmonic generation in a periodically poled potassium - titanyl - phosphate waveguide , '' phys . rev . a * 87 * , 013836 ( 2013 ) . b. brecht , a. eckstein , a. christ , h. suche , and c. silberhorn , `` from quantum pulse gate to quantum pulse shaper engineered frequency conversion in nonlinear optical waveguides , '' new . j. phys . * 13 * , 065029 ( 2011 ) . m. karpiski , c. radzewicz , and k. banaszek , `` disperion - based control of modal characteristics for parametric down - conversion in a multimode waveguide , '' , 878880 ( 2012 ) . p. j. mosley , a. christ , a. eckstein , and c. silberhorn , `` direct measurement of the spatial - spectral structure of waveguided parametric down - conversion , '' , 233901 ( 2009 ) . m. karpiski , c. radzewicz , and k. banaszek , `` generation of spatially pure photon pairs in a multimode nonlinear waveguide using intermodal dispersion , '' proc . spie * 8518 * , quantum communications and quantum imaging x , 85180j ( 2012 ) . c. k. hong , z. y. ou , and l. mandel `` measurement of subpicosecond time intervals between two photons by interference , '' , 20442046 ( 1987 ) . w. p. grice , r. erdmann , i. a. walmsley , and d. branning , `` spectral distinguishability in ultrafast parametric down - conversion , '' , r2289r2292 ( 1998 ) . h. shih and c. o. alley , `` new type of einstein - podolsky - rosen - bohm experiment using pairs of light quanta produced by optical parametric down convertion , '' , 29212924 ( 1988 ) . t. zhong , x. hu , f. n. c. wong , k. k. berggen , t. d. roberts and p. battle , `` high - quality fiber - optics polarization entanglement distribution at 1.3@xmath2 m telecom wavelength , '' , 13921394 ( 2010 ) . m. h. rubin `` transverse correlation in optical spontaneous parametric down - convertion , '' , 53495360 ( 1996 ) . j. p. torres , k. banaszek , and i. a. walmsley `` engineering nonlinear optic sources of photonic entanglement , '' progress in optics * 56 * , 227331 ( 2011 ) . h. herrmann , x. yang , a. thomas , a. poppe , w. sohler and c.silberhorn `` post - selection free , integrated optical source of non - degenerate , polarization entangled photon pairs , '' , 2798127991 ( 2013 ) . f. steinlechner , p. trojek , m. jofre , h. weier , d. perez , t. jennewein , r. ursin , j. rarity , m. w. mitchell , j. p. torres , h. weinfurter and v. pruneri , `` a high - brightness source of polarization - entangled photons optimized for applications in free space , '' , 96409649 ( 2012 ) .
on note @xmath0a , b[$ ] o @xmath1 et @xmath2 $ ] ladhrence de @xmath3 dans @xmath4 . soit @xmath5 un corps et @xmath6 . on se donne un espace probabilis @xmath7 sur lequel existent une famille croissante de tribus @xmath8 et une famille dcroissante de tribus @xmath9 . suivant yasue @xcite , on introduit la on note @xmath10 lensemble des processus @xmath11 dfinis sur @xmath12 , valeurs dans @xmath13 et tels que : @xmath11 soit @xmath14 et @xmath15 adapt , pour tout @xmath16 @xmath17 , lapplication @xmath18 de @xmath19 dans @xmath20 est continue , pour tout @xmath21 les quantits @xmath22 $ ] , et @xmath23 $ ] , existent dans @xmath20 , et enfin les applications @xmath24 et @xmath25 sont continues de @xmath3 dans @xmath20 . + le complt de @xmath26 pour la norme @xmath27 , est encore not @xmath10 , et simplement @xmath28 quand @xmath29 . les quantits @xmath30 et @xmath31 sont introduites par edward nelson dans sa thorie dynamique des diffusions browniennes ( _ _ cf__@xcite ) . soit @xmath32 linjection @xmath33 . on note : @xmath34 . le problme dextension consiste trouver un oprateur @xmath35 satisfaisant : * ( recollement ) @xmath36 sur @xmath37 , * ( @xmath4-linarit ) @xmath38 est @xmath4-linaire , * ( reconstruction ) si lon note @xmath39 o @xmath40 et @xmath41 sont des @xmath4-formes linraires , on suppose que lapplication de @xmath42 dans @xmath42 : @xmath43 est inversible . loprateur @xmath38 tend donc la drivation classique ( _ cf_(i ) ) en un oprateur linaire ( _ cf_(ii ) ) sur @xmath28 et la connaissance de @xmath44 induit celle de @xmath45 et @xmath46 ( _ cf_(iii ) ) . on obtient : les seuls oprateurs @xmath47 vrifiant ( i ) , ( ii ) et ( iii ) sont : @xmath48 on crira @xmath49 et @xmath50 . la dfinition des itrs de @xmath38 et @xmath51 ncessite lextension de ces oprateurs aux processus valeur complexe . dans la suite , on tend @xmath38 et @xmath51 par @xmath52linarit aux processus complexes , _ i.e. _ pour tous @xmath53 , @xmath54 . . on note @xmath55 lensemble des processus @xmath56 tels que pour tout @xmath57 , @xmath58 existe en tout point de @xmath3 . on donne la dfinition ( [ bonnesdiffusions ] ) un ensemble @xmath59 qui permet de montrer que @xmath28 nest pas trivial . en effet @xmath60 ( _ cf _ @xcite p.26 ) . le calcul de @xmath61 combine de faon non triviale les quantits @xmath30 et @xmath31 . titre dexemple , on obtient sur @xmath62 , @xmath63 . la partie relle de @xmath64 concide donc avec lacclration postule par nelson comme quantit la plus pertinente pour dcrire une notion dacclration pour une diffusion brownienne ( cf @xcite p.@xmath65 ) . en utilisant @xmath66 , on construit des analogues stochastiques doprateurs diffrentiels non linaires . [ stochastisation ] [ stoca ] on appelle plongement stochastique , relatif lextension @xmath67 , dun oprateur @xmath68 qui scrit sous la forme : @xmath69 avec @xmath70 , @xmath71 , loprateur @xmath72 agissant sur @xmath55 . + un oprateur @xmath68 crit sous la forme : @xmath73 @xmath74 , est plong en @xmath75 agissant sur un sous - ensemble de @xmath28 dpendant de certaines proprits de @xmath76 . un oprateur de la forme @xmath77 peut se rcrire @xmath78 qui se plonge alors en @xmath79 . ce dernier nest gal @xmath80 que dans certains cas ( cf @xcite p.@xmath81 ) . ceci montre en particulier que le plongement stochastique nest pas une application , il dpend du choix dcriture de loprateur . la notion de plongement doprateur stend de faon naturelle celle de plongement dquation dfinie par un oprateur @xmath68 dordre @xmath82 : @xmath83 . on dfinit lquation plonge par : @xmath84 o @xmath85 . on soccupe dsormais du cas lagrangien . on appelle lagrangien admissible une fonction @xmath86 de classe @xmath87 en sa premire variable @xmath88 et holomorphe en sa deuxime variable @xmath89 , et relle quand @xmath89 est relle . lquation @xmath90 sappelle quation deuler - lagrange . soit @xmath91 un lagrangien admissible . le plongement stochastique de ( [ el ] ) est donn par @xmath92 on sait que lquation ( [ el ] ) provient dun principe de moindre action ( cf @xcite p.@xmath93 ) . existe - il un principe de moindre action stochastique permettant lobtention de lquation ( [ els ] ) ? nous montrons dans @xcite chap.@xmath94 que tel est bien le cas et on donne un lemme montrant la cohrence de la procdure de plongement vis - - vis des principes de moindre action ainsi dfinis . considrons le lagrangien admissible @xmath95 o @xmath96 et @xmath97 est une fonction de classe @xmath87 . lquation ( [ el ] ) associe est lquation de newton @xmath98 . lquation de newton plonge est alors @xmath99 et concide avec lquation deuler - lagrange plonge ( [ els ] ) . on se propose dtudier un rsultat sur la densit dun processus solution de cette quation . on donne dans @xcite p.@xmath100 , suivant @xcite et @xcite , un espace sur lequel nous pourrons calculer les drives du premier ordre @xmath30 et @xmath31 et les drives du second ordre @xmath101 , @xmath102 , @xmath103 et @xmath104 . prenons @xmath1050,1[$ ] . soit @xmath106 un mouvement brownien standard dans @xmath107 dfini sur un espace probabilis filtr @xmath108 . * @xmath11 est solution sur @xmath19 de leds : @xmath109 o @xmath110 , @xmath111 et @xmath112 sont des fonctions mesurables vrifiant lhypothse : il existe une constante @xmath113 telle que pour tous @xmath114 : + @xmath115 et @xmath116 , * pour tout @xmath16 , @xmath117 possde une densit @xmath118 en @xmath119 , * en posant @xmath120 , pour tout @xmath121 , pour tout @xmath122 , pour tout ouvert born @xmath123 , * les fonctions @xmath124 et @xmath125 appartiennent @xmath126 , sont bornes et toutes leurs drives du premier et second ordre sont bornes . on notera @xmath127 ( resp . @xmath128 ) le sous - ensemble de @xmath59 form par les diffusions do nt le coefficient est constant gal @xmath129 ( resp . do nt le drift est un gradient ) , et on pose @xmath130 . si @xmath140 alors il existe des fonctions @xmath141 et @xmath142 diffrentiables sur @xmath143 telles que + @xmath144 car @xmath145 et @xmath124 est un gradient . on choisit @xmath146 . les fonctions @xmath141 et @xmath142 sont galement introduites par nelson dans @xcite p.@xmath147 . + on pose @xmath148 et @xmath149 . * dmonstration . * des expressions @xmath149 et @xmath154 , on dduit + @xmath155 . lquation de newton plonge peut scrire @xmath156 car @xmath97 est rel . + or @xmath157 . donc @xmath158 et avec ( [ deriv_fonc ] ) il vient + @xmath159 . le lemme de schwarz donne : + @xmath160 , et @xmath161 $ ] , et par consquent : + @xmath162 . en intgrant sur @xmath143 les fonctions des deux membres de la dernire quation , il apparat des constantes quon peut rendre nulles en ajoutant une fonction de @xmath163 convenable dans @xmath142 . le rsultat sen dduit . la partie relle de lquation de newton plonge concide avec lquation de newton stochastique propose par nelson dans sa thorie dynamique des diffusions browniennes ( @xcite p.@xmath165 ) . sa partie imaginaire correspond lquation @xmath166 . nous conjecturons que cette dernire impose que le drift de @xmath11 doit tre un gradient , et donc quil nest pas utile de le supposer dans le thorme ( [ schro ] ) . 7 arnold v.i . , _ mathematical methods of classical mechanics _ , 2d edition , springer , 1989 . cresson j. , darses s. , stochastic embedding of dynamical systems , arxiv : math.pr/0509713 , 112.p , 2005 . millet a , nualart d. , sanz m. , integration by parts and time reversal for diffusion processes , annals of probability 1989 , vol . 17 , no . 1 , 208 - 238 . nelson e. , _ dynamical theories of brownian motion _ , second edition , princeton , 2001 . thieullen m. , second order stochastic differential equations and non - gaussian reciprocal diffusions , proba . theory and rel . fields 97 , 231 - 257 ( 1993 ) . yasue k , stochastic calculus of variations , journal of functional analysis 41 , 327 - 340 ( 1981 ) .	on dfinit un oprateur agissant sur des processus stochastiques , qui tend la drivation classique sur les fonctions dterministes diffrentiables . on utilise cet oprateur pour dfinir une procdure associant aux oprateurs diffrentiels et quations diffrentielles ordinaires leurs analogues stochastiques . elle est appele plongement stochastique . en plongeant les systmes lagrangiens , nous obtenons une quation deuler - lagrange stochastique , qui dans le cas des systmes lagrangiens naturels est appele quation de newton plonge . cette dernire contient lquation de newton stochastique introduite par nelson dans sa thorie dynamique des diffusions browniennes . enfin , on considre une diffusion drift gradient , coefficient de diffusion constant et possdant une densit de probabilit . on dmontre alors quune condition ncessaire pour que cette diffusion soit solution de lquation de newton plonge , est que sa densit soit le carr du module dune fonction donde solution dune quation de schrdinger linaire . _ pour citer cet article : a. nom1 , a. nom2 , c. r. acad . sci . paris , ser . i 340 ( 2005 ) . _ 0.5 * abstract * 0.5stochastic embedding of lagrangian systems . we define an operator which extends classical differentiation from smooth deterministic functions to certain stochastic processes . based on this operator , we define a procedure which associates a stochastic analog to standard differential operators and ordinary differential equations . we call this procedure stochastic embedding . by embedding lagrangian systems , we obtain a stochastic euler - lagrange equation which , in the case of natural lagrangian systems , is called the embedded newton equation . this equation contains the stochastic newton equation introduced by nelson in his dynamical theory of brownian diffusions . finally , we consider a diffusion with a gradient drift , a constant diffusion coefficient and having a probability density function . we prove that a necessary condition for this diffusion to solve the embedded newton equation is that its density be the square of the modulus of a wave function solution of a linear schrdinger equation . _ to cite this article : a. nom1 , a. nom2 , c. r. acad . sci . paris , ser . i 340 ( 2005 ) . _ probability theory / dynamical systems ,
the ads / cft correspondence @xcite has provided a new approach to tackle strongly coupled theories . this has recently boosted the search for string realizations of qcd - like theories @xcite . a crucial ingredient in these realizations is a ( compact ) warped extra dimension that plays the role of the energy scale in the qcd - like theory . it is therefore interesting to look for properties of qcd in the strong regime that can be derived from weakly coupled theories in five - dimensions . examples of this type of properties have already been found in high - energy hadron scattering @xcite , string breaking @xcite , hadron form factors and hadron spectroscopy @xcite-@xcite . in refs . @xcite a five - dimensional model was proposed to study the breaking of the chiral symmetry in qcd . the model was described in terms of infinite weakly coupled resonances , similar to qcd in the large-@xmath0 limit . the vector sector was studied and several relations among couplings and masses were derived based only on the ( warped ) five - dimensionality of the space . the predictions of the model showed a good agreement with qcd . here we will extend this analysis to the scalar and pseudoscalar sector . we will calculate the scalar and pseudoscalar two - point correlator and show that they have a similar behavior to that in qcd . we will obtain the masses and couplings of the resonances , pointing out the implications of working with a warped extra dimension . we will also calculate the scalar form factor of the pseudo - goldstone boson ( pgb ) and the pseudo(scalar ) contributions to the @xmath1 coefficients of the low - energy chiral lagrangian . a prediction for the quark masses will also be given . we will compare our results with the qcd experimental data whenever this is available . the 5d model proposed to study the properties of qcd with 3 flavors consists in a theory where the chiral symmetry u(3)@xmath2u(3)@xmath3 is gauged in the 5d bulk . is broken by the anomaly that , although it will not be studied here , can also be incorporated in extra - dimensional models along the lines of ref . ] parity is defined as the interchange @xmath4 . the bulk fields are the gauge bosons @xmath5 and a complex scalar field @xmath6 transforming as a ( @xmath7,@xmath8 ) . this scalar plays the role of the operator @xmath9 in qcd whose vacuum expectation value ( vev ) is responsible for the breaking of the chiral symmetry . the 5d metric in conformal coordinates is defined as @xmath10 where @xmath11 is the warp factor . we will work within ads@xmath12 @xmath13 where @xmath14 is the ads curvature radius . the ads@xmath12 metric will guarantee conformal invariance of the model at high energies . the fifth dimension is assumed to be compact , @xmath15 @xcite . the boundary at @xmath16 generates a mass gap in the model ( breaks the conformal symmetry at energies @xmath17 ) , while the boundary at @xmath18 is only needed to regulate the theory . when performing calculations one must take the limit @xmath19 and eliminate the divergences that one encounters by properly adding uv - boundary counterterms @xcite . the action is given by @xmath20 where @xmath21\ , . \label{la5}\ ] ] the covariant derivative is defined as @xmath22 where @xmath23 and @xmath24 , with tr@xmath25=\delta_{ab}$ ] ( and similarly for @xmath26 and @xmath27 ) . for the value of the scalar mass we take @xmath28 that , by the ads / cft dictionary , corresponds to associate the scalar @xmath6 with a cft operator of dimension 3 such as @xmath29 . solving the equation of motion for @xmath6 we obtain @xmath30 where @xmath31 and @xmath32 can be written in terms of the value of @xmath33 at the boundaries @xmath34 where we have defined @xmath35 it can be shown that a nonzero @xmath36 corresponds to an explicit breaking of the chiral symmetry in the uv , while a nonzero @xmath32 corresponds to a spontaneous breaking of the chiral symmetry in the ir . therefore the value of @xmath32 is determined dynamically by minimizing the action . in order to get a nonzero value for @xmath32 in the chiral limit ( @xmath37 ) we add a potential for @xmath6 on the ir - boundary : @xmath38 an origin for this type of potentials can be found in string constructions @xcite . to determine the value of @xmath32 , or equivalently the value of @xmath39 , we must minimize the effective 4d action obtained after substituting eq . ( [ vev ] ) into the 5d action . for @xmath19 , this is given by @xmath40+v(\xi)\frac{l^4}{l_1 ^ 4}\right\}\ , , \label{energy}\ ] ] that is minimized for @xmath41 this 5d model depends on 5 parameters : for @xmath39 by means of eq . ( [ xi ] ) . in the following we will take @xmath42 and treat @xmath39 as a parameter . ] @xmath36 , @xmath43 , @xmath44 , @xmath39 and @xmath45 . the value of @xmath36 is related to the quark masses as we will see below . the values of @xmath43 , @xmath44 and @xmath39 were determined in ref . @xcite from the gauge sector of the theory . by using the qcd values for @xmath0 , @xmath46 and @xmath47 , it was found @xcite : @xmath48 our predictions will be given using the above values ( although in certain cases we will study the dependence of the predictions on @xmath39 ) . this leaves the scalar sector of the theory depending only on one parameter , @xmath45 . an estimate of its value can be obtained using naive dimensional analysis ( nda ) that gives @xmath49 . we define @xmath50 where @xmath51 corresponds to a real scalar and @xmath52 to a real pseudoscalar under parity . since we will be considering the chiral limit @xmath53 , we have @xmath54 and the symmetry breaking pattern u(3)@xmath2 u(3)@xmath55 u(3)@xmath56 . under su(3)@xmath56 we have that both @xmath51 and @xmath52 transform as @xmath57 . we will work in the unitary gauge . this corresponds to add the gauge fixing terms @xmath58 ^ 2\ , , \\ { \cal l}_{gf}^a&=-\frac{m_5 a}{2\xi_a}{\mathop{\rm tr}}\left[\partial_{\mu}a_{\mu}-\frac{\xi_a}{a}\partial_5 ( a a_5)-\xi_a \sqrt{2}a^2 v p\right]^2\ , , \end{aligned}\ ] ] where @xmath59 , and take the limit @xmath60 , _ i.e. _ @xmath61 the above equation will allow us to write @xmath52 as a function of @xmath62 in the 5d lagrangian . after integration by parts , the 5d quadratic terms for the scalar @xmath51 and the pseudoscalar @xmath62 are given by @xmath63s\big\ } , \\ & { \cal l}_{a_5}=-\frac{am_5}{2 } { \mathop{\rm tr}}\big\{a_5\big[\partial^2 { \cal d}+{\cal d}(2v^2a^2{\cal d})\big]a_5\big\ } , \end{aligned}\ ] ] where @xmath64 is a differential operator defined by @xmath65 the scalar and pseudoscalar field has also 4d boundary terms that , after using the 5d equation of motion for @xmath62 ( _ i.e. _ , @xmath66 ) and eq . ( [ xi ] ) , can be written as , as we will do later to perform a kaluza - klein decomposition of the sector . ] @xmath67\big{\|}_{l_0}^{l_1 } -a^4v(s)\big\|_{l_1}\nonumber\\ & + & m_5a^3{\mathop{\rm tr } } [ s]\partial_5 v\big\|_{l_0}\ , , \end{aligned}\ ] ] where @xmath68\big\|_{l_1}+{\cal o}(s^3 ) , \qquad m^2_s=\frac{4\lambda\xi^2}{l^2}-\frac{3m_5}{2l}+ { \cal o}(\widetilde m_q)\ , .\ ] ] to cancel the quadratic terms on the ir - boundary of eq . ( [ lbound ] ) we impose the conditions @xmath69s\big{\|}_{l_1}=0 , \qquad a_5\big{\|}_{l_1}=0\ , .\ ] ] the boundary conditions on the uv - boundary will be specified later . the interactions between scalars and pseudoscalars that we will be considering are @xmath70\ , , \\ \label{lp4 } { \cal l}_{a_5 ^ 4}&=&\frac{m_5}{96a^9 v^6}{\mathop{\rm tr}}\big[\big(\partial_5(aa_5)\overleftrightarrow{\partial_{\mu}}\partial_5(aa_5)\big)^2\big]\ , .\end{aligned}\ ] ] the @xmath71 interaction is absent . this is a consequence of the u(3)@xmath56 invariance and the fact that only dimension - four operators are considered in eq . ( [ la5 ] ) . this interaction , however , could be induced by higher - dimensional operators or loop effects . with the above lagrangian for the scalar and pseudoscalar sector we can calculate any relevant physical quantity . we will be considering two approximations . first , we will be working at the tree - level . according to eq . ( [ fit ] ) this corresponds to work in the large-@xmath0 limit . since loop effects are expected to be of order @xmath72 , our predictions for qcd quantities will have a @xmath73 uncertainty . second , we will take the chiral limit @xmath74 . for the pseudoscalar sector this limit will be taken in the following way . we will first perform the calculations with @xmath75 and fixed @xmath76 ( this is equivalent to @xmath77 and @xmath78 ) ) . next we will take the limit @xmath19 . this procedure simplifies the calculations and avoids singularities at @xmath18 . in this section we will calculate the scalar and pseudoscalar two - point correlator . in qcd these are defined as @xmath79 where @xmath80 and @xmath81 . the correlators @xmath82 can be obtained from the generating functional @xmath83 according to @xmath84 where @xmath85 and @xmath86 are the scalar and pseudoscalar external sources coupled to qcd : @xmath87+h.c . , \quad \phi = m_q+s+ip_s\ , .\ ] ] the ads / cft correspondence tells us @xcite that @xmath83 is obtained in the 5d theory by integrating out the bulk fields restricted to a given uv - boundary value . these boundary values play the role of the external sources coupled to qcd . in particular , for the 5d scalar field we have @xmath88 where the constant @xmath89 will be determined by matching with the qcd correlators in the uv as we will see later . up to the quadratic order in the fields , eq . ( [ uvbc ] ) leads to @xmath90 let us calculate @xmath91 at the quadratic level for @xmath51 and @xmath62 . by solving the equations of motion from eq . ( [ l2 ] ) with the boundary conditions of eqs . ( [ irbc ] ) and ( [ uvbcb ] ) , and substituting the solution back into the action , we get ( in momentum space ) and the longitudinal part of @xmath92 that we are not writing . ] @xmath93+\frac{1}{2}\pi_p(p^2){\mathop{\rm tr}}[p_s^2 ] + \gamma_s{\mathop{\rm tr}}[s]\ , .\ ] ] for a ads@xmath12 space @xmath94 can be given analytically at the tree - level . we obtain @xmath95 \ , , \ ] ] where @xmath96 are bessel functions , @xmath97 is the euclidean momentum and @xmath98 is determined by the ir - boundary condition of eq . ( [ irbc ] ) : @xmath99 taking the limit @xmath19 we find @xmath100\ , .\ ] ] the divergent terms for @xmath19 can be absorbed in a bare mass and a bare kinetic term for @xmath85 . after this renormalization the correlator is finite . for large momentum @xmath101 , we find @xmath102 as expected from the conformal symmetry . matching with qcd in which at large momentum we have @xmath103 we obtain , using eq . ( [ fit ] ) , @xmath104 the next to leading terms in the large momentum expansion in eq . ( [ largeps ] ) are suppressed exponentially , contrary to qcd where one finds that the scalar correlator has power corrections . this is because we assumed , for simplicity , that the scalar had a potential only on the ir - boundary . in more realistic models such as those arising from string theories the scalar potential is present in the 5d bulk ( although peaked towards the ir ) . in these cases the scalar correlator has power corrections . also , if the 5d metric deviate in the ir from ads or if higher - dimensional operators are included in eq.([la5 ] ) , then power corrections can be present in @xmath94 . for small momentum @xmath105 can be approximated by @xmath106 + { \cal o}(p^2)\ , .\ ] ] the scalar correlator eq . ( [ pis ] ) can also be written as a sum over infinitely narrow resonances , similarly as in large-@xmath0 qcd : @xmath107 therefore the masses of the scalar resonances can be determined by finding the poles of eq . ( [ pis2 ] ) , _ i.e. _ , by the equation @xmath108 . in fig . [ figms ] we plot the value of the mass of the first and second scalar resonance as a function of @xmath45 for @xmath109 . the first resonance mass ranges from @xmath110 mev ( @xmath111 ) to @xmath112 mev ( @xmath113 ) . we compare this value with the masses of the @xmath114 states ( since these are the qcd scalars whose masses are not very sensitive to @xmath115 ) . we see that for a value of @xmath45 close to its nda estimate , @xmath116 , the mass of the first scalar resonance is closer to that of @xmath117 than to that of @xmath118 . nevertheless we must recall that we are working in the large-@xmath0 limit and then corrections can be as large as @xmath73 . consequently we can not discard to associate @xmath119 with @xmath118 . the scalar decay constants @xmath120 are determined by the residues of @xmath94 . we obtain @xmath121 for @xmath122 we obtain @xmath123 gev and @xmath124 mev , while for the second resonance we get @xmath125 mev and @xmath126 mev . using this result we can calculate the value of the coupling @xmath127 defined in ref . we obtain @xmath128 mev ( taking the value of @xmath129 from eq . ( [ b0 ] ) ) very close to the value used in ref . @xcite : @xmath130 mev . for @xmath109 . the dashed lines show the experimental values for the masses of the scalar resonances @xmath117 and @xmath118.__,width=321 ] to calculate the pseudoscalar correlator @xmath131 we must rely on numerical analysis . only for small and large momentum we are able to give analytical results . for large momentum @xmath101 we have @xmath132\ , \qquad { \rm where}\ \ \ c^p_6=-\frac{64}{5}\frac{3\tilde n_c\ , \xi^2}{l_1 ^ 6 } \ , .\ ] ] again the divergences can be cancelled by adding a proper mass and a kinetic term for the pseudoscalar @xmath86 on the uv - boundary . from eqs . ( [ largeps ] ) and ( [ largepp ] ) we can obtain the correlator @xmath133 at large momentum . it drops as @xmath134 . comparing with @xmath135 , we find @xmath136 @xcite in strong disagreement with qcd in which one has @xmath137 . this can be improved if , as we said , we consider more realistic theories where the scalar potential is present in the 5d bulk and therefore @xmath94 has power corrections . at low momentum before taking @xmath138 @xcite . ] and for @xmath139 we find @xmath140 where @xmath141 @xmath142 is the axial - vector correlator calculated in ref . @xcite . the first term of eq . ( [ smallpp ] ) shows a pole at @xmath143 as expected due to the presence of the massless pgb . by looking at the poles of @xmath131 we can find the pseudoscalar masses . the lowest mode is the massless pgb of the spontaneous chiral symmetry breaking . there is a nonet of pgbs but we must recall that the inclusion of the u(1)@xmath144-anomaly will give mass to the singlet @xcite . the mass of the first massive resonance is shown in fig . [ figmp ] as a function of @xmath39 . we see that its value is far from the mass of the @xmath145 state . nevertheless , we find that , for @xmath146 , @xmath147 is close to the mass of @xmath148 suggesting that this could be the state to be associated with our first massive pseudoscalar resonance . for this resonance we find a decay constant @xmath149 mev . . the shadow band shows the experimental value for @xmath148.__,width=321 ] finally , we calculate the linear term in eq . ( [ leff ] ) to be associated in qcd with the @xmath29 condensate : @xmath150 . we find @xmath151 to study the interactions it is convenient to perform a kaluza - klein ( kk ) decomposition of the 5d fields : @xmath152 we impose the following boundary conditions on the uv - boundary : @xmath153 that cancel the boundary terms of eq . ( [ lbound ] ) . the wave - functions of the kk - modes @xmath154 are given by @xmath155 \ \ \ \stackrel{l_0\to 0 } { \longrightarrow } \ \ \ \frac{z^2}{n_{s_n}l_1 ^ 2 } j_1(m_{s_n}z ) \ , , \label{wfv}\ ] ] where @xmath156 is a constant fixed by canonically normalizing the fields , @xmath157 . in fig . [ figkkmodes ] we plot the wave - functions of the first two kk - modes . scalar resonance , the pgb and the first massive pseudoscalar for @xmath109 and @xmath158.__,width=321 ] the equation that determines the wave - functions of the pseudoscalars can be obtained from eq . ( [ l2 ] ) . this is given by @xmath159 the lowest state , @xmath160 , is the pgb that in the limit @xmath19 is massless . its wave - function is given by @xmath161 \ , , \ ] ] where @xmath162 is determined by the condition @xmath163 . the wave - function of the massive modes must be obtained numerically from eq . ( [ kkp ] ) with the normalization condition @xmath164 . the wave - functions of @xmath165 and @xmath166 are shown in fig . [ figkkmodes ] . the couplings between the resonances are easily obtained by integrating the 5d interactions over @xmath167 with the corresponding wave - functions . the coupling of a scalar to two pgbs comes from eq . ( [ lspp ] ) . we obtain @xmath168\ , , \ ] ] where @xmath169 is given by @xmath170 ^ 2}{2a^3v^3}\ , .\ ] ] in fig . [ figgn ] we show the coupling of the first modes as a function of @xmath45 for @xmath171 . we find that @xmath169 becomes smaller as @xmath172 increases . this property is also present in the coupling between a vector resonance and two pgbs , and it is due to the oscillatory behaviour of the kk wave - functions . associating @xmath119 with @xmath117 , we find that @xmath173 mev for @xmath122 , and the prediction of the 5d model for the @xmath174 coupling is @xmath175 gev@xmath176 for @xmath109 . in the notation of ref . @xcite we find @xmath177 mev to be compared to the value @xmath178 mev given there . if the width of @xmath117 is dominated by the decay to @xmath179 we find @xmath180 unfortunately , the experimental value of the width of @xmath117 has a large uncertainty @xmath181 mev @xcite . scalar resonance to two pgbs as a function of @xmath45 for @xmath109 ( solid line ) and @xmath182 ( dashed line).__,width=321 ] by integrating the heavy scalar resonances we obtain the following four - pgb interaction @xmath183 -\frac{1}{3}{{\mathop{\rm tr}}}^2[(\partial_{\mu}\pi)^2]\right\ } \sum_n\frac{g_{n\pi\pi}^2}{p^2+m_{s_n}^2}\ , , \ ] ] from the scalar octet and @xmath184 \sum_n\frac{g_{n\pi\pi}^2}{p^2+m_{s_n}^2}\ , , \ ] ] from the scalar singlet . the sum over the kk - modes in eqs . ( [ l4pi8 ] ) and ( [ l4pi1 ] ) is dominated by the first resonance . at large momentum we find that the first resonance gives @xmath185 of the total contribution and this percentage rises to @xmath186 at zero momentum ( for @xmath122 ) . therefore , as in the vector case @xcite , we find that the scalar mediation of the four - pgb interaction is dominated by the exchange of the first resonance . four - pgb interactions can also arise from eq . ( [ lp4 ] ) . we find @xmath187\ , \ \ \text{where}\ \ \ \ \ g_{\pi^4}=\frac{1}{24m_5l^2}\int dz\ , \frac{[\partial_5(af^\pi)\big]^4}{a^9v^6}\ , .\ ] ] at high energies the four - pgb amplitude arising from eq . ( [ pion4 ] ) grows as @xmath188 . nevertheless , this bad energy behavior of the four - pgb amplitude is cured by the contribution arising from eqs . ( [ l4pi8 ] ) and ( [ l4pi1 ] ) that cancels the @xmath189 terms . this occurs thanks to the sum rule @xmath190 eq . ( [ sumrule1 ] ) is a property of any 5d model in which the breaking of the chiral symmetry is realized by the higgs mechanism . we can also calculate the coupling of the pgb to the source @xmath85 that defines the scalar form factor of the pgb . apart from a contact piece given by @xmath191\ , , \ ] ] this coupling is mediated by the octet and singlet scalar resonances that gives respectively @xmath192- \frac{1}{3}{{\mathop{\rm tr}}}[(\partial_{\mu}\pi)^2]{\mathop{\rm tr}}[s]\right\ } \sum_n\frac{g_{n\pi\pi}f_{s_n}m_{s_n}}{p^2+m_{s_n}^2 } \ , , \nonumber\\ { \cal l}^{(1)}_{\pi^2s}&= & \frac{1}{3}{{\mathop{\rm tr}}}[(\partial_{\mu}\pi)^2]{\mathop{\rm tr}}[s ] \sum_n\frac{g_{n\pi\pi}f_{s_n}m_{s_n}}{p^2+m_{s_n}^2}\ , .\end{aligned}\ ] ] the scalar form factor of the pgb is then given by ( normalized to unity at zero momentum ) @xmath193 at low momentum the sum in eq . ( [ fspi ] ) is dominated by the first resonance that gives @xmath194 of the total contribution ( for @xmath122 ) . at large momentum we find that the form factor goes as @xmath195 , as expected from the conformal symmetry @xcite . the cancellation of the constant term in @xmath196 occurs due to the sum rule @xmath197 this sum rule is fulfilled in any 5d model whose metric approaches to ads@xmath12 for @xmath198 ( conformal theories in the uv ) . in eq . ( [ sumrule2 ] ) we find that the first two resonances give a similar contribution , while the contributions of the heavier resonances tend to cancel out . therefore we see that @xmath196 is very well approximated by the exchange of only the first two resonances . at energies below the massive resonances our 5d model is described by the qcd chiral lagrangian . in this section we calculate the ( pseudo)scalar contributions to the coefficients of the chiral lagrangian and compare these results with the qcd values . up to @xmath199 , the chiral lagrangian for the octet of pgb , @xmath200 , is given by @xcite @xmath201\ , , \label{l2}\ ] ] where @xmath202 and @xmath203 the prediction of our model for @xmath204 is given in eq . ( [ fpi ] ) . it gives @xmath205 for the prediction of @xmath129 we can use eq . ( [ cond ] ) : @xmath206 that leads to @xmath207 notice that @xmath208 as it should be , since the first term of eq . ( [ smallpp ] ) can also be deduced by integrating out the pgb at tree - level in the chiral lagrangian . the relation @xmath208 also leads to the right matching of eq . ( [ direct ] ) with the chiral lagrangian . the value of the quark masses @xmath115 is related to the vev of @xmath6 on the uv - boundary . using eqs . ( [ mqdef ] ) , ( [ uvbc ] ) and ( [ alpha ] ) we obtain was not calculated . ] @xmath209 from the chiral lagrangian we have @xmath210\ , , \ ] ] that for @xmath211 mev and @xmath212 mev gives @xmath213 the value of the quark masses in eq . ( [ masses ] ) are scale independent . this is because we took @xmath28 that corresponds , by the ads / cft dictionary , to fix the dimension of @xmath115 to be exactly one . in qcd however the quark masses evolve with the energy scale @xmath214 . to minimize this discrepancy we must compare our predictions with the experimental values of the quark masses taken at the lowest energy scale ( @xmath215 gev ) . from ref . @xcite we have @xmath216 mev and @xmath217 mev at @xmath218 gev in good agreement with eq . ( [ masses ] ) . at @xmath219 the chiral lagrangian is given by @xcite @xmath220 + l_2 \,{\mathop{\rm tr}}\big [ d_\mu u^\dagger d_\nu u\big ] { \mathop{\rm tr}}\big [ d^\mu u^\dagger d^\nu u\big ] + l_3 \,{\mathop{\rm tr}}\big [ d_\mu u^\dagger d^\mu u d_\nu u^\dagger d^\nu u\big]\ , \nonumber\\ & + & l_4 \,{\mathop{\rm tr}}\big [ d_\mu u^\dagger d^\mu u\big ] { \mathop{\rm tr}}\big [ u^\dagger\chi + \chi^\dagger u \big ] + l_5 \,{\mathop{\rm tr}}\big [ d_\mu u^\dagger d^\mu u \left ( u^\dagger\chi + \chi^\dagger u \right)\big]\ , \nonumber\\ & + & l_6 \,{\rm tr}^2\big [ u^\dagger\chi + \chi^\dagger u \big ] + l_7 \,{\rm tr}^2\big [ u^\dagger\chi - \chi^\dagger u \big ] + l_8 \,{\mathop{\rm tr}}\big[\chi^\dagger u \chi^\dagger u + u^\dagger\chi u^\dagger\chi\big ] \nonumber \\ & - & i l_9 \,{\mathop{\rm tr}}\big [ f_r^{\mu\nu } d_\mu u d_\nu u^\dagger + f_l^{\mu\nu } d_\mu u^\dagger d_\nu u\big ] + l_{10 } \,{\mathop{\rm tr}}\big [ u^\dagger f_r^{\mu\nu } u f_{l\mu\nu } \big ] \ , . \label{l4}\end{aligned}\ ] ] at tree - level , the ( pseudo)scalar resonances only contribute will not be studied here since it arises from integrating out the singlet pgb that becomes massive when the u(1)@xmath144 anomaly is considered . ] to @xmath221 . the contributions to the coefficients @xmath44 and @xmath222 coming from the octet and singlet scalar can be read from eqs . ( [ l4pi8 ] ) and ( [ l4pi1 ] ) . we obtain @xmath223 the octet and singlet contribution to the coefficient @xmath44 cancels out , as expected from large-@xmath0 @xcite , and only @xmath222 gets a nonzero scalar contribution . for @xmath122 and @xmath109 ( @xmath224 ) we obtain @xmath225 ( @xmath226 ) . adding the vector contribution to @xmath222 calculated in ref . @xcite we get @xmath227 ( @xmath228 ) to be compared with the experimental value @xcite @xmath229 . the scalar contribution to @xmath230 and @xmath231 can be obtained from eq . ( [ sff8 ] ) : @xmath232 as expected from large-@xmath0 , the total contribution to @xmath233 is zero . the value of @xmath234 is shown in fig . [ figl8 ] as a function of @xmath235 for @xmath171 . for @xmath236 gev we obtain @xmath237 in good agreement with experiments . @xmath234 can also be calculated from the axial - vector correlator @xcite : @xmath238 for @xmath139 with @xmath239 fixed , was given for @xmath240 and therefore the last term of eq . ( [ l5ap ] ) was not present . this last term arises due to the @xmath39 dependence on @xmath36 - see eq . ( [ xi ] ) . also a factor @xmath241 was missing in eq . ( 67 ) of ref . @xcite and therefore the prediction of @xmath234 given there was a factor @xmath242 larger . ] we obtain @xmath243 + \frac{f^4_\pi l_1 ^ 4}{192\lambda\xi^4 } \simeq 1.2\cdot 10^{-3}\left[1 - 0.23\left(\frac{4}{\xi}\right)^{\frac{2}{3 } } + 0.09\left(\frac{10^{-3}}{\lambda}\right ) \left(\frac{4 } { \xi}\right)^{\frac{8}{3 } } \right]\ , .\ ] ] finally , the coefficient @xmath244 can be computed from the correlators @xmath82 . we have @xmath245\right\|_{p^2=0 } \ , , \qquad l_8^{(1)}=0\ , .\end{aligned}\ ] ] then @xmath246 , as expected from large-@xmath0 . using eqs . ( [ smallps ] ) , ( [ smallpp ] ) and ( [ b0 ] ) in the above equation , we obtain @xmath247 \simeq 8\cdot 10^{-4}\left[1 -0.27\left(\frac{4}{\xi}\right)^{\frac{2}{3 } } + 0.11\left(\frac{10^{-3}}{\lambda}\right ) \left(\frac{4 } { \xi}\right)^{\frac{8}{3 } } \right]\ , .\ ] ] notice that this expression is only valid for @xmath139 with @xmath239 fixed . in fig . [ figl8 ] we show the exact value of @xmath248 as a function of @xmath235 . for @xmath123 gev and @xmath109 we obtain @xmath249 again in good agreement with the experimental value . from fig . [ figl8 ] one can see that small values of @xmath235 are preferred . the coefficient @xmath248 can also be written as @xmath250 \ , , \ ] ] that shows that in the limit where the chiral symmetry is restored , @xmath251 and @xmath252 , only the first term remains . for @xmath146 we find that the first term still dominates ( it gives @xmath253 of the total contribution for @xmath122 ) since the other resonances , being so heavy , are not very sensitive to chiral symmetry breaking . and @xmath248 as a function of @xmath235 . the horizontal line corresponds to the experimental value with the error bands @xcite.__,width=321 ] and @xmath248 as a function of @xmath235 . the horizontal line corresponds to the experimental value with the error bands @xcite.__,width=321 ] we have analyzed the scalar and pseudoscalar sector of a five - dimensional model proposed to study mesons in qcd . we have calculated the scalar and pseudoscalar two - point correlator and we have obtained the mass spectrum and interactions . this has allowed us to determine the ( pseudo)scalar contribution to the scalar form factor of the pgb as well as the contribution to the @xmath1 coefficients of the chiral lagrangian . we have also found two interesting sum rules for the scalar couplings and masses of the resonances that are fulfilled generically in ads@xmath12 models . comparing with the experimental data , we have found a good agreement for the @xmath1 predictions ( see fig . [ figl8 ] ) and the quark masses . for the first massive pseudoscalar resonance we have obtained a mass around @xmath254 mev , quite different from the mass of the lowest qcd pseudoscalar resonance @xmath145 . this has suggested us to associate this state to @xmath148 . we have also given predictions for the scalar couplings and decay constants but the absence of clean experimental data has not allowed us to compare them with qcd . previous approaches to calculate the scalar and pseudoscalar spectrum and/or determine their contribution to @xmath1 can be found in refs . in particular , the analysis of refs . @xcite has certain similarity with ours . @xcite work in the large-@xmath0 limit where qcd is described as a theory of infinite hadron resonances . these sets of infinite hadrons , however , are approximated in refs . @xcite by taking only the lowest modes , and their masses and couplings are determined by demanding a good high - energy behaviour of the correlators and form factors . in our approach we have shown that the correlators and form factors have the correct high - energy behaviour since this is dictated by the conformal symmetry . we have also found that , in certain cases , it can be a good approximation to take only the lowest resonance . therefore in these cases our approach and that of refs . @xcite give similar results . nevertheless , we have showed that the single - resonance approximation is not always justified ( for example in eq . ( [ sumrule2 ] ) ) and this approximation can lead to large errors in the determination of the scalar parameters . the analysis carried out here can be extended to study three - point or four - point correlation functions or to incorporate the effects of @xmath255 either in the mass spectrum or in the interactions . also the effects of higher - dimensional operators or departures from ads@xmath12 in the ir - boundary can be studied . these effects are important to study the power corrections in the correlators . we leave this analysis for the future . we would like to thank rafel escribano , matthias jamin , santi peris and eduard de rafael for very useful discussions . this work was partly supported by the mcyt and feder research project fpa2002 - 00748 and dursi research project 2001-sgr-00188 . the work of ld was supported by the spanish education office ( mecd ) under an fpu scholarship . for certain realizations of string theories describing flavor see : a. karch and e. katz , jhep * 0206 * ( 2002 ) 043 ; m. kruczenski , d. mateos , r. c. myers and d. j. winters , jhep * 0307 * , 049 ( 2003 ) ; t. sakai and j. sonnenschein , jhep * 0309 * , 047 ( 2003 ) ; c. nunez , a. paredes and a. v. ramallo , jhep * 0312 * ( 2003 ) 024 ; j. babington , j. erdmenger , n. j. evans , z. guralnik and i. kirsch , phys . d * 69 * ( 2004 ) 066007 ; n. j. evans and j. p. shock , phys . d * 70 * ( 2004 ) 046002 ; k. ghoroku and m. yahiro , phys . b * 604 * , 235 ( 2004 ) ; k. suzuki , arxiv : hep - th/0411076 ; s. kuperstein , jhep * 0503 * ( 2005 ) 014 ; d. bak and h. u. yee , phys . d * 71 * , 046003 ( 2005 ) ; a. paredes and p. talavera , nucl . b * 713 * , 438 ( 2005 ) ; i. kirsch and d. vaman , phys . d * 72 * ( 2005 ) 026007 ; r. apreda , j. erdmenger and n. evans , arxiv : hep - th/0509219 .	we study the scalar and pseudoscalar sector in a five - dimenional model describing chiral symmetry breaking . we calculate the scalar and pseudoscalar two - point correlator , the mass spectrum and interactions . we also obtain the scalar and pseudoscalar contributions to the coefficients of the chiral lagrangian and determine the scalar form factor of the pseudo - goldstone bosons . most quantities show a good agreement with qcd . uab - ft-588 + october 2005 + + +
kohn - sham ( ks ) density functional theory@xcite ( dft ) is nowadays one of the most popular methods for electronic structure calculations both in chemistry and solid - state physics , thanks to its combination of low computational cost and reasonable performances . the accuracy of a ks - dft result is limited by the approximate nature of the exchange - correlation energy density functional @xmath0 $ ] . simple approximations ( local - density approximation and generalized gradient corrections ) for @xmath0 $ ] provide practical estimates of thermodynamical , structural and spectroscopic properties of atoms , molecules and solids . however , with the current approximations , ks - dft is still lacking in several aspects , in particular it fails to handle near - degeneracy correlation effects ( rearrangement of electrons within partially filled shells ) and to recover long - range van der waals interaction energies . the inaccuracy of ks - dft stems from our lack of knowledge of @xmath0 $ ] , and much effort is put nowadays in finding new approximations to this term ( for recent reviews , see , e.g. , refs . ) . a popular trend in the development of new ks @xmath0 $ ] is the use of the exact exchange functional @xmath1 $ ] ( in terms of the ks orbitals ) , and thus the search for an approximate , compatible , correlation functional @xmath2 $ ] . an exact expression for @xmath2 $ ] is the coupling - constant integral@xcite e_c[n]=_0^_phys d_0^dr_12 4r_12 ^ 2f_c^(r_12 ) , [ eq_ecfromf ] where the interaction between the electrons is adiabatically turned on from @xmath3 to the coulomb repulsion @xmath4 by varying a real parameter @xmath5 ( typical examples are @xmath6 , with @xmath7 , or @xmath8 , with @xmath9 ) . the one - electron density @xmath10 is ( ideally ) kept independent of @xmath5 and equal to the one of the physical system by means of a suitable external potential @xmath11 . in eq . ( [ eq_ecfromf ] ) the correlation part of the spherically and system - averaged pair density ( intracule density ) @xmath12 is defined as follows . for each @xmath5 , take the square of the many - electron wavefunction @xmath13 ground - state of the hamiltonian @xmath14 , h^=-_i=1^n+_ij=1^n w^(\|_i-_j\|)+_i=1^n v^(_i ) , and integrate it over all variables but the scalar electron - electron distance @xmath15 , f^(r_12 ) = _ _ 1 ... _n @xmath16 . the correlation part @xmath12 is then defined as @xmath17 , where the intracule density of the ks system is @xmath18 ( and yields the hartree plus the exchange energy ) . the traditional dft approach to the construction of approximate @xmath2 $ ] is based on the idea of universality . for example , the familiar local - density approximation ( lda ) consists in transfering , in each point of space , the pair density from the uniform electron gas to obtain an approximation for @xmath12 in eq . ( [ eq_ecfromf ] ) . in a couple of recent papers , @xcite we have started to explore a different way of constructing @xmath2 $ ] , based on an `` average pair density functional theory '' ( apdft ) , which was inspired by the seminal work of overhauser @xcite and its subsequent extensions . @xcite in this approach , we solve a set of radial ( one - dimensional ) schrdinger - like equations that give , in principle , the exact @xmath19 along the dft adiabatic connection . in practice , this formalism contains an unknown effective electron - electron interaction that needs to be approximated . the apdft equations must be solved for each system , and combined self - consistently with the ks equations.@xcite preliminary applications of this approach , combined with a simple approximation @xcite for the effective electron - electron interaction that enters in the formalism , gave accurate intracule densities @xmath20 and correlation energies @xmath2 $ ] for the he isoelectronic series . @xcite katriel _ et al . _ @xcite have recently tested most of the currently available correlation energy functionals in the high - density ( weakly - correlated ) limit of the he and of the hooke s atom isoelectronic series , finding that , while several functionals are accurate for the he sequence , none is satisfactory for the hooke s atom series . motivated by their findings , in this work we compute the correlation energy and the intracule density in the high - density limit of the two series via the apdft approach , @xcite finding accurate results in both cases . the paper is organized as follows . in the next sec . [ sec_wcl ] we recall the basic equations that define the high - density limit of the he and hooke s atom sequences , to which we apply , in sec . [ sec_eff ] and [ sec_ec ] , the formalism of refs . to compute the intracule density and the correlation energy . in sec . [ sec_lda ] we also analyze the failure of lda in this limit from the point of view of @xmath20 . the last sec . [ sec_conc ] is devoted to conclusions . the two hamiltonians analyzed in this paper read @xmath21 we are interested in the high - density ( weakly - correlated ) limit , which corresponds to @xmath22 and @xmath23 . by switching to scaled coordinates @xmath24 , with @xmath25 ( he series ) and @xmath26 ( hooke s series ) , both hamiltonians have the form h = ( --+(s_1)+(s_2)+ ) ( _ 0+_1 ) , [ eq_hgeneral ] where @xmath27 for the he series , and @xmath28 for the hooke s atom series . we thus study pertubatively the system described by @xmath29 . the order zero of the one - electron density @xmath30 and of the intracule density @xmath20 , in scaled units , is simply @xmath31 these functions are correctly normalized , so that if we switch back to coordinates @xmath32 we have @xmath33 , etc . the first - order correction to the scaled density , @xmath34 is given by n^(1)(s ) = 2 n^(0)(s ) ( s ) , [ eq_densord1 ] where @xcite ( s ) = ---++ s+(-2s)-(s ) , [ eq_chi ] for the he isoelectronic series , with @xmath35 , ( -x)=-_x^ dt , and @xcite @xmath36,\end{aligned}\ ] ] for the hooke s atom isoelectronic series . by definition , the kohn - sham hamiltonian describes a non - interacting system that has the same density of the physical , interacting , system . thus , the first - order change in the electron density of eq . ( [ eq_densord1 ] ) corresponds to a first - order change in the ks system . therefore , we write the scaled intracule @xmath37 up to orders @xmath38 as f(s_12)=f^(0)(s_12)+ + o(^2 ) , [ eq_expf ] where we have separated the first - order correction into a kohn - sham part and a correlation part . the ks part @xmath39 is entirely determined by the first - order density @xmath40 of eq . ( [ eq_densord1 ] ) , f_ks^(1)(s_12)=dn^(0)(+_12 ) n^(0)(s)(s ) , [ eq_fks1 ] and is reported in appendix [ app_fks1 ] in analytic form for the he isoelectronic series , while is obtained numerically for the case of the hooke s series . the total first - order intracule @xmath41 is known analytically in the case of the hooke s series,@xcite f^(1)(s_12 ) = . we are interested in calculating @xmath42 and the corresponding second - order correlation energy @xmath43 with the method of refs . , in which the intracule density @xmath20 of the physical system is obtained from a set of effective equations , which for two - electron systems reduce to ( r_12)= ( r_12 ) , [ eq_eff ] with @xmath44 . equation ( [ eq_eff ] ) can be derived by considering@xcite a set of hamiltonians characterized by a real parameter @xmath45 , h^= -_i=1^n+_ij=1^n w^(\|_i-_j\|)+_i=1^n v_ne(_i ) , f^(r_12)=f(r_12)[eq_adiaapdft ] that describe a set of systems in which the external potential is turned off as @xmath46 , and the intracule density is kept fixed , equal to the one of the physical system , by means of a suitable electron - electron interaction @xmath47 . in the case @xmath48 , when @xmath49 we have a translationally - invariant system ( the center - of - mass degree of freedom is described by a plane wave ) of two fermions in a relative bound state ( similar to the case of positronium , but with a different interaction ) . this relative bound state is such that the square of the wavefunction for the relative coordinate @xmath50 is equal to @xmath20 of the starting physical system , and is thus described by eq . ( [ eq_eff]).@xcite for more than two electrons , in the case of a confined system ( atom , molecule ) , the limit @xmath46 in eq . ( [ eq_adiaapdft ] ) describes a cluster of fermions , and eq . ( [ eq_eff ] ) becomes an approximation@xcite for the internal degrees of freedom of the cluster . here we focus on the high - density limit of the hamiltonians of eq . ( [ eq_hgeneral ] ) and we thus stick to the case @xmath48 . in general , the effective electron - electron interaction @xmath51 in eq . ( [ eq_eff ] ) is not known , and must be approximated . in the case of the he series , we have found@xcite that a simple approximation based on the original idea of overhauser@xcite gives very accurate results for @xmath52 . in what follows we analyze the performance of the same approximation in the very @xmath22 limit , and we extend our study to the @xmath23 limit of the hooke s atom series . of course , in the special case of the hooke s series , the hamiltonian ( [ eq_hgeneral ] ) is exactly separable into center - of - mass and relative coordinates , so that the exact @xmath51 is directly available . however , the point here is to check whether the same approximate @xmath51 that accurately describes the he series is capable to yield also good results for the hooke s series , since this seems to be not the case for the currently available correlation energy functionals.@xcite the construction of an approximation for the e - e effective potential @xmath53 starts with the decomposition @xcite w_eff(r_12)=w_eff^ks(r_12)+w_eff^c(r_12 ) , [ eq_wdecomp ] where @xmath54 is the potential that generates the kohn - sham @xmath55 via eq . ( [ eq_eff ] ) , and @xmath56 is a correlation potential that needs to be approximated . in the usual dft language , eq . ( [ eq_wdecomp ] ) implies that we are treating exchange exaclty . in scaled units @xmath57 , using standard perturbation theory we obtain the equation for the first - order contribution to @xmath58 [ see eq . ( [ eq_expf ] ) ] , that separates into the kohn - sham and the correlation parts : @xmath59 \psi^{(1)}_{\rm ks } = \left[\epsilon_{\rm ks}^{(1 ) } -w_{\rm eff}^{\rm ks\,(1)}\right ] \psi^{(0 ) } \label{eq_ks1 } \\ \left[-\nabla^2+w_{\rm eff}^{\rm ks\,(0)}-\epsilon^{(0)}\right ] \psi^{(1)}_c = \left[\epsilon_{c}^{(1 ) } -w_{\rm eff}^{c\,(1)}\right ] \psi^{(0 ) } , \label{eq_corr1}\end{aligned}\ ] ] where @xmath60 , @xmath61 , @xmath62 , and @xmath63 in eq . ( [ eq_ks1 ] ) , @xmath39 is exatly known for both series , so that we can also obtain @xmath64 by inversion . we thus concentrate on the correlation part , since we want to test approximations for @xmath65 . defining @xmath66 and @xmath67 , we have u_c = u_0 . following the method of refs . we look for a solution of the kind @xmath68 . the function @xmath69 is then given by y(x)=_0^x_0^x u_0 ^ 2(x ) [ w_eff^c(1)(x)-_c^(1)]dx [ eq_y ] the constant @xmath70 is fixed by requiring the proper normalization , _ 0^f_c^(1)(x ) x^2 dx=0 _ 0^ y(x ) u_0 ^ 2(x)dx=0 . the other integration constant has been fixed in eq . ( [ eq_y ] ) by setting equal to zero an unphysical term @xmath71 that would make @xmath72 diverge for large @xmath73 . , divided by @xmath74 , as a function of the scaled variable @xmath75 for the he isoelectronic series . the `` exact '' results are obtained from the accurate wavefunctions of ref . . approximate results at finite @xmath76 using the `` overhauser model '' are taken from ref . . the @xmath77 result corresponds to eq . ( [ eq_y ] ) with the potential of eq . ( [ eq_ovscaled ] ) . ] in refs . an approximation for @xmath78 was built as an average `` overhauser - type '' potential,@xcite w_eff^c(r_12)(+ - ) ( _ s - r_12 ) , [ eq_ov ] where @xmath79 is the heaviside step function and @xmath80 is related to the average density , or , better to the dimension of the system . for two - electron atoms it was simply estimated as@xcite _ s=()^-1/3 , [ eq_avrs ] where = dn()^2 . [ eq_avn ] the idea beyond this approximation is the following . the e - e correlation potential @xmath56 changes the kohn - sham @xmath58 into the physical one , and must thus keep the information on the one - electron density ( which is the same in the two systems ) while turning on the e - e interaction @xmath81 . in eqs . ( [ eq_ov])-([eq_avn ] ) this information is approximately kept via the average density @xmath82 . in scaled units , the overhauser potential to first order in @xmath38 , to be used in eq . ( [ eq_corr1 ] ) , becomes w_eff^c(1)(s_12)(+- ) ( _ s - s_12 ) , [ eq_ovscaled ] where , if we adopt the prescription of eqs . ( [ eq_avrs])-([eq_avn ] ) , @xmath83 for the he series and @xmath84 for the hooke s atom series . equation ( [ eq_y ] ) with the potential of eq . ( [ eq_ovscaled ] ) can be evaluated analytically as a function of @xmath85 and @xmath86 for both series , although the final expressions are cumbersome and will not be reported here . the resulting @xmath42 for the he series is shown in fig . [ fig_fcall ] , together with the corresponding scaled quantity , @xmath87 , for some finite @xmath76 . [ since @xmath88 , and @xmath89 , the quantity to be compared with @xmath90 is @xmath87 . ] for finite @xmath76 we show both the `` exact '' result @xcite and the approximate result @xcite obtained with the overhauser - type potential of eqs . ( [ eq_ov])-([eq_avn ] ) . we see that the @xmath76 dependence of the short - range part of @xmath91 is very well captured by this simple approximation . figure [ fig_fcall ] also suggests that the @xmath22 limit of the short - range part of @xmath91 is well described by this approach . in fig . [ fig_fc1hooke ] we show the result for @xmath92 in the case of the hooke s series from the overhauser potential compared to the exact one , finding very accurate agreement . [ see eq . ( [ eq_expf ] ) ] , for the hooke s series . the exact values are compared with the results from the overhauser - type approximation of eq . ( [ eq_ovscaled ] ) . ] [ see eq . ( [ eq_expf ] ) ] : the kohn - sham part and the correlation part . ] [ see eq . ( [ eq_expf ] ) ] : the kohn - sham part and the correlation part . ] the ks and the correlation components of @xmath93 are shown in fig . [ fig_ord1 ] for both series . we see that in the case of the he series the ks and the correlation parts have roughly the same depth , while in the case of the hooke s series the correlation part is much deeper than the ks one . this is due to the fact that the ks part gives the change in the e - e distance probability distribution only due to the first - order change in the one - electron density . in the case of the hooke s series the first - order change in the density is much smaller , since the harmonic confining external potential is stronger than the coulombic one . indeed , the function @xmath94 of eq . ( [ eq_densord1 ] ) in the case of the he series is about twice the one for the hooke s atom series . the apd @xmath95 gives the correlation contribution to second order to the expectation @xmath96 of the coulomb electron - electron repulsion operator , @xmath97 , v_ee= , where @xmath98 , and v_ee_c^(2 ) = _ 0 ^ 4 s_12_c^(1)(s_12 ) ds_12 . our @xmath92 from the overhauser potential give @xmath99 ha for the he sequence ( to be compared with the exact @xcite value , @xmath100 ha ) , and @xmath101 ha for the hooke s series ( to be compared with the exact @xcite value , @xmath102 ha ) . the error is thus 9 mh for the he series and 4 mh for the hooke s series . the correlation energy can then be otbained via the adiabatic connection formula of eq . ( [ eq_ecfromf ] ) , which for @xmath43 reads e_c^(2)=_0^_phys d_0^ds_12 4s_12 ^ 2_c^(1)(s_12 ) , [ eq_adiabatic ] where @xmath103 is the first - order correlated part of @xmath104 for the system with interaction @xmath105 . if we were able to calculate the exact @xmath103 for any @xmath106 , the resulting @xmath43 from eq . ( [ eq_adiabatic ] ) would be independent of the choice of @xmath106 . however , when we deal with approximations , we can obtain better results with some choices rather than others . as in ref . , we build an overhauser - type potential along the adiabatic connection as w_eff^c,(s_12;_s)=w^(s_12 ) - _ w^(\|- _ 12\|)d , [ eq_vefflambda ] where , in scaled units , if we stick with the choice of eqs . ( [ eq_avrs])-([eq_avn ] ) , @xmath107 for the he series and @xmath108 for the hooke s series . the idea behind eq . ( [ eq_vefflambda ] ) is that the average density @xmath82 ( and thus the average @xmath86 ) is kept fixed to mimic the fact that the one - electron density does not change along the adiabatic connection while we turn on the e - e interaction . if we set @xmath109 , eq . ( [ eq_vefflambda ] ) simply gives the overhauser potential of eq . ( [ eq_ovscaled ] ) with a multiplying factor @xmath5 in front . from eq . ( [ eq_y ] ) , we see that this corresponds to @xmath110 , as in the exact case . i.e. , the simple approximation of eq . ( [ eq_vefflambda ] ) has the correct scaling behavior in the @xmath111 limit . our result for @xmath43 with the linear adiabatic connection thus gives an error of 4.5 mh for the he series and 2 mh for the hooke s series . a choice for @xmath106 that separates short- and long - range effects is the `` erf '' adiabatic connection @xcite , @xmath112 , for which eq . ( [ eq_adiabatic ] ) becomes e_c^(2)=_0^d_0^ds_12 4s_12 ^ 2_c^(1)(s_12 ) e^-^2s_12 ^ 2 . [ eq_erfconn ] the overhauser - type potential corresponding to this interaction is reported in the appendix of ref . . for the he isoelectronic series with @xmath52 , the overhauser - type approximation combined with the `` erf '' adiabatic connection gives@xcite correlation energies with errors within 4 mh , better than the linear adiabatic connection that gives errors within 10 mh . in the weakly - correlated limit , instead , we obtained , via eq . ( [ eq_erfconn ] ) , @xmath113 ha for the he series and @xmath114 ha for the hooke s series . the errors with respect to the exact values , 6 mh and 4 mh , respectively , are thus slightly worse than those obtained with the linear adiabatic connection . , of order @xmath115 for the he series [ see eq . ( [ eq_expf ] ) ] : the present calculation is compared with the lda approximation ( panel a ) . panel b shows the same quantities multiplied by @xmath116 : the integral under each curve gives the correlation part of the second order contribution to the expectation value @xmath96 , which diverges in the case of lda . ] , of order @xmath115 for the he series [ see eq . ( [ eq_expf ] ) ] : the present calculation is compared with the lda approximation ( panel a ) . panel b shows the same quantities multiplied by @xmath116 : the integral under each curve gives the correlation part of the second order contribution to the expectation value @xmath96 , which diverges in the case of lda . ] , of order @xmath26 for the hooke s atom series [ see eq . ( [ eq_expf ] ) ] : the exact result is compared with the lda approximation ( panel a ) . panel b shows the same quantities multiplied by @xmath116 : the integral under each curve gives the correlation part of the second order contribution to the expectation value @xmath96 , which diverges in the case of lda . ] , of order @xmath26 for the hooke s atom series [ see eq . ( [ eq_expf ] ) ] : the exact result is compared with the lda approximation ( panel a ) . panel b shows the same quantities multiplied by @xmath116 : the integral under each curve gives the correlation part of the second order contribution to the expectation value @xmath96 , which diverges in the case of lda . ] as a further element of comparison , we also computed the first - order @xmath117 within the local - density approximation ( lda ) , f^(1)lda_c(s_12)=_0 g_c(_f ( ) s_12 ; _ s ( ) ) d , [ eq_lda ] where @xmath118 is the pair - correlation function of the uniform electron gas @xcite of density @xmath119 , and _ f ( ) = [ 3 ^ 2 n^(0)()]^1/3 , _ s()=^-1/3 . [ eq_rsforalpha ] with these definitions , the density parameter @xmath120 of the uniform electron gas is locally proportional to @xmath38 . we have numerically evaluated the right - hand - side of eq . ( [ eq_lda ] ) at smaller and smaller @xmath38 ( i.e. , at larger and larger @xmath76 and @xmath121 ) , for @xmath122 . as @xmath38 decreases , the results tend to a well defined curve , shown in figs . [ fig_lda_he ] and [ fig_lda_hooke ] , together with the result from the overhauser model ( he series ) and the exact result ( hooke s series ) . since , as shown by eq . ( [ eq_lda ] ) , the @xmath111 limit corresponds to the @xmath123 limit of the uniform electron gas pair - correlation function @xmath124 , to better understand the lda result for @xmath91 we now analyze more in detail the high - density behavior of @xmath124 . this analysis extends and completes the one done in ref . . when @xmath123 , the short - range part of @xmath124 scales as g_c(x , r_s0)= r_s g_c^(1)(x)+o(r_s^2 r_s ) , x = r_12/r_s , [ eq_gchd ] where the function @xmath125 does not depend explictily on @xmath120 and has been computed by rassolov _ et al . _ it is accurately recovered by the model @xmath124 of ref . that we have used in the evaluation of eq . ( [ eq_lda ] ) . the scaled variable @xmath73 is locally proportional to the scaled variable @xmath85 [ see eq . ( [ eq_rsforalpha ] ) ] . equation ( [ eq_gchd ] ) thus shows that the short - range part ( corresponding to values of the scaled variable @xmath73 not too large ) of @xmath124 in the @xmath123 limit has a scaling similar to the one of the he and hooke s series in the @xmath111 limit . this is also reflected by a good performance of lda for @xmath126 , as shown by figs . [ fig_lda_he ] and [ fig_lda_hooke ] . however , the high - density electron gas is an extended system with important long - range correlations that are not present in finite systems like atoms and molecules . in fact , the scaling of eq . ( [ eq_gchd ] ) is not valid when @xmath127 : it has been shown that the long - range part of @xmath124 scales as @xcite g_c(x1,r_s)r_s^2 h(v ) , where @xmath128 is another scaled variable , @xmath129 , which is thus locally proportional to @xmath130 . the function @xmath131 has the following asymptotic behaviors : h(v1 ) v^-2 , h(v1 ) v^-4 , which are also correctly included in the model @xmath124 of ref . . when @xmath123 ( i.e. , @xmath111 ) , even for very large @xmath73 the scaled variable @xmath128 is small , so that the long - range ( @xmath127 ) behavior of @xmath124 is more and more dominated by the small @xmath128 part of @xmath131 , i.e. , it behaves more and more like @xmath132 rather than like @xmath133 . it is this increasing dominance of the `` short - range component of the long - range part '' that causes the @xmath134 behavior in the correlation energy per electron of the high - density electron gas , and thus the divergence of the lda correlation energy in the large-@xmath76 and large-@xmath121 limit of the he and hooke s atom sequences ( see , e.g. , ref . ) . in fact , when @xmath22 ( or @xmath23 ) , the high - density long - range behavior of @xmath124 affects the long - range part of @xmath135 in eq . ( [ eq_lda ] ) . the small-@xmath128 behavior @xmath136 of the function @xmath131 is related to the @xmath81 divergence of the coulomb potential at _ small _ @xmath50 . for this reason , the @xmath134 high - density behavior of the correlation energy is still present in a uniform electron gas with screened ( or short - range only ) coulomb interaction ( e.g.,@xcite @xmath137 ) , while is absent in an electron gas with long - range - only interaction ( e.g.,@xcite @xmath138 ) . we have computed the intracule density and the correlation energy for the high - density ( weakly - correlated ) limit of the he and hooke s atom isoelectronic series via an approach@xcite based on an `` average pair density functional theory '' ( apdft ) , and inspired by the seminal work of overhauser.@xcite unlike the currently available correlation energy functionals analyzed in ref . , the apdft approach gives accurate results for both series . in its present formulation , the apdft approach works well for two - electron systems and for the uniform electron gas . its extension to many - electron systems of nonuniform density is a big challenge , and we are presently exploring several different paths to achieve this ambitious goal.@xcite we have also analyzed the lda failure in the same weakly - correlated limit of the he and hooke s atom series , in terms of the long - range part of the intracule density . the results of katriel _ et al._@xcite show that higher - order functionals such as pbe@xcite and tpss@xcite can reasonably fix the lda problems in the case of the he isoelectronic series , but are much less satisfactory for the hooke s atom sequence , yielding a wrong scaling in the @xmath23 limit ( pbe ) or a correct scaling with an error of about 40% on the asymptotic value of the correlation energy ( tpss ) . as stressed by katriel _ et al._@xcite , these differences in performances for the two series raise serious doubts on the universality of currently available correlation energy functionals . the accuracy of the results obtained via the apdft approach for both series suggests that the effort towards its generalization to many - electron systems of nonuniform density could be really worthwhile . it is a pleasure to dedicate this methodological paper to dennis salahub , who did pioneering work not only in the applications of dft , but also in the exploration of new methodologies in the dft framework . for the he isoelectronic series eq . ( [ eq_fks1 ] ) corresponds to @xmath140 + 81 e^{2x}(x-1)[{\rm ei}(-6\,x ) \nonumber \\ & & -{\rm ei}(-4\,x ) ] + e^{-2x}[164 + 27 ( 3+x\,(9 + 4\,x\,(3 + 2\,x ) ) ) [ { \rm ei}(-2\,x ) \nonumber \\ & & -\gamma-\log(x ) ] + 3\,x\,[-163 + 6\,x\,(15+x\,(7 + 10\,x))-27\log(4/3 ) ] \nonumber \\ & & -162\log(2)+81\log(3)]\big\},\end{aligned}\ ] ] where @xmath141 and the function @xmath142 have been defined after eq . ( [ eq_chi ] ) . 99 kohn , w. _ rev . phys . _ * 1999 * , _ 71 _ , 1253 . mattsson , a.e . _ science _ * 2002 * , _ 298 _ , 759 . fiolhais , c. ; nogueira , f. ; marques , m. ( eds . ) , _ a primer in density functional theory _ ; springer - verlag : berlin , 2003 . perdew , j. p. ; ruzsinszky , a. ; tao , j. ; staroverov , v. n. ; scuseria , g. e. ; csonka , g. i. _ j. chem . phys . _ * 2005 * , _ 123 _ , 062201 . harris , j. ; jones , r. _ j. phys . f _ * 1974 * , _ 4 _ , 1170 . + langreth , d.c . ; perdew , j.p . _ solid state commun . _ * 1975 * , _ 17 _ , 1425 . + gunnarsson , o. ; lundqvist , b.i . b _ * 1976 * , _ 13 _ , 4274 . yang , w. _ j. chem . phys . _ * 1998 * , _ 109 _ , 10107 . gori - giorgi p. ; savin , a. _ phys . a. _ * 2005 * , _ 71 _ , 032513 . gori - giorgi , p. ; savin , a. _ philos . mag . _ * 2006 * , _ 86 _ , 2643 . gori - giorgi , p. ; savin , a. submitted to _ int . b _ ; arxiv : cond - mat/0611324 . arxiv.org eprint archive . . ( accessed nov 13 , 2006 ) . overhauser , a.w . j. phys . _ * 1995 * , _ 73 _ , 683 . gori - giorgi , p. ; perdew , j.p . b _ * 2001 * , _ 64 _ , 155102 . davoudi , b. ; polini , m. ; asgari , r. ; tosi , m.p . _ phys . rev . b _ * 2002 * , _ 66 _ , 075110 . corona , m. ; gori - giorgi , p. ; perdew , j.p . b _ * 2004 * , _ 69 _ , 045108 . + nagy , i. ; diez muio , r. ; juaristi , j.i . ; echenique , p.m. _ phys . rev . b _ * 2004 * , _ 69 _ , 233105 . katriel , j. ; roy , s. ; springborg , m. _ j. chem . _ * 2006 * , _ 124 _ , 234111 . schwartz , c. _ ann . _ * 1959 * , _ 6 _ , 156 . white , r.j . ; byers brown , w. _ j. chem . phys . _ * 1970 * , _ 53 _ , 3869 . + cioslowski , j. ; pernal , k. _ j. chem . phys . _ * 2000 * , _ 113 _ , 8434 . _ j. chem . _ * 2006 * , _ 125 _ , 184104 . dalgarno , a. ; lewis , j.t . ( london ) _ * 1955 * , _ a233_,70 . + dalgarno , a. _ proc . ( london ) _ * 1956 * , _ a69 _ , 784 . + dalgarno a. ; stewart , a.l . _ proc . ( london ) _ * 1956 * _ a238 _ , 276 . young , w.h . ; march , n. h. _ phys . rev . _ * 1958 * , _ 109 _ , 1854 . freund , d.e . ; huxtable , b.d . ; morgan iii , j.d . _ phys . rev . * 1984 * , _ 29 _ , 980 . we used an improved version ( provided to us by c. umrigar ) of the accurate variational wavefunctions described in this work to obtain one - electron densities @xmath10 and functions @xmath20 . see also umrigar , c.j . ; gonze , x. _ phys . a _ * 1994 * , _ 50 _ , 3827 . see , e.g. , huang , c .- j . ; umrigar , c.j . _ phys . * 1997 * , _ 56 _ , 290 . savin , a. ; colonna , f. ; pollet , r. _ int . j. quantum chem . _ * 2003 * , _ 93 _ , 166 , and references therein . savin , a. in _ recent developments and applications of modern density functional theory _ , edited by seminario , j.m . elsevier : amsterdam , 1996 ; p. 327 . + leininger , t. ; stoll , h. ; werner , h .- j . ; savin , a. _ chem . phys . lett . _ * 1997 * , _ 275 _ , 151 . + pollet , r. ; savin , a. ; leininger , t. ; stoll , h. _ j. chem . phys . _ * 2002 * , _ 116 _ , 1250 . pollet , r. ; colonna , f. ; leininger , t. ; stoll , h. ; werner , h .- j . ; savin , a. _ int . j. quantum chem . _ * 2003 * _ 91 _ , 84 . toulouse , j. ; colonna , f. ; savin , a. _ phys . rev . a _ * 2004 * , _ 70 _ , 062505 . gori - giorgi , p. ; perdew , j.p . b _ * 2002 * , _ 66 _ , 165118 . burke k. ; angulo j.c . ; perdew j.p . * 1994 * , _ 50 _ , 297 . rassolov , v. a. ; pople , j. a. ; ratner , m. a. _ phys . b _ * 1999 * _ 59 _ , 15625 . + rassolov , v. a. ; pople , j. a. ; ratner , m. a. _ phys . rev . b _ * 2000 * , _ 62 _ , 2232 . wang , y. ; perdew , j. p. _ phys . b _ * 1991 * , _ 44 _ , 13298 . perdew , j.p . ; wang , y. _ phys . rev . b _ * 1992 * , _ 46 _ , 12947 . perdew , j.p . ; mcmullen , e.r . ; zunger , a. _ phys . * 1981 * _ 23 _ , 2785 . zecca , l. ; gori - giorgi , p. ; moroni , s. ; bachelet , g.b . b _ * 2004 * , _ 70 _ , 205127 . paziani , s. ; moroni , s. ; gori - giorgi , p. ; bachelet , g.b . * 2006 * , _ 73 _ , 155111 . toulouse , j. ; gori - giorgi , p. ; savin , a. _ int . j. quantum chem . _ * 2006 * , _ 106 _ , 2026 . perdew j.p . ; burke k. ; ernzerhof m. _ phys . lett . _ * 1996 * , _ 77 _ , 3865 . tao j. ; perdew j.p . ; staroverov v.n . ; scuseria g.e . lett . _ * 2003 * , _ 91 _ , 146401 .	the `` extended overhauser model '' [ overhauser , a.w . _ can . j. phys . _ * 1995 * , _ 73 _ , 683 ] for the calculation of the spherically and system - averaged pair density ( apd ) has been recently combined with the kohn - sham equations to yield realistic apd and correlation energies . in this work we test this approach in the high - density ( weakly - correlated ) limit of the he isoelectronic series and of the hooke s atom isoelectronic series . unlike many of the commonly used energy functionals , the overhauser approach yields accurate correlation energies for both series .
the standard model of particle physics provides a solid foundation from which physical phenomena of strong and electroweak nature from high - energy scattering to atomic and molecular structure can be successfully described . the standard model itself does not predict precise values of fundamental constants such as the fine structure constant @xmath6 or the electron - to - proton mass ratio @xmath7 , but rather accepts the experimentally observed values as input parameters to the physical theory . speculative theories which go beyond the standard model , such as string theories , suggest that these constants may vary in time or space @xcite , enticing both theorists and experimentalists alike to contrive favorable means for detecting variations of these constants . one method for determining variations or limitations on variations in @xmath1 and @xmath2 is from analysis of atomic or molecular absorption lines originating from interstellar space . comparison with laboratory spectra can , in principle , reveal variations on cosmological time or distance scales . employing this method , groups have reported evidence for non - zero variations in both @xmath1 @xcite and @xmath2 @xcite , though other analyses have shown no variation at similar levels of accuracy @xcite ( see also comment @xcite ) . complementary to the astrophysical studies , terrestrial experiments probe @xmath1- and @xmath2-variation on much smaller time and distance scales , but benefit from the high - precision and reproducibility that the spectroscopic experiments offer . given in terms of temporal drift , laboratory experiments have placed stringent , model - free constraints on variations of both @xmath1 @xcite and @xmath2 @xcite : @xmath8 with the dot signifying a derivation with respect to time . both the astrophysical and laboratory methods may realize significant gains by utilizing atomic or molecular species which have enhanced sensitivity to @xmath1- and @xmath2-variation . much theoretical effort has been dedicated to identifying such systems , with a general strategy being to locate accidental near - degeneracies within energy spectra . in this spirit , measurements have been proposed for multiply - charged ions @xcite , diatomic @xcite and more complex @xcite molecules , and even nuclei @xcite . note that `` near - degeneracy '' is a relative term here ; for example , the ground and anomalously low - lying excited state of the @xmath9th nucleus being separated by an interval of 7.6 ev @xcite are nearly - degenerate relative to the typical energy scale of nuclear excitation ( @xmath10 kev ) . the relative sensitivity to @xmath1- and @xmath2-variation for a given transition may be parameterized in terms of dimensionless coefficients @xmath11 and @xmath12 , defined by the relation @xmath13 with @xmath14 being the transition energy . `` typical '' transitions have sensitivity coefficients on order of unity or less . for example , in a diatomic molecule a typical fine structure transition has @xmath15 , @xmath16 , while a typical vibrational transition has @xmath17 , @xmath18 . near - degeneracies resulting from a cancellation between fine structure and vibrational intervals can , however , lead to transitions with sensitivity coefficients orders of magnitude larger than unity @xcite . several atomic , molecular , or nuclear transitions which enjoy large enhancement , however , may be irrelevant for astrophysical studies or may prove unfavorable for spectroscopic experiments , and therefore may be of limited utility . in this paper we consider the molecular ion nh@xmath0 as a candidate for measuring variation in @xmath1 and @xmath2 . nh@xmath0 possesses an accidental near - degeneracy between its ground ( @xmath19 ) and first excited ( @xmath20 ) electronic states , these being separated by only a few hundred @xmath4 ( for scale , the next electronic state is @xmath21 higher @xcite ) . as we will show , this near - degeneracy results in enhanced sensitivity coefficients of order 10100 for a number of transitions within the rotational spectrum . moreover , nh@xmath0 is a light molecule which has been suggested to be a component in interstellar clouds @xcite , though to - date it has not been detected in such media @xcite . further still , motivated by the search for an electron electric dipole moment , leanhardt _ et al . _ @xcite have argued that high precision spectroscopy may be performed on molecular ions within a paul trap , with experiments now underway for hff@xmath0 @xcite . these considerations suggest that nh@xmath0 could serve as a valuable probe of @xmath1- and @xmath2-variation in future astrophysical or laboratory studies . the rotational spectrum of nh@xmath0 was analyzed experimentally in some detail several years ago by kawaguchi and amano @xcite and more recently by hbers _ et al . _ both papers tabulate molecular parameters which , together with the appropriate effective hamiltonian , are capable of furnishing the rotational spectrum . evolution of the spectrum with respect to @xmath1- and @xmath2-variation may be determined so long as the the scaling of the molecular parameters with respect to @xmath1 and @xmath2 is known . a list of molecular parameters relevant to the @xmath5 and @xmath3 states of nh@xmath0 are provided in table [ tab : params ] along with their lowest - order scalings with @xmath1 and @xmath2 . lcciclcciclcparam . & scaling & & & param . & scaling & & & param . & scaling + + @xmath22 & @xmath23 & & & @xmath24 & @xmath23 & & & @xmath25 & + @xmath26 & @xmath2 & & & @xmath27 & @xmath28 & & & @xmath29 & @xmath23 + @xmath30 & @xmath31 & & & @xmath32 & @xmath28 & & & @xmath33 & @xmath34 + @xmath35 & @xmath36 & & & @xmath37 & @xmath38 & & & @xmath39 & @xmath23 + @xmath40 & @xmath23 & & & @xmath41 & @xmath31 & & & @xmath42 & @xmath23 + @xmath43 & @xmath28 & & & @xmath44 & @xmath36 & & & @xmath45 & @xmath28 + figure [ fig : spec ] illustrates the rotational spectra of nh@xmath0 for @xmath46 and @xmath47 vibrational states . we see that the @xmath48 state is in such close proximity to the ground @xmath49 state that the corresponding rotational spectra begin to overlap for relatively low ( @xmath50 ) levels of the @xmath51 ladder . for @xmath47 , the states are closer and even the lowest rotational levels are seen to be heavily intermixed . this close proximity of @xmath5 and @xmath3 states allows for sizable coupling between the two via spin - orbit interaction . this is accounted for in the effective hamiltonian through the non - standard molecular parameters @xmath39 , @xmath42 , and @xmath52 appearing in table [ tab : params ] . in general , this coupling results in noticeable perturbations to the rotational spectrum , with one example being that the ground @xmath53-doublet interval is doubled in size @xcite . the effect is more pronounced for close @xmath5 and @xmath3 rotational levels of similar angular momentum @xmath54 and parity @xmath32 ( these being conserved quantum numbers ) . one noteworthy case is the @xmath55 and @xmath56 rotational levels ( see fig . [ fig : spec ] ) . due to spin - orbit coupling , the energy eigenstates are nearly equal admixtures of the unperturbed @xmath5 and @xmath3 states . in table [ tab : params ] we have decomposed the term energy for a given electronic and vibrational state into three contributions , @xmath57 where the respective terms correspond to the non - relativistic electronic , relativistic electronic , and vibrational contributions . for each vibrational subspace , only @xmath58 is required to produce the spectrum , this being supplied by refs . however , to determine the overall @xmath1- and @xmath2-dependence we require the partial contributions . the vibrational part may be inferred from ref . @xcite , and for the @xmath46 subspace it is found to be @xmath59 corresponding to the difference in zero - point energies . for the @xmath47 subspace this difference is a factor of 3 larger . the relativistic electronic contribution can not be extracted from experiment ; we have determined it by _ ab initio _ calculation to be @xmath60 the details of this computation are reserved for the appendix . constraining @xmath58 to the experimental value implies a non - relativistic electronic contribution of @xmath61 this last part is insensitive to both @xmath1- and @xmath2-variation . finally , in our analysis we choose to neglect the hyperfine splitting of the rotational levels . according to hbers _ et al . _ @xcite the largest hyperfine constants are of order 100 mhz , or @xmath62 . as long as the smallest @xmath14 considered are of the order @xmath63 , the hyperfine interaction does not change these transition energies by more than a few percent . the same must be true for sensitivity coefficients @xmath11 and @xmath12 . cd .. 4.3d .. 2.3d .. 4.2d .. 4.2d .. 1.3cic cd .. 4.3d .. 2.3d .. 4.2d .. 4.2d .. 1.3 @xmath64 & & & & & & & & @xmath64 & & & & & + + & + @xmath65 & 0.454 & 0.454 & 2.08 & 1.39 & 0.564 & & & @xmath65 & 16.431 & 9.455 & -2.29 & 7.29 & 0.493 + @xmath66 & 34.450 & 0.677 & 3.56 & 0.88 & 0.403 & & & @xmath66 & 64.878 & 0.418 & 0.21 & 20.94 & 0.442 + @xmath67 & 97.065 & 0.799 & 4.57 & 0.81 & 0.336 & & & @xmath67 & 131.543 & 4.582 & 1.16 & -6.68 & 0.249 + @xmath68 & 189.470 & 0.892 & 5.56 & 1.00 & 0.285 & & & @xmath68 & 222.450 & 8.464 & 2.87 & -5.13 & 0.199 + @xmath69 & 311.925 & 1.019 & 6.45 & 1.55 & 0.246 & & & @xmath69 & 340.220 & 9.783 & 3.65 & -4.91 & 0.177 + @xmath70 & 464.416 & 1.276 & 6.80 & 2.56 & 0.215 & & & @xmath70 & 485.857 & 9.626 & 4.21 & -4.92 & 0.162 + @xmath71 & 646.814 & 1.854 & 6.22 & 3.98 & 0.190 & & & @xmath71 & 659.709 & 8.654 & 4.76 & -5.10 & 0.151 + @xmath72 & 858.917 & 3.187 & 4.83 & 5.62 & 0.168 & & & @xmath72 & 861.832 & 7.243 & 5.46 & -5.51 & 0.142 + @xmath73&1100.473 & 6.435 & 3.03 & 7.44 & 0.144 & & & @xmath73&1092.120 & 5.611 & 6.56 & -6.39 & 0.133 + @xmath74&1371.183 & 15.041 & 1.01 & 9.12 & 0.107 & & & @xmath74&1350.354 & 3.885 & 8.65 & -8.34 & 0.126 + @xmath75&1699.955 & 29.249 & 3.22 & -4.56 & 0.076 & & & @xmath76&1636.231 & 2.135 & 14.22 & -13.94 & 0.119 + @xmath77&2020.150 & 21.488 & 2.89 & -2.27 & 0.097 & & & @xmath78&1949.389 & 0.398 & 68.76 & -70.30 & 0.112 + @xmath79&2373.578 & 18.952 & 2.39 & -0.73 & 0.100 & & & @xmath79&2290.724 & 1.310 & -18.81 & 20.53 & 0.106 + @xmath80&2756.628 & 18.488 & 2.00 & 0.16 & 0.097 & & & @xmath80&2658.832 & 2.981 & -7.48 & 8.88 & 0.101 + @xmath81&3167.736 & 19.044 & 1.73 & 0.72 & 0.093 & & & @xmath81&3052.821 & 4.610 & -4.40 & 5.75 & 0.096 + @xmath82&3605.981 & 20.256 & 1.54 & 1.12 & 0.088 & & & @xmath83&3472.169 & 6.194 & -3.00 & 4.34 & 0.091 + + & + @xmath66 & 105.405 & 0.244 & -1.97 & 2.91 & 1.592 & & & @xmath84 & 151.564 & 16.803 & 1.86 & -0.67 & 1.170 + @xmath67 & 195.934 & 0.617 & -1.02 & 2.34 & 0.885 & & & @xmath85 & 237.233 & 20.745 & 2.76 & -1.79 & 0.664 + @xmath68 & 317.542 & 1.082 & -0.55 & 2.13 & 0.598 & & & @xmath86 & 352.021 & 20.618 & 3.41 & -2.37 & 0.457 + @xmath69 & 469.655 & 1.710 & -0.02 & 2.73 & 0.446 & & & @xmath87 & 495.901 & 18.888 & 3.91 & -2.72 & 0.353 + @xmath88 & 655.253 & 3.410 & 19.71 & -53.15 & 0.190 & & & @xmath88 & 668.781 & 16.511 & 4.36 & -2.99 & 0.291 + @xmath71 & 863.581 & 0.996 & -11.10 & -0.22 & 0.286 & & & @xmath89 & 870.530 & 13.906 & 4.86 & -3.27 & 0.251 + @xmath72&1103.851 & 0.150 & -81.53 & -126.86 & 0.236 & & & @xmath90&1100.956 & 11.276 & 5.50 & -3.64 & 0.221 + @xmath91&1374.211 & 5.020 & 1.73 & 19.93 & 0.165 & & & @xmath91&1359.806 & 8.724 & 6.44 & -4.25 & 0.198 + @xmath74&1694.241 & 20.595 & 3.65 & -4.78 & 0.140 & & & @xmath92&1646.767 & 6.297 & 8.04 & -5.39 & 0.179 + @xmath76&2017.566 & 15.979 & 2.66 & -1.11 & 0.159 & & & @xmath75&1961.476 & 4.008 & 11.39 & -7.90 & 0.164 + @xmath78&2373.114 & 15.516 & 2.01 & 0.32 & 0.151 & & & @xmath77&2303.527 & 1.858 & 22.27 & -16.23 & 0.150 + @xmath93&2757.519 & 16.311 & 1.65 & 0.95 & 0.140 & & & @xmath93&2672.636 & 0.158 & -239.61 & 185.81 & 0.139 + @xmath94&3169.622 & 17.721 & 1.43 & 1.29 & 0.130 & & & @xmath94&3069.899 & 2.041 & -17.09 & 14.15 & 0.129 + @xmath95&3608.690 & 19.585 & 1.30 & 1.53 & 0.120 & & & @xmath95&3492.954 & 3.787 & -8.62 & 7.58 & 0.120 + ccrdddr @xmath96 & @xmath97 & & & & & + + + @xmath98 & @xmath99 & 443.287 & -24.658 & -3.74 & 17.82 & 0.005 + @xmath100 & @xmath101 & 643.977 & -0.983 & -76.84 & 249.78 & 0.002 + @xmath100 & @xmath102 & 643.977 & -7.866 & -4.68 & 30.10 & 0.165 + @xmath103 & @xmath102 & 649.676 & -2.167 & -48.45 & 193.99 & 0.001 + + @xmath104 & @xmath105 & 0.000 & -6.976 & -13.68 & 22.48 & 0.385 + @xmath106 & @xmath107 & 21.486 & 5.055 & 19.73 & -33.94 & 0.208 + @xmath108 & @xmath109 & 58.529 & -5.931 & -19.95 & 25.71 & 0.222 + @xmath110 & @xmath111 & 77.352 & 12.474 & 3.48 & -8.68 & 0.023 + @xmath112 & @xmath111 & 83.854 & 18.976 & 5.44 & -6.48 & 0.003 + @xmath113 & @xmath114 & 104.966 & -29.795 & -0.10 & 5.94 & 0.099 + @xmath115 & @xmath116 & 190.451 & -26.037 & -1.96 & 7.87 & 0.003 + ccrdddr @xmath117 & @xmath118 & & & & & + + @xmath119 & @xmath120 & 134.760 & 3.217 & 30.68 & -11.53 & 0.039 + @xmath120 & @xmath121 & 216.488 & -5.962 & -6.89 & 0.05 & 0.020 + @xmath121 & @xmath122 & 331.403 & -8.817 & -0.94 & -2.13 & 0.008 + @xmath122 & @xmath123 & 477.013 & -8.844 & 0.94 & -3.03 & 0.004 + @xmath123 & @xmath124 & 652.270 & -7.439 & 2.18 & -3.95 & 0.002 + before proceeding , we must briefly discuss our management of the available experimental data . the effective hamiltonian employed by hbers _ et al . _ @xcite differs subtly from the earlier work of kawaguchi and amano @xcite , with this difference extending to the underlying definitions and therefore scalings of the molecular parameters found in table [ tab : params ] . for the discussion of the previous section we chose to be consistent with hbers _ et al . _ , who in turn follow closely the methodical development of the effective hamiltonian given in the book of brown and carrington @xcite . however , in our analysis we have calculated the rotational spectrum and sensitivity coefficients by following both references as independently as possible , taking care to scale molecular parameters as appropriate for each case . in principle , results following from ref . @xcite would be preferred , as the data is more accurate and the effective hamiltonian formulation is more transparent . unfortunately , ref . @xcite lacks molecular parameters for the @xmath47 subspace and in itself does not provide information for the vibrational contribution to the term energy in the @xmath46 subspace [ i.e. , eq . ( [ eq : tvib ] ) ] . in this section , we choose to maintain consistency by presenting our results following exclusively from ref . implications of this particular choice will be discussed more in sec . [ sec : accuracy ] . in practice , the dimensionless sensitivity coefficients for a given transition are found from the relation ( here @xmath125 ) @xmath126 where @xmath127 is the energy difference between levels and @xmath128 is the difference between _ dimensional _ sensitivity coefficients defined by @xmath129 scaling molecular parameters accordingly , we obtain rotational energy levels by diagonalizing the effective hamiltonian for multiple values of @xmath2 and @xmath1 in the neighborhood of the known present - day values ; numerical differentiation is then used to obtain the coefficients @xmath130 for each level . note that proper interpretation of eq . ( [ eq : qx ] ) requires us to specify our employed unit system , namely atomic units ( i.e. , the atomic unit of energy @xmath131 is assumed constant ) . a more subtle point is that the sensitivity coefficients @xmath132 , despite being dimensionless , also depend on this specification of atomic units . differences @xmath133 , on the other hand , do not depend on the choice of unit system . tables [ tab : omegadoublets ] [ tab_pi3pi1v1 ] display our results for select ( low - frequency ) transitions within the nh@xmath0 rotational spectrum . along with the transition energy @xmath14 and sensitivity coefficients @xmath11 and @xmath12 , we also tabulate reduced matrix elements for electric dipole transition amplitudes , @xmath134 where @xmath135 is a spherical component of the electric dipole operator and @xmath136 is an angular momentum projection along the @xmath137-axis in a space - fixed frame . in the non - relativistic limit all off - diagonal amplitudes between electronic states @xmath138 , @xmath139 , and @xmath140 vanish . these amplitudes appear only after spin - orbit interaction mixes these state . as spin - orbit coupling is embedded in the effective hamiltonian , only the dipole moments of the @xmath51 and @xmath140 states in the molecule - fixed frame are further needed to determine the off - diagonal amplitudes . we have calculated the dipole moments for these states to be 0.91 a.u . and 0.82 a.u . , respectively ( see appendix ) . sensitivity of @xmath53-doublet transitions to @xmath1- and @xmath2-variation has been discussed previously in refs . @xcite . in nh@xmath0 the situation is complicated by the proximity and strong spin - orbit coupling of the @xmath5 and @xmath3 states , resulting in less predictable behavior of the intervals . in some instances this leads to anomalously small transition energies @xmath14 , with subsequently magnified sensitivity coefficients [ see eq . ( [ eq : qx ] ) ] . one example is the @xmath141 doublet of the @xmath142 state . here we find that the spin - orbit coupling leads to a more than 10-fold decrease in the transition energy , with large sensitivity coefficients @xmath143 and @xmath144 following . full results for @xmath53-doublet transitions are given in table [ tab : omegadoublets ] . another mechanism for enhancement is the state - mixing itself , as exemplified by the @xmath145 doublet of the @xmath142 state . indeed , spin - orbit coupling has a significant effect on this transition energy , even causing @xmath14 to change sign . this modification of @xmath14 does not promote enhancement , however , as the absolute value of @xmath14 is found to increase by about 50% . enhancement here comes rather from the numerator of eq . ( [ eq : qx ] ) . as mentioned in sec . [ sec : rotspec ] , the negative parity state in this doublet is nearly an equal admixture of @xmath5 and @xmath3 due to spin - obit coupling of near resonant levels . the @xmath5 portion brings no enhancement , while the @xmath3 portion provides larger @xmath146 and @xmath147 values expected of @xmath148 transitions . in fact , @xmath147 may be readily estimated : @xmath149 \\ & \approx&-210~\mathrm{cm}^{-1},\end{aligned}\ ] ] where the first term is from a difference in vibrational energies ( with a factor of 1/2 from the @xmath34 scaling ) and the second term is from a difference in rotational energies [ @xmath150 for both states and @xmath151 is the difference in @xmath152 ; the leading 1/2 is a weight factor accounting for the fact that the eigenstate is about `` half '' @xmath3 . the comparatively small interval , @xmath153 , implies a sensitivity coefficient @xmath154 . this agrees with our tabulated result for this transition , wherein @xmath155 and @xmath156 . from the preceding discussion , we can infer that sensitivity coefficients @xmath11 and @xmath12 will be enhanced for @xmath157 cross transitions having @xmath14 comparable to the @xmath53-doublet intervals ( @xmath158 ) . from the perspective of electronic transitions , these are anomalously small @xmath14 . there are a number of such transitions in the spectrum , and in tables [ tab_sigpiv0 ] we present results for transitions with @xmath159 . again we find sensitivity coefficients on the order of @xmath160 . transitions between @xmath51 and @xmath140 states require the change of the electronic spin , @xmath161 . thus , we would expect strong suppression of transition amplitudes . this is found to be generally true for the @xmath46 subspace . the exception is the transition complementary to @xmath53-doublet transition described in the previous section ( involving the other state which is a nearly equal admixture of @xmath5 and @xmath3 ) ; this has an `` enhanced '' transition amplitude on the order of the @xmath53-doublet transition amplitudes . for the @xmath47 subspace , the difference in term energies is smaller ( @xmath162 compared to @xmath163 for @xmath46 ) and the mixing caused by spin - orbit interaction is generally stronger . because of this , there are several sufficiently strong transitions . fine - structure transitions between @xmath138 and @xmath139 states constitute another source of low - frequency transitions ; following from our analysis of @xmath53-doubling , we may suspect enhancement of @xmath11 and @xmath12 here as well . in the hund s case ` a ' limit , these transitions require spin - flip ( i.e. , @xmath164 , for spin - projection @xmath165 on the internuclear axis ) and are therefore forbidden in the non - relativistic limit . the interaction with the @xmath140 state can open some of these transitions . it is a second order effect and the transition amplitudes are significantly smaller than for @xmath157 transitions . table [ tab_pi3pi1v1 ] lists a number such transitions within the @xmath47 subspace , with the most sensitive transition having @xmath166 and @xmath167 . as mentioned in sec . [ sec : results ] , we analyzed the nh@xmath0 rotational spectrum following the papers of kawaguchi and amano @xcite and hbers _ et al . _ @xcite as independently as possible . a comparison of these results allows us to assess the accuracy of our data presented in tables [ tab : omegadoublets ] [ tab_pi3pi1v1 ] . for the majority of the transitions , our values of @xmath11 and @xmath12 agree to within a few percent following both references . this agreement , however , deteriorates for a handful of the transitions , with these transitions tending to be the ones with the largest @xmath11 and @xmath12 . the bulk of the discrepancy in these cases can be attributed to a difference in @xmath14 [ i.e. , the denominator of eq . ( [ eq : qx ] ) ] , which in the most extreme case differs by a factor of two . direct measurement of the transition energies can remove this uncertainty , provided that the sensitivity coefficients are corrected appropriately , @xmath168 with this prescription , theoretical uncertainty is then due to calculated @xmath146 and @xmath147 . we find that our @xmath146 are consistent to @xmath169 , with agreement being somewhat better for @xmath147 at @xmath170 . we emphasize that our comparison here is limited to the @xmath46 subspace , as hbers _ et al . _ do not give data for the @xmath47 subspace . the above analysis effectively gauges the spread in @xmath146 and @xmath147 arising from uncertainty in the molecular parameters of table [ tab : params ] ( we note that the difference between molecular parameters tabulated in refs . @xcite is typically more than 1-@xmath171 , even after the disparity in effective hamiltonians is accounted for @xcite ) . the notable exclusion is the relativistic electronic contribution to the term energy . this is supplied by our computed value , @xmath172 , for which we estimate an uncertainty of 10% ( see appendix ) . taking values within this window , we find deviations in @xmath146 to be under 10% for most transitions , but up to 30% for a select few . since @xmath29 is insensitive to variations in @xmath2 , the @xmath147 are not affected . finally , we mention that our analysis is based on the lowest - order scaling of the molecular parameters . in principle , higher - order contributions alter the @xmath1- and @xmath2-dependence of the parameters . the higher - order contributions are small , and we expect their effects on @xmath146 and @xmath147 to be negligible . altogether , we estimate uncertainty in our @xmath146 and @xmath147 to be about 30% and 5% , respectively . with experimental values of @xmath14 and subsequent application of eq . ( [ eq : qcorr ] ) , this uncertainty extends to our tabulated sensitivity coefficients @xmath11 and @xmath12 . for most transitions , the 30% estimate is highly conservative . for laboratory experiments , it is useful to quantify absolute shift in addition to relative shift . the largest @xmath146 and @xmath147 have magnitudes of about @xmath173 and @xmath174 , respectively . for temporal variations of @xmath1 and @xmath2 at the current laboratory limits , ( [ eq : lablimits ] ) , the resulting drift in @xmath14 is latexmath:[\[\begin{aligned } here @xmath1-variation is neglected , as it s laboratory constraint is three orders of magnitude tighter than for @xmath2-variation . we may conclude that if hertz - level precision is achievable with molecular - ion spectroscopy , then nh@xmath0 represents a good system to probe for @xmath2-variation in the laboratory . the natural linewidth gives a fundamental limit to spectroscopic precision . the contribution to the natural linewidth from a given decay channel @xmath176 is proportional to the product @xmath177 , with @xmath14 being the transition energy and @xmath178 being the electric dipole matrix element connecting the two states . for the @xmath46 subspace , the dominant decay channels are the rotational transitions @xmath179 . as the rotational splitting @xmath180 grows with @xmath181 , so too does the natural linewidth . for large @xmath181 , where the hund s case ` b ' limit is realized , we find the linewidth to be @xmath182 where @xmath183 is the dipole moment in the molecule - fixed frame . with the experimental values of @xmath26 @xcite and our calculated values of @xmath183 , we establish that @xmath184 for rotational levels @xmath185 . for the @xmath47 subspace , decay to the ground vibrational state opens up as well . the vibrational transition energy @xmath186 is larger than the rotational transition energies @xmath180 , whereas the dipole matrix element is suppressed in comparison . we find the resulting contribution to the linewidth to be @xmath187 @xmath188 being a factor of order unity which accounts for dependence of @xmath183 on internuclear separation @xmath189 ( as with @xmath183 , evaluation of @xmath188 at the equilibrium separation is implicit in the first expression ) . with experimental values of @xmath186 @xcite and computed values of @xmath188 ( see appendix ) , we find @xmath190 for the @xmath5 state and @xmath191 for the @xmath3 state . assuming practical limitations @xmath192 , depending on statistics , we conclude that the natural linewidth should allow measurement of the transition energies at the required level of precision . to summarize , we have analyzed the sensitivity of the nh@xmath0 rotational spectrum to variations in the fine structure constant @xmath1 and electron - to - proton mass ratio @xmath2 . we find enhanced sensitivity for a number of low - frequency transitions within both the @xmath46 and @xmath47 spectra , having sensitivity coefficients on the order of @xmath160 . the enhanced sensitivity for these transitions is attributed to the near degeneracy of the ground @xmath19 and excited @xmath20 electronic states and the significant spin - orbit coupling between them . these results could prove useful in future astrophysical and laboratory searches for @xmath1- and @xmath2-variation . this work was supported by the marsden fund , administered by the royal society of new zealand . mk further acknowledges support from rfbr grant 11 - 02 - 00943 and fqxi mini grant . vf further acknowledges support by the arc . the potential energy curves and the dipole moments of the @xmath193 and @xmath194 states were obtained using the molpro computational package @xcite , within the multireference configuration interaction approach with single and double excitations ( mrcisd ) . correlation consistent dunning aug - cc - pvqz basis sets were employed for both atoms @xcite ; convergence of calculated spectroscopic constants with respect to the basis set was verified . the second order douglas - kroll - hess hamiltonian doukro74,hes86 was used to account for scalar relativistic effects . the contribution of spin - orbit coupling was calculated by employing the breit - pauli ( bp ) operator . within the molpro package , the lowest - order one- and two - electron spin - orbit bp operators are used for computing the matrix elements between internal configurations ( no electrons in external orbitals ) , while for contributions of external configurations a mean - field one - electron fock operator is employed . the error introduced by this approximation is generally negligible @xcite . the calculated potential energy curves were used to extract the relativistic electronic contribution to the term energy , @xmath172 . agreement between our calculated spin - orbit splitting , @xmath195 , and the experimental spin - orbit constant @xmath196 @xcite gives testament to our accuracy . our accuracy is likely limited by higher - order ( @xmath197 ) effects , arising primarily from the spin - orbit mixing between the @xmath51 and @xmath140 states . this mixing is already explicitly accounted for in the effective hamiltonian by the parameters @xmath39 , @xmath42 , and @xmath52 , and its effect on the potential energy curves here represents somewhat of an intrusion . we estimate these effects to be a few @xmath198 , leading us to ascribe 10% accuracy to our calculated value of @xmath172 . we have computed the dipole moments in the molecule - fixed frame ( origin at the center - of - mass ) to be 0.91 a.u . and 0.82 a.u . for the @xmath51 and @xmath140 states , respectively . cheng _ et al . _ have previously computed the @xmath51 dipole moment to be 0.7897 a.u . @xcite , and we suspect that their result is more accurate . high - accuracy is not required for the dipole moments , and this agreement is sufficient for our purposes . we also compute dependence of the dipole moments on internuclear separation ; for the factor @xmath188 appearing in eq . ( [ eq : gamvib ] ) , we obtain values 0.73 and 2.0 for @xmath5 and @xmath3 , respectively . 51ifxundefined [ 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/revmodphys.75.403 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.87.091301 [ * * , ( ) ] @noop ( ) , link:\doibase 10.1103/physrevlett.96.151101 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.98.240801 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.92.121302 [ * * , ( ) ] link:\doibase 10.1051/0004 - 6361/200811475 [ * * , ( ) ] @noop * * , ( ) @noop ( ) , link:\doibase 10.1103/physrevlett.99.239001 [ * * , ( ) ] link:\doibase 10.1126/science.1154622 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.100.150801 [ * * , ( ) ] link:\doibase 10.1103/physreva.80.064502 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.105.120801 [ * * , ( ) ] link:\doibase 10.1103/physreva.73.034101 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.99.150801 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.100.043202 [ * * , ( ) ] @noop * * , ( ) link:\doibase 10.1103/physreva.80.022118 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.91.241302 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.91.011301 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.100.043201 [ * * , ( ) ] @noop * * , ( ) link:\doibase 10.1103/physrevlett.106.100801 [ * * , ( ) ] @noop * * , ( ) @noop ( ) , link:\doibase 10.1103/physrevlett.97.092502 [ * * , ( ) ] link:\doibase 10.1103/physrevc.79.034302 [ * * , ( ) ] @noop * * , ( ) link:\doibase 10.1103/physrevc.79.064303 [ * * , ( ) ] @noop * * , ( ) link:\doibase 10.1103/physrevlett.102.210801 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.104.200802 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.98.142501 [ * * , ( ) ] link:\doibase 10.1103/physreva.82.022106 [ * * , ( ) ] @noop _ _ ( , , ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop ( ) , @noop ( ) , @noop * * , ( ) @noop * * , ( ) @noop _ _ ( , , ) @noop `` , '' link:\doibase 10.1103/physrevd.67.083507 [ * * , ( ) ] link:\doibase 10.1063/1.478678 [ * * , ( ) ] \doibase doi : 10.1016/0003 - 4916(74)90333 - 9 [ * * , ( ) ] link:\doibase 10.1103/physreva.33.3742 [ * * , ( ) ] @noop * * , ( ) link:\doibase 10.1103/physreva.75.012502 [ * * , ( ) ]	we identify the molecular ion nh@xmath0 as a potential candidate for probing variations in the fine structure constant @xmath1 and electron - to - proton mass ratio @xmath2 . nh@xmath0 has an anomalously low - lying excited @xmath3 state , being only a few hundred @xmath4 above the ground @xmath5 state . being a light molecule , this proximity is such that rotational levels of the respective states are highly intermixed for low angular momenta . we find that several low - frequency transitions within the collective rotational spectrum experience enhanced sensitivity to @xmath1- and @xmath2-variation . this is attributable to the close proximity of the @xmath5 and @xmath3 states , as well as the ensuing strong spin - orbit coupling between them . suggestions that nh@xmath0 may exist in interstellar space and recent predictions that trapped - ion precision spectroscopy will be adaptable to molecular ions make nh@xmath0 a promising system for future astrophysical and laboratory studies of @xmath1- and @xmath2-variation .
@xmath0-hypergeometric systems ( or gkz - systems or simply hypergeometric systems ) are systems of linear partial differential equations on the complex affine space @xmath5 . although they were already considered in works by j. hrabowski @xcite and by i. m. gelfand , m. i. graev and a. v. zelevinsky @xcite , the systematic study of hypergeometric systems started with the paper by i. m. gelfand , m. m. kapranov and a. v. zelevinsky @xcite . each of these systems , denoted by @xmath6 , is determined by a pair @xmath7 where @xmath8 is an integer @xmath9 matrix of rank @xmath10 and @xmath11 is a parameter vector . an @xmath0-hypergeometric system @xmath12 is defined by the @xmath10 euler operators @xmath13 for @xmath14 and the toric operators @xmath15 associated with each pair @xmath16 such that @xmath17 . here @xmath18 stands for the monomial differential operator @xmath19 and @xmath20 . for generic parameters @xmath21 , the holomorphic solutions of @xmath6 at nonsingular points can be described by using the so - called @xmath22hypergeometric series ( see @xcite , @xcite ; see also @xcite and subsection [ gevrey - solutions - at - infinity ] ) . we are interested in divergent @xmath22hypergeometric series solutions . to study them we use the notion of slopes defined in the general setting by y. laurent @xcite . if the matrix @xmath0 is pointed ( which means that the column vectors of @xmath23 lie in an open half - space with boundary passing through the origin in @xmath24 ) , m. schulze and u. walther @xcite have described the slopes of @xmath6 with respect to coordinates varieties , generalizing previous work in @xcite , @xcite and @xcite . these slopes are closely related to the irregularity of the system @xcite and the existence of non convergent gevrey series solutions of this system . gevrey series solutions of @xmath6 were studied by the second author in @xcite ( see also @xcite and @xcite where the authors treat particular cases ) . modified hypergeometric systems were introduced by the fourth author @xcite in order to study solutions of hypergeometric systems along a curve @xmath25 for @xmath26 , @xmath27 . each of them is determined by a tuple @xmath28 where @xmath0 and @xmath29 are as before , @xmath30 and @xmath31 . we denote by @xmath32 ( or simply @xmath33 ) the matrix @xmath34 throughout this paper , we do not always assume that @xmath0 is pointed , but we assume that @xmath35 is . note that when @xmath0 is pointed , then @xmath35 also is . ( @xcite ) [ def : mgkz ] we call the following system of differential equations @xmath36 a _ modified @xmath0-hypergeometric system _ : @xmath37 with @xmath38 running over all @xmath39 such that @xmath40 . here we denote @xmath41 . the modified system is defined on the space @xmath42 with coordinates @xmath43 . let @xmath44 ( or @xmath45 ) be the weyl algebra in @xmath46 . the left ideal in @xmath44 generated by the operators in ( [ eq : mgkz1 ] ) , ( [ eq : mgkz2 ] ) , and ( [ eq : mtoric ] ) is also denoted by @xmath36 if no confusion arises . the left @xmath44-module @xmath47 is denoted by @xmath48 . by @xcite the @xmath44-module @xmath48 is holonomic for any @xmath49 . the system @xmath6 is a summand of the modified system on the space @xmath50 . more precisely , denote @xmath51 with coordinates @xmath52 and consider the map @xmath53 defined by @xmath54 . the pullback image of the ideal @xmath36 by @xmath55 equals the ideal of differential operators on @xmath56 generated by @xmath6 and @xmath57 . notice that this last ideal is nothing but the hypergeometric ideal associated with the matrix @xmath58 and the parameter vector @xmath59 . as usual we denote this ideal by @xmath60 . let us consider the local analytic situation . let @xmath61 be the sheaf of holomorphic differential operators on @xmath62 and @xmath63 the quotient sheaf @xmath64 . by the previous observation , the hypergeometric @xmath61-module @xmath65 is the extension to @xmath66 of the pullback module @xmath67 considered as a @xmath68module on the first space @xmath69 . since @xmath55 is a biholomorphic map both @xmath70-modules @xmath71 and @xmath72 are isomorphic . we can describe solutions of the original hypergeometric system @xmath6 associated to the weight vector @xmath26 via solutions of the modified system . series solutions of @xmath6 have been studied in @xcite and @xcite where the authors constructed convergent series solutions associated to the regular triangulation induced by a generic weight vector @xmath4 . the construction is generalized as follows @xcite : assume that @xmath11 is very generic ( this condition is essential in the construction ) . suppose that the initial ideal @xmath73 has a solution of the form @xmath74 , @xmath75 . then the monomial @xmath74 can be extended to a formal series solution @xmath76 of @xmath6 . we call the series @xmath77 a series solution of @xmath6 associated to the weight vector @xmath4 . the series is divergent in general . we are interested in giving an explicit expression of a solution of @xmath6 whose asymptotic expansion is @xmath77 . a standard method , in the theory of ordinary differential equations , to construct such an expression is the laplace integral representation and the borel transformation of divergent series . this method has been successful in the study of global analytic properties of solutions of ordinary differential equations , see , e.g. , the book by w. balser @xcite and the references therein . then it is a natural problem to construct a laplace integral representation corresponding to the divergent series solution @xmath77 associated to the weight vector @xmath4 . our modified system , which is a system of differential equations for @xmath78 , is used to give an answer to this problem . a key ingredient of our study is the fact that the modified hypergeometric system @xmath36 is transformed into the hypergeometric system @xmath79 associated with the matrix @xmath80 and parameter @xmath81 , by the formal inverse fourier transform @xmath82 , @xmath83 , which enables us to study the modified system using the theory of hypergeometric systems ( see subsection [ section - fourier ] ) . for example , for @xmath84 and @xmath85 the system @xmath6 is generated by @xmath86 . the modified system @xmath36 for @xmath87 , @xmath31 is generated by the three operators @xmath88 , and the inverse fourier transformation of @xmath36 is generated by @xmath89 which is equal to @xmath90 . we study the behavior of solutions of @xmath36 near the hyperplane @xmath91 in the space @xmath62 and we will give laplace integral representations of its solutions . the structure of the paper is as follows : in section [ generalities - slopes ] we recall y. laurent s definition of the slopes of a finitely generated @xmath44-module with respect to a hypersurface . in section [ section - irregularity - gkz ] we recall the use of umbrellas for the description of the slopes of a hypergeometric system given by schulze and walther @xcite , then we summarize the construction of the gevrey solutions given in @xcite and extend some of these results to the case of the gevrey solutions at infinity . in section [ section : irregularity - mgkz ] we prove that the formal inverse fourier transform , with respect to @xmath92 , of a modified hypergeometric system is an @xmath0-hypergeometric system and we use this fact to describe the slopes of the former by using the umbrella of the latter . we construct , associated with any slope of the modified system @xmath63 , a basis of its gevrey solutions , modulo convergent power series , when the parameters @xmath29 and @xmath93 are very generic , see theorem [ formal - modulo - convergent ] . moreover , for @xmath11 very generic we construct a basis of formal series solutions of the modified system for any @xmath94 and @xmath26 , see theorem [ theorem - dimension-1 ] . if in addition @xmath4 is generic this basis is reduced to a single element . later , in sections [ section : borel ] and [ section : borel_transform_revisited ] , we prove under some assumptions that this solution is an asymptotic expansion of a laplace integral representation of a solution by the borel summation method ( theorem [ borel - thm ] and section [ section : borel_transform_revisited ] ) . as an application , we give an asymptotic error evaluation of finite sums of formal series solutions of original @xmath0-hypergeometric systems with irregular singularities studied in , e.g. , @xcite and @xcite ( see theorem [ borel - thm ] , the inequality ( [ eq : asymptotic_expansion ] ) , and section [ section : borel_transform_revisited ] ) . example [ example12 ] illustrates the application for the simplest @xmath0 . several integral representations have been studied for solutions of regular holonomic hypergeometric systems ( see , e.g. , @xcite , @xcite , @xcite ) . they play a prominent role in the study of @xmath0-hypergeometric functions . however , there have been few studies of integral representations for solutions of irregular @xmath0-hypergeometric systems . we would like to point out that integral representations of holomorphic solutions of irregular @xmath0-hypergeometric systems have been recently given by a. esterov and k. takeuchi @xcite by using the so called rapid decay homology cycles . our laplace integral representation , which we propose in this paper for giving an analytic meaning to divergent series solutions , is different from their representation : the integrand of our representation is an @xmath0-hypergeometric function associated to a homogenized configuration of @xmath0 ( section [ section : borel ] ) . in section [ section : borel_transform_revisited ] we illustrate how , under some conditions , the study of the irregularity of @xmath63 along @xmath92 gives an analytic meaning to the gevrey series solutions of @xmath95 , along coordinate varieties , constructed in @xcite . more precisely we prove ( see proposition [ bounded - growth ] ) that they are asymptotic expansions of certain holomorphic solutions of @xmath95 . an interplay of algebra ( slopes and formal series ) and analysis ( borel summation method ) is a main point of this paper . in order to make a comprehensible presentation to readers from several disciplines , we often review some well - known facts to experts . we hope that our style is successful . acknowledgements : we wish to thank j. gonzlez - meneses , m. granger and d. mond for their help , suggestions and comments . we are very grateful to an anonymous referee , whose thoughtful suggestions have improved this article . recall that @xmath96 is the weyl algebra @xmath97 . in this section we write @xmath98 , @xmath99 . the variable @xmath100 is also denoted by @xmath101 and @xmath102 by @xmath103 . let @xmath104 be a linear form @xmath105 such that @xmath106 for @xmath107 , inducing the so - called @xmath108filtration on the ring @xmath44 . if @xmath109 for all @xmath110 , the associated graded ring @xmath111 is isomorphic to a polynomial ring in @xmath112 variables @xmath113 with complex coefficients . this polynomial ring is @xmath108-graded , the @xmath108-degree of a monomial @xmath114 being @xmath115 . if we need to emphasize the coefficients of the linear form we simply write @xmath116 for @xmath117 with @xmath106 for all @xmath110 . if @xmath118 and @xmath119 then the corresponding @xmath120 filtration is nothing but the usual order filtration on @xmath44 ( which is also called the @xmath121-filtration ) . if @xmath122 and @xmath123 then the corresponding @xmath120 filtration is nothing but the malgrange - kashiwara filtration on @xmath44 ( also known as the @xmath124-filtration ) with respect to @xmath91 in the remainder of this section we assume @xmath109 for all @xmath110 ; we say then that @xmath125 is a _ weight vector _ for the weyl algebra @xmath44 . all the @xmath44modules appearing here are left @xmath44-modules unless stated otherwise . we denote by @xmath126 the hyperplane defined by @xmath91 . let @xmath127 be a finitely generated @xmath44-module . to the @xmath108-filtration on @xmath44 we associate a _ good _ @xmath108-filtration on @xmath127 , by means of a finite presentation . the associated @xmath111-module @xmath128 is then finitely generated . the radical of the annihilating ideal @xmath129 , which is independent of the _ good _ @xmath108filtration on @xmath127 , defines an affine algebraic subset of the cotangent space @xmath130 . this algebraic set is called the @xmath108-_characteristic variety _ of @xmath127 and it is denoted by @xmath131 the results stated so far are well known in @xmath44-module theory and generalize @xcite which treats the case of the @xmath121filtration in @xmath44 . the case of a general @xmath108filtration has been studied for example in @xcite and , with more details , in ( * ? ? ? * section 3.2 ) in the micro - differential setting which is slightly different from the one in this paper . see also ( * ? ? ? * section 2 ) for an equivalent treatment better adapted to effective computations for modules on the weyl algebra . we consider a special type of @xmath108-filtration : for any real number @xmath132 we denote by @xmath133 either the linear form @xmath134 or the filtration on @xmath44 given by the @xmath135dimensional weight vector @xmath136 where @xmath137 is placed in the @xmath138-component . here @xmath121 ( resp . @xmath124 ) stands for the order filtration on @xmath44 ( resp . the malgrange - kashiwara filtration with respect to @xmath92 ) . [ def : slope ] ( * ? ? ? * section 3.4 ) let @xmath127 be a finitely generated @xmath44module . consider the projection @xmath139 defined by @xmath140 for any real number @xmath141 , let @xmath142 be the closure of the projection by @xmath143 of the irreducible components of the @xmath133characteristic variety @xmath144 that are not @xmath145bihomogeneous . the real number @xmath146 is said to be a _ slope _ of @xmath127 along @xmath92 at @xmath147 if and only if @xmath148 . as proved by y. laurent , see ( * ? ? ? * section 3.4 ) , any slope is a rational number and the set of slopes of @xmath127 is finite . moreover , y. laurent also proved _ loc . _ that @xmath149 is a slope of @xmath127 along @xmath92 at @xmath147 if and only if , in a neighborhood of @xmath150 , @xmath151 is not locally constant for @xmath152 with @xmath153 small enough . the irregularity of a holonomic system with respect to a smooth hypersurface ( ( * ? ? ? * df . 6.3.1 ) ) is deeply related with the slopes of the system defined with respect to the given hypersurface @xcite . recall that @xmath0 is a @xmath9 integer matrix of rank @xmath10 whose columns @xmath154 generate @xmath155 as @xmath156module . we denote by @xmath6 the hypergeometric ideal associated with @xmath0 and the parameter vector @xmath21 @xcite and by @xmath157 the corresponding hypergeometric system ( also known as @xmath158system ) . this system is the quotient of @xmath159\langle \partial_1,\ldots , \partial_n\rangle$ ] , the weyl algebra of order @xmath3 , by the left ideal @xmath6 . in this section we denote @xmath160 . we recall in this subsection some results from @xcite , where @xmath0 is assumed to be pointed , i.e. the columns of @xmath0 lie in a open half - space defined by a hyperplane passing through the origin in @xmath161 . these results will not be applied to our matrix @xmath0 but only to the matrix @xmath80 ( see subsection [ slopes - mhgs ] ) , which we assume to be pointed throughout this article . we denote by @xmath162 the @xmath110-th column of @xmath0 for @xmath163 . the @xmath108-characteristic variety of the hypergeometric system @xmath164 has been described , in a combinatorial way , by m. schulze and u. walther @xcite for any pointed matrix @xmath0 and any filtration @xmath165 such that @xmath166 for all @xmath167 . the case @xmath168 was first studied by a. adolphson @xcite . the main tool for their description is the notion of @xmath169-_umbrella _ that we define here for the sake of completeness . first of all , the @xmath169umbrella only depends on @xmath0 and on the coefficients @xmath170 of the linear form @xmath116 . 2.7 ) [ special - umbrella ] we assume that @xmath171 for all @xmath110 . the @xmath169polyhedron @xmath172 is the convex hull in @xmath161 of the set @xmath173 . the @xmath169-umbrella @xmath174 is the set of faces of @xmath175 which do not contain zero . in particular , @xmath174 contains the empty face . by @xmath176 we denote the subset of faces of dimension @xmath177 . we identify each face @xmath178 of @xmath174 with the set @xmath179 and with @xmath180 . the @xmath169-umbrella is then an abstract cell complex . when not all the @xmath170 are strictly positive then both definitions of the @xmath169-polyhedron and the @xmath169-umbrella are a little bit more involved . we refer to ( * ? ? ? 2.7 ) for these precise definitions in the general case . see also subsection [ slopes - mhgs ] . * cor . 4.17 ) [ sw ] the @xmath108characteristic variety of @xmath181 is given by : @xmath182 where @xmath183 is the closure of the conormal @xmath184 of the torus orbit @xmath185 . moreover , it is proved in ( * ? ? ? * lemma 3.14 ) that @xmath183 meets @xmath186 for all @xmath187 . thus theorem [ sw ] provides a description of the slopes of a hypergeometric system at the origin along any coordinate variety @xmath188 via considering the @xmath189parameter family of filtrations @xmath190 , @xmath141 , where @xmath124 is the @xmath124-filtration filtration with respect to the coordinate variety @xmath191 is defined by assigning the weight -1 ( resp . the weight 1 ) to the variables @xmath192 ( resp . @xmath193 ) for @xmath194 and the weight @xmath195 to the remaining variables . ] along @xmath66 : ( * ? ? ? 4.18 ) [ sw2 ] the real number @xmath146 is a slope of @xmath196 along @xmath66 at the origin if and only if @xmath197 is not locally constant at @xmath198 . [ remark - slope - at - any - p ] when @xmath188 is a coordinate hyperplane then the set of slopes of @xmath164 along @xmath66 at the origin coincides with the set of slopes of @xmath181 along @xmath66 at any point @xmath199 . this is proved in ( * ? ? ? 5.9 ) by using the comparison result in ( * ? ? ? 2.4.2 ) . in this subsection we extend some of the results from @xcite to the case of the gevrey solutions of a hypergeometric system at infinity in the direction of a coordinate hyperplane that we may assume to be @xmath200 . this construction is used later in the study of the gevrey solutions along @xmath92 of a modified hypergeometric system , see subsection [ gevrey - solutions - modulo - convergent - series ] . let us denote by @xmath201 the sheaf of holomorphic functions on @xmath160 . for @xmath202 , we denote by @xmath203 the formal completion of @xmath201 along @xmath66 , whose germs at @xmath204 are of the form @xmath205 where all the @xmath206 are holomorphic functions in a common neighborhood of @xmath207 . notice that the restriction of @xmath208 to @xmath66 , denoted by @xmath209 , is a subsheaf of @xmath210 . for any real number @xmath211 , we also consider the sheaf @xmath212 of gevrey series along @xmath66 of order @xmath211 which is defined to be the subsheaf of @xmath210 whose germs @xmath213 at @xmath204 satisfy @xmath214 we denote @xmath215 and use @xmath216 for the sheaf of gevrey series along @xmath66 of order less than @xmath211 . if @xmath213 belongs to @xmath217 for some @xmath211 , we say that the _ index _ of the gevrey series @xmath213 is @xmath211 . we also write @xmath218 . we denote by @xmath219 the sheaf of linear differential operators on @xmath62 with holomorphic coefficients . if @xmath220 is a coherent @xmath70-module , z. mebkhout has defined in ( * ? ? ? 6.3.1 ) the irregularity of order @xmath211 of @xmath220 with respect to @xmath66 to be the complex of ( sheaves of ) vector spaces @xmath221 and has proved that , for all @xmath222 $ ] , this complex is a perverse sheaf on @xmath66 when @xmath220 is holonomic ( * ? ? ? by the comparison theorem ( * ? ? * th.2.4.2 ) @xmath211 is a slope of @xmath220 with respect to @xmath66 if and only if @xmath211 is a gap in the filtration @xmath223 on the irregularity @xmath224 . the perversity result implies , in particular , that at a generic point @xmath199 only the first cohomology space of previous complexes is possibly non zero and thus it is worth studying the stalk @xmath225 . recall that @xmath226 is a full rank matrix with @xmath227 for all @xmath228 and @xmath229 . following @xcite and @xcite , for any vector @xmath230 we can define a series @xmath231_{u_{-}}}{[v+u]_{u_{+ } } } x^{v+u } \label{phiv}\ ] ] where @xmath230 verifies @xmath232 and @xmath233 . here @xmath234 , @xmath235 is the negative support of @xmath236 , @xmath237_{u}=\prod_{i } [ v_i ] _ { u_i } $ ] and @xmath238_{u_i}=\prod_{j=1}^{u_i } ( v_i -j + 1)$ ] is the pochhammer symbol for @xmath239 , @xmath240 . the series @xmath241 is annihilated by the hypergeometric ideal @xmath242 if and only if the negative support of @xmath243 is minimal , i.e. , @xmath244 with @xmath245 ( see ( * ? ? ? * section 3.4 ) ) . when @xmath21 is _ very generic _ , i.e. , when @xmath246 is not in a locally finite countable union of zarisky closed sets , there is a basis of the gevrey solution space of @xmath247 along @xmath66 given by series @xmath241 for suitable vectors @xmath230 ( see ( * ? ? ? 6.2 and 6.7 ) ) . for any subset @xmath248 we denote by @xmath249 the submatrix of @xmath0 given by the columns of @xmath0 indexed by @xmath250 and we denote @xmath251 . we say that @xmath252 is a @xmath253-_simplex _ with respect to @xmath0 ( or simply that @xmath178 is a @xmath253-simplex ) if the columns of @xmath254 determine a basis of @xmath161 . if it is so , we can reorder the variables in order to have @xmath255 without loss of generality . then a basis of @xmath256 associated with @xmath178 is given by the columns of the matrix : @xmath257 a vector @xmath230 satisfying @xmath258 for all @xmath259 and @xmath260 can be written as @xmath261 for some @xmath262 . since @xmath29 is very generic then the negative support of @xmath263 is the empty set and hence @xmath264 is annihilated by @xmath265 . moreover , the summation index set @xmath266 in the series @xmath264 is given by the integer vectors in an affine translate of the positive span of the columns of @xmath267 . the sum of the coordinates of the @xmath110-th column of @xmath267 is @xmath268 where @xmath269 means the sum of the coordinates . we have the following . * theorem 3.11)[theorem - gevrey - solutions - hypergeometric ] under the above conditions the series @xmath264 is a gevrey solution of @xmath181 along @xmath270 with index @xmath271 at points in certain relatively open subset of @xmath272 . in particular , if @xmath273 for all @xmath274 then @xmath264 is convergent . by corollary [ sw2 ] , a real number @xmath275 is a slope of @xmath276 along @xmath277 if and only if @xmath278 belongs to the hyperplane @xmath279 supported on a facet @xmath280 of the convex hull of @xmath281 such that @xmath282 . in particular , for any @xmath253-simplex @xmath283 it is easy to check that @xmath284 . we say in this case that @xmath178 is a @xmath253-simplex corresponding to the slope @xmath275 of @xmath196 along @xmath66 . the following theorem is a summary of some of the results from @xcite . its last statement uses results from @xcite and in @xcite . for the definition of a regular triangulation see , e.g. , ( * ? ? ? 8) . [ hypergeometric - gevrey - solutions ] assume that @xmath21 is very generic and that @xmath275 is a slope of @xmath196 along @xmath277 . for any @xmath253-simplex @xmath178 corresponding to @xmath211 one can construct @xmath285 many linearly independent gevrey solutions @xmath264 of @xmath286 along @xmath66 with index @xmath211 by varying @xmath287 in a set @xmath288 so that @xmath289 is a set of representatives of the group @xmath290 . moreover , if we repeat this construction for all the @xmath253-simplices @xmath291 corresponding to @xmath211 which belong to a suitable regular triangulation for @xmath0 and take the classes modulo @xmath292 we obtain a basis for the space of solutions of @xmath196 in the space @xmath293 for points @xmath199 in a relatively open set of @xmath66 . we have exhibited the construction of the gevrey solutions of @xmath294 along @xmath295 corresponding to each slope @xmath275 of @xmath196 along @xmath66 for @xmath29 very generic . let us construct gevrey solutions of @xmath196 at infinity . in other words , we are going to construct gevrey solutions of the projectivized hypergeometric system treated in ( * ? ? ? * section 5 ) at a generic point at infinity . we use the following notation : @xmath296 is @xmath297 with coordinates @xmath298 and @xmath299 so that @xmath300 . denote @xmath301 . take @xmath302 where @xmath124 is the @xmath124-filtration along @xmath66 . notice that @xmath303 is not locally constant at @xmath304 if and only if @xmath305 belongs to the hyperplane @xmath279 supported on a facet @xmath280 of the convex hull of @xmath281 such that @xmath306 . [ hypergeometric - gevrey - solutions - infinity ] assume that @xmath307 is very generic and that there exists @xmath275 such that @xmath305 belongs to a hyperplane @xmath279 as above . for any @xmath253-simplex @xmath308 one can construct @xmath285 many linearly independent gevrey series @xmath264 along @xmath309 with index @xmath211 by varying @xmath310 in a set @xmath288 so that @xmath311 is a set of representatives of the group @xmath312 . the classes of these series modulo convergent series @xmath313 are solutions of @xmath181 in @xmath314 . moreover , if we repeat this construction for all the @xmath253-simplices @xmath291 as above which belong to a suitable regular triangulation for @xmath0 and take the classes modulo @xmath315 then we obtain a basis for the space of solutions of @xmath196 in the space @xmath316 for points @xmath317 in a relatively open set of @xmath309 . _ proof.- _ by reordering the variables we may assume @xmath318 without loss of generality . let @xmath319 be the matrix @xmath267 but with the last column multiplied by @xmath320 . take a vector @xmath230 such that @xmath260 , @xmath258 for all @xmath321 and @xmath322 . it is clear that such a vector can be taken as @xmath323 , as before , for some @xmath324 . moreover , the summation index set @xmath266 in the series @xmath264 is given by the integer vectors in an affine translate of the positive span of the columns of @xmath325 . notice that the sum of the coordinates of the @xmath110-th column of @xmath325 is @xmath326 for @xmath327 while the sum of the coordinates of the last column is @xmath328 . this implies that @xmath329 is a gevrey series along @xmath309 with order @xmath146 . to prove that @xmath211 is the gevrey index of @xmath330 one can use a slightly modified version of lemma 3.8 in @xcite . moreover , since @xmath29 is very generic then the negative support of @xmath263 is @xmath331 which is not minimal and hence @xmath264 is not annihilated by @xmath265 ( see ( * ? ? ? * section 3.4 ) ) . however , it can be checked that for any differential operator @xmath332 the series @xmath333 is either zero or a polynomial in @xmath334 with coefficients that are convergent power series in the variables @xmath335 . now we finish the proof following ( * ? ? ? * sec . 7 ) and ( * ? ? ? [ remark - gevrey - solutions - infinity ] a slightly weaker version of the first paragraph of theorem [ hypergeometric - gevrey - solutions - infinity ] can also be proved without the very genericity assumption in @xmath21 . more precisely , for all @xmath21 if @xmath303 is not locally constant at @xmath304 we can construct a gevrey series along @xmath309 of index @xmath149 which is a solution of @xmath196 modulo the space @xmath315 ( of gevrey series along @xmath309 with order less than @xmath211 ) by methods similar to the ones in ( * ? ? ? * section 4 ) . the following corollary is a particular case of ( * ? ? ? * conjecture 5.18 ) . [ slopes - hypergeometric - at - infinity ] the real number @xmath275 is a slope of @xmath196 along @xmath336 if and only if @xmath303 is not locally constant at @xmath337 . _ proof.- _ for the only if direction of the proof we refer to ( * ? ? ? * section 5 ) . let us prove the if direction . by theorem [ hypergeometric - gevrey - solutions - infinity ] and remark [ remark - gevrey - solutions - infinity ] one can construct a gevrey series of index @xmath149 that is a solution of @xmath181 along @xmath309 ( modulo @xmath338 ) . so , the result follows from the comparison theorem for the slopes @xcite . [ example - slope - infinity ] set @xmath339 and @xmath340 . we have that the kernel of @xmath0 is generated by @xmath341 and thus the hypergeometric ideal @xmath265 is generated by @xmath342 , @xmath343 and @xmath344 . take @xmath345 and notice that for @xmath346 we have that @xmath347 belongs to the line @xmath348 determined by @xmath349 and @xmath350 . by corollary [ slopes - hypergeometric - at - infinity ] we have that @xmath346 is a slope of @xmath196 along @xmath351 . indeed , if we consider @xmath352 for @xmath353 we have that @xmath354 is a gevrey series along @xmath309 of index @xmath346 if @xmath355 . moreover , we have that @xmath356 for @xmath357 . for each of the three biggest @xmath358 verifying that @xmath359 we have that @xmath360 , which is convergent at any @xmath361 . the classes modulo @xmath313 of these three series @xmath354 form a basis for the space of solutions of @xmath196 in @xmath362 , @xmath361 . if @xmath29 is very generic then @xmath363 . let @xmath44 be the weyl algebra @xmath97 . the variable @xmath100 is also denoted by @xmath101 and @xmath102 by @xmath103 . let @xmath104 be a linear form @xmath105 such that @xmath106 for @xmath107 inducing the so - called @xmath108filtration on the ring @xmath44 . if @xmath109 for all @xmath110 , the associated graded ring @xmath111 is isomorphic to a polynomial ring in @xmath112 variables @xmath364 with complex coefficients . this polynomial ring is @xmath108-graded , the @xmath108-degree of a monomial @xmath114 being @xmath115 . if we need to emphasize the coefficients of the linear form we will simply write @xmath116 for @xmath117 with @xmath106 for all @xmath110 . if @xmath118 and @xmath119 then the corresponding @xmath120 filtration is nothing but the usual order filtration on @xmath44 ( which is also called the @xmath121-filtration ) . if @xmath122 and @xmath123 then the corresponding @xmath120 filtration is nothing but the malgrange - kashiwara filtration on @xmath44 ( we also say the @xmath124-filtration ) with respect to @xmath91 . in the remainder of this section we will assume @xmath109 for all @xmath110 ; we will say then that @xmath125 is a weight vector for the weyl algebra . we define the ring isomorphism @xmath365 of @xmath44 by @xmath366 , @xmath367 . the isomorphism @xmath365 is called the _ fourier transform _ on @xmath44 with respect to the variable @xmath100 . the inverse transform @xmath368 is given by @xmath369 , @xmath83 . let @xmath125 be a weight vector for the weyl algebra . the fourier transforms @xmath365 and @xmath368 induce isomorphisms in @xmath370 and we denote them also by @xmath365 and @xmath368 respectively . analogously , if @xmath371 is the affine algebraic set defined by an ideal @xmath372 we write @xmath373 and @xmath374 for the algebraic set defined by @xmath375 and @xmath376 respectively . we define the fourier transform of the weight vector @xmath125 by the formula @xmath377 we notice that @xmath378 . we will also write @xmath379 if @xmath116 . [ finf(u , v)fp ] for any operator @xmath380 , we have @xmath381 _ proof_. we prove it in the case @xmath382 . other cases can be reduced to this case . we put @xmath383 . we assume that @xmath384 and @xmath385 . then , we have @xmath386 . since @xmath387 and @xmath388 , we have @xmath389 . applying the inverse fourier transform , we obtain @xmath390 . suppose that @xmath391 . then , we have @xmath392 . thus , we obtain the conclusion . proposition [ finf(u , v)fp ] yields the following simple , but important claim for the grbner deformation method . [ corollary : fourier ] for any left ideal @xmath393 in @xmath44 , we have @xmath394 _ proof_. since @xmath395 is spanned by @xmath396 , @xmath397 as a @xmath398-vector space , the conclusion follows from the previous proposition . we retain the notations of @xcite , which are explained in the introduction . we are interested in the slopes of the modified system @xmath399 at any point along @xmath400 . we notice that @xmath36 is the fourier transform of @xmath401 where @xmath402 , i.e. @xmath403 recall that @xmath404 where @xmath405 and @xmath406 . thus @xmath407 . in order to obtain the slopes of the modified system @xmath48 , we need to study the @xmath133initial ideal of the modified hypergeometric ideal @xmath36 , for @xmath408 . therefore , applying corollary [ corollary : fourier ] and ( [ fouriermgkz = gkz ] ) , we have @xmath409 using ( [ fourier2 ] ) , we have @xmath410 on the other hand , since @xmath411 is a hypergeometric system and the matrix @xmath80 is pointed , theorem [ sw ] gives a description of the irreducible components of @xmath412 in terms of the @xmath413umbrella . we notice here that the last coordinate of @xmath414 equals @xmath415 . we recall the definition ( * ? ? ? 2.7 ) of the umbrella in this case . first of all let us recall _ loc . , _ that if @xmath416 , @xmath417 are two points and @xmath418 is a hyperplane in @xmath419 containing neither @xmath416 nor @xmath417 , then the convex hull @xmath420 of @xmath416 and @xmath417 relative to @xmath418 , is the unique line segment joining @xmath416 and @xmath417 and not meeting @xmath418 . let us denote the @xmath421 column of @xmath80 by @xmath422 , say @xmath423 for @xmath424 and @xmath425 . for simplicity let us write @xmath426 instead of @xmath413 . we view @xmath427 as points in @xmath428 . as @xmath429 is pointed , there exists a linear form @xmath430 on @xmath431 such that @xmath432 for all @xmath110 . let @xmath433 be such that @xmath434 for @xmath163 and @xmath435 . 2.7 ) [ umbrella ] the @xmath426polyhedron @xmath436 is the convex hull @xmath437 where @xmath438 is the projective closure of the affine hyperplane @xmath439 . the @xmath426-umbrella @xmath440 is the set of faces of @xmath436 which do not contain the origin . in particular , @xmath440 contains the empty face . figure 1 shows three @xmath413umbrellas for @xmath441 and @xmath442 . in each case the shaded region is the polyhedron @xmath436 . for @xmath443 the point @xmath444 belongs to the line passing through @xmath445 and @xmath446 which means that @xmath447 is a slope of the system along @xmath448 . using definition [ def : slope ] , equation and ( * ? ? ? 4.12 ) we get the following : [ fouriersw ] the real number @xmath146 is a slope of @xmath48 along @xmath92 at @xmath147 if and only if @xmath449 is not locally constant at @xmath198 . let us denote by @xmath450 the matrix with columns @xmath451 , @xmath452 , and let @xmath453 be the convex hull of @xmath454 and the origin . with this notation , corollary [ fouriersw ] can be rephrased as follows . [ fouriersw2 ] the real number @xmath146 is a slope of @xmath455 along @xmath92 at @xmath147 if and only if there exists a facet @xmath280 of @xmath456 such that @xmath282 and @xmath457 , where @xmath279 is the hyperplane that contains @xmath280 . [ remark - slopes - modify - infinity ] we note that the @xmath44-modules @xmath48 and @xmath458 agree on @xmath459 under the change of the variable @xmath460 in @xmath461 . we regard both @xmath44-modules as extensions of each other . in this paper , when we say the irregularity of @xmath458 along @xmath462 , it means the irregularity of @xmath48 along @xmath400 . using this terminology , corollary [ fouriersw2 ] provides a description of the slopes of the modified system @xmath48 along @xmath463 by using @xmath464 instead of @xmath456 . until the end of this section we will denote either @xmath465 or @xmath466 for the @xmath467umbrella with @xmath116 since definition [ umbrella ] does not depend on @xmath108 but only on @xmath468 moreover , for any subset @xmath469 we denote by @xmath470 the vector with coordinates equal to the ones of @xmath4 indexed by @xmath250 , i.e. , @xmath471 . in the following two lemmas we assume for simplicity that @xmath472 , @xmath163 . in fact , this can be assumed without loss of generality since @xmath80 is pointed . [ lemma - umbrellas-1 ] for any sufficiently small real number @xmath141 , we have that @xmath473 in particular , if all the facets in @xmath474 contain exactly @xmath10 columns of @xmath0 , then the set of facets of @xmath475 which contain @xmath476 is @xmath477 . _ proof_. take @xmath478 such that @xmath479 and set @xmath480 . by definition [ umbrella ] , @xmath481 if and only if there is a ( unique ) vector @xmath482 ( which depends on @xmath483 ) such that @xmath484 if @xmath485 , @xmath486 and @xmath487 if @xmath488 . the equation for @xmath489 is equivalent to @xmath490 and the other equalities and inequalities can then be written as @xmath491 for @xmath492 and @xmath493 for @xmath494 . thus , denoting @xmath495 the limit of the vector @xmath496 when @xmath483 tends to zero , we have that @xmath497 for @xmath492 and @xmath498 for @xmath499 . this proves that @xmath500 . it follows for @xmath501 that @xmath502 for @xmath503 and @xmath504 for @xmath505 . this proves that @xmath506 . this proves the inclusion @xmath507 in the statement . the other inclusion can be proved in a similar way , now starting from the existence of vectors @xmath508 and @xmath430 corresponding to @xmath509 and @xmath510 as above and considering @xmath511 with @xmath512 . [ lemma - umbrellas-2 ] for any sufficiently large real number @xmath141 we have that @xmath513 in particular , if all the facets in @xmath514 contain exactly @xmath10 columns of @xmath0 , then @xmath515 . _ proof_. recall that @xmath516 if and only if there is a ( unique ) vector @xmath517 ( which could depend on @xmath483 ) verifying the equalities @xmath518 for @xmath519 and the inequalities @xmath520 for @xmath521 . assume to the contrary that there is a facet @xmath522 such that @xmath523 for all @xmath524 and @xmath525 big enough . since @xmath526 and @xmath527 we have that the corresponding @xmath528 is independent of @xmath529 and @xmath530 which is a contradiction . thus , any facet of @xmath531 contains @xmath476 . if we write @xmath532 for @xmath533 then @xmath534 if and only if there is a unique vector @xmath482 as above ( now depending on @xmath483 ) . in particular , @xmath535 and using the limit @xmath536 of the vector @xmath537 when @xmath483 tends to infinity we get that @xmath538 for @xmath492 and @xmath539 for @xmath540 . thus , @xmath541 . moreover , we obtain that @xmath542 by using the vector @xmath543 . the other inclusion can be proved in a similar way , now starting from the existence of vectors @xmath536 and @xmath430 corresponding to @xmath544 and @xmath545 as above and considering @xmath546 . [ regularity - along - t ] the following conditions are equivalent : 1 . @xmath548 is constant for all @xmath141 . 3 . @xmath549 does not have slopes along @xmath92 . _ proof_. since any umbrella is determined by its facets , lemma [ lemma - umbrellas-1 ] and lemma [ lemma - umbrellas-2 ] prove that ( a ) is equivalent to ( b ) . corollary [ fouriersw ] finishes the proof . note that condition ( a ) in proposition [ regularity - along - t ] implies that there is a common refinement ( the one given by considering the faces of @xmath548 not containing @xmath476 ) of the polyhedral complex subdivisions induced by the umbrellas @xmath550 and @xmath514 . in particular , when @xmath4 is generic , condition ( a ) means that @xmath4 induces a regular triangulation of @xmath0 that refines the polyhedral complex subdivision induced by the weight vector @xmath551 . in other words , @xmath4 is a perturbation of @xmath551 for the matrix @xmath0 . for example , this condition is satisfied if either the row span of the matrix @xmath0 contains the vector @xmath552 or @xmath4 is the sum of a vector in the row span of the matrix @xmath0 and a vector @xmath553 with @xmath554 . even when @xmath411 is regular holonomic and @xmath555 has no slopes along @xmath92 , the latter system is irregular if @xmath196 is ( see example [ curious - example ] ) . [ curious - example ] take @xmath556 and @xmath557 . the system @xmath181 is irregular along @xmath558 with unique slope @xmath559 . if we choose @xmath560 and consider the matrix @xmath561 we have that the hypergeometric system @xmath562 is regular holonomic for all @xmath563 and @xmath31 by a well known result of hotta @xcite . however , the modified system @xmath564 has a slope @xmath559 along @xmath463 because @xmath565 has the slope @xmath566 along @xmath92 ( see corollary [ fouriersw2 ] and remark [ remark - slopes - modify - infinity ] ) . let @xmath55 be the map ( [ eq : gr - deformation ] ) defined in the introduction . since the @xmath70-modules @xmath567 and @xmath63 are isomorphic when restricted to @xmath568 , the slopes of both modules along any coordinate subspace @xmath272 not contained in @xmath92 coincide in @xmath569 . moreover , the map @xmath570 also induces an isomorphism for their spaces of gevrey solutions along @xmath272 . we study convergent and formal power series solutions of the modified @xmath0-hypergeometric module @xmath63 . as was said before , the map @xmath55 ( [ eq : gr - deformation ] ) induces an isomorphism between the @xmath70-modules @xmath567 and @xmath63 when restricted to @xmath571 , and also an isomorphism between their corresponding spaces of holomorphic solutions . more precisely , for any germ of a holomorphic function @xmath572 at a point @xmath573 in @xmath571 , the function @xmath572 is a solution of @xmath63 if and only if @xmath574 is a germ of a solution of @xmath575 at the point @xmath576 such that @xmath577 . we can rewrite this as follows : the morphism @xmath578 is an isomorphism of sheaves of vector spaces . as a consequence the holonomic ranks of both modules coincide @xmath579 and this last rank equals the one of @xmath6 for any @xmath4 , see ( * ? ? * theorem 1 ) . recall that if @xmath29 is generic then @xmath580 , where @xmath581 is the normalized volume of the matrix @xmath0 @xcite , while in general @xmath582 @xcite . we describe the solutions of @xmath63 in the space @xmath583 of formal power series with respect to @xmath584 . more generally , we also describe the solutions of @xmath63 in the space @xmath585 for any finite set @xmath586 ( see theorem [ theorem - dimension-1 ] ) . we will use notations in @xcite . let @xmath280 be the weight vector @xmath587 inducing the malgrange - kashiwara @xmath124filtration along @xmath588 on the ring @xmath45 . let @xmath589:={{\mathbb c}}[\partial_1 , \ldots , \partial_n , t]$ ] be the toric ideal associated with @xmath80 , i.e. , the binomial ideal generated by the operators in ( [ eq : mtoric ] ) . [ inn_tau - inn_w ] for all @xmath590 we have @xmath591 { { \rm in}}_w(i_a)$ ] . _ proof_. recall that @xmath592 and we can write @xmath593 notice that @xmath594 if @xmath595 and @xmath596 and that @xmath597 if @xmath595 and @xmath598 . the conclusion follows by a straightforward groebner basis argument because a groebner basis of @xmath599 with respect to @xmath600 ( resp . of @xmath601 with respect to @xmath4 ) is given by a set of binomials with the same form as the ones defining the ideal . recall that the _ indicial polynomial _ ( also called @xmath417-_function _ ) of @xmath602 along @xmath92 is the polynomial @xmath603 $ ] such that @xmath604 is the monic generator of @xmath605 $ ] where @xmath606 . moreover , we have by ( * ? ? ? 3 ) that for @xmath29 and @xmath4 generic , the indicial polynomial of @xmath607 along @xmath92 is @xmath608 where @xmath609 , @xmath610 is the set of _ top - dimensional standard pairs _ of @xmath127 ( see ( * ? ? ? 3.2 ) ) and @xmath611 is the vector defined as @xmath612 for @xmath613 and @xmath260 , which is also an exponent of @xmath6 with respect to @xmath4 ( see ( * ? ? ? * lemma 4.1.3 ) ) . [ perturbation ] we say that a ( generic ) vector @xmath614 is a ( generic ) perturbation of @xmath615 , with respect to @xmath0 , if there exists @xmath616 such that @xmath617 . [ remark - number - of - exponents ] if @xmath618 is generic then @xmath619 is a monomial ideal and it is well known that its degree equals the cardinality of its set of top - dimensional standard pairs @xmath620 . moreover , for very generic @xmath307 there are exactly @xmath621 many exponents of @xmath265 with respect to @xmath618 ; see ( * ? ? ? 3.4 ) and ( * ? ? ? 4.10 ) . [ formal - solution - w - perturbation ] let @xmath21 be very generic and @xmath622 . there is a generic perturbation @xmath614 of @xmath4 such that for any exponent @xmath623 of @xmath265 with respect to @xmath618 the series @xmath624 , for @xmath625 , is a solution of @xmath63 of the form @xmath626 , with @xmath627 and @xmath628 for some @xmath629 . _ _ since @xmath29 is very generic , for any generic @xmath618 an exponent @xmath243 of @xmath265 with respect to @xmath618 can be written as @xmath630 where @xmath631 is a top - dimensional standard pair of @xmath619 , see ( * ? ? ? 3.4 ) . in particular , @xmath632 , @xmath633 , @xmath260 and @xmath612 for all @xmath613 . the series @xmath634 is either a holomorphic solution or a gevrey solution of @xmath247 along a coordinate subspace @xmath635 at any point @xmath207 in a non empty relatively open set @xmath636 in @xmath272 ( see theorem [ theorem - gevrey - solutions - hypergeometric ] and ( * ? ? ? 3.11 ) for the details ) and since @xmath29 is very generic we have that @xmath637 . the expression @xmath638 ( resp . @xmath639 ) formally satisfies the equations defining @xmath640 ( resp . @xmath641 ) . we will prove that we can write @xmath642 and that it is a gevrey series along @xmath643 . recalling the expression of @xmath241 ( [ phiv ] ) it is enough to prove that for all @xmath644 we have @xmath645 and that the coefficient of @xmath646 in @xmath213 , i. e. @xmath647_{u_{-}}}{[v+u]_{u_{+ } } } x^{v+u},\ ] ] is a convergent series in an open neighborhood of some @xmath629 , both the neighborhood and @xmath207 independent of @xmath648 . we can take a generic perturbation @xmath614 of @xmath4 of the form @xmath649 with @xmath650 for @xmath153 and @xmath651 small enough and @xmath652 is generic . take any @xmath653 and let us prove that @xmath654 . since @xmath243 is an exponent of @xmath265 with respect to @xmath618 we have by ( * ? ? ? * ( 3.30 ) ) that @xmath655 . hence , since last inequality holds for @xmath153 and @xmath656 small enough we have that @xmath657 . thus we have that @xmath657 for all @xmath658 . notice that when @xmath659 for all @xmath653 then the set @xmath660 is finite and hence @xmath661 is clearly convergent . in general , @xmath660 is not finite , but we will see that @xmath661 is still convergent at some point @xmath629 . since @xmath637 the set @xmath662 is a finite union of shifted copies of the form @xmath663 where @xmath664 is the set of columns of @xmath267 and @xmath665 satisfies @xmath666 . the series @xmath661 is convergent if and only if all the series @xmath667_{u_{-}}}{[v+u]_{u_{+ } } } x^{v+u}$ ] are convergent . since @xmath29 is very generic , @xmath668 , and the convergence of each series @xmath669 is equivalent to the convergence of the series @xmath670 . thus , it is enough to see that for any column @xmath671 of @xmath267 such that @xmath672 we have that @xmath673 . notice that @xmath674 implies @xmath675 and so @xmath676 . hence , since this holds for @xmath677 small enough , we have that @xmath678 . we have proved that @xmath213 is a formal solution of @xmath679 along @xmath92 and it is clear that @xmath680 . from the expression of the @xmath669 and ( * ? ? ? * th . 3.11 ) any @xmath681 is convergent at any point in @xmath682 for some @xmath683 . let @xmath614 be a generic perturbation of @xmath615 as in the proof of lemma [ formal - solution - w - perturbation ] . [ f_m ] if @xmath684 is a solution of @xmath63 for some @xmath685 , @xmath147 , with @xmath686 , then : 1 . @xmath687 is a solution of @xmath688 . 2 . for all @xmath689 , @xmath681 is a holomorphic solution of @xmath690 , where this last module is the hypergeometric system associated with the matrix @xmath450 and the parameter @xmath691 . @xmath692 , where @xmath693 is the indicial polynomial of @xmath607 along @xmath92 . if @xmath29 is very generic then @xmath694 for some exponent @xmath243 of @xmath265 with respect to @xmath618 . _ the proof of ( a ) and ( b ) are straightforward . let us prove ( c ) . by ( a ) and using ( * ? ? ? * theorem 2.5.5 ) we have that @xmath695 is a solution of @xmath696 . recall by definition of @xmath693 that @xmath697 $ ] where @xmath698 . thus , the differential operator @xmath604 annihilates @xmath695 . this implies , using for example ( * ? ? ? * lemma 1.3.2 ) , that @xmath699 and this implies that @xmath700 . let us prove ( d ) . by ( b ) we have that @xmath701 is a holomorphic solution of @xmath702 and thus it can be written as a nilsson series at the origin with respect to a vector @xmath703 that is a perturbation of @xmath704 ( see @xcite , @xcite , @xcite ) and , in particular , it makes sense to consider the initial form of @xmath701 with respect to @xmath703 . on the other hand , using lemma [ inn_tau - inn_w ] , we have that @xmath705 annihilates @xmath701 . since @xmath29 is very generic , @xmath706 ( * ? ? ? this implies that @xmath701 is a solution of @xmath707 and hence , @xmath708 is a solution of @xmath709 for @xmath649 . thus , since @xmath29 is very generic @xmath710 for @xmath711 and @xmath243 an exponent of @xmath12 with respect to @xmath618 . hence , using ( b ) , we also have that @xmath712 . although we assume in this paper that @xmath80 is pointed it turns out that in this section this fact is only used in the proof of ( d ) in lemma [ f_m ] . however , let us notice that if @xmath80 is not pointed and @xmath4 is generic then @xmath713 $ ] , @xmath714 and so @xmath715 . thus , by ( c ) in lemma [ f_m ] the modified system @xmath63 does not have any solution in @xmath716 for all @xmath685 and @xmath147 . [ degree - w - initial - toric ] since @xmath649 with @xmath717 for sufficiently small @xmath153 and @xmath651 we have that @xmath718 for @xmath719 . in particular , @xmath720 and @xmath721 have the same degree . let us denote by @xmath722 the dimension of the space of @xmath723-solutions of a @xmath44-module @xmath127 at a point @xmath207 . [ theorem - dimension-1 ] assume @xmath11 is very generic , @xmath615 and @xmath94 . then @xmath724 if @xmath725 for all the exponents @xmath243 of @xmath726 with respect to @xmath618 . we also have that @xmath727 in particular , if @xmath4 is generic we also have @xmath728 for all generic @xmath147 , @xmath729 and any exponent @xmath243 of @xmath265 with respect to @xmath4 . _ proof_. the first statement follows from lemma [ f_m ] , ( d ) . the inequality @xmath730 follows from lemma [ formal - solution - w - perturbation ] , remark [ remark - number - of - exponents ] and remark [ degree - w - initial - toric ] . on the other hand , since the differential operators defining @xmath63 belong to the weyl algebra we have that any solution @xmath731 of @xmath63 decomposes as a finite sum of solutions , each of them in a space @xmath732 . recall by the proof of lemma [ f_m ] that any solution @xmath733 of @xmath63 verifies that @xmath734 for an exponent of @xmath12 with respect to @xmath618 . this last fact , remark [ remark - number - of - exponents ] , remark [ degree - w - initial - toric ] and a slightly modified version of ( * ? ? ? * proposition 2.5.7 ) prove that @xmath735 . finally , the last statement follows from the second statement and from the fact that if @xmath29 is very generic , @xmath736 is generic and @xmath243 and @xmath737 are two different exponents of @xmath265 with respect to @xmath4 , then @xmath738 . [ remark3 ] let @xmath639 be the series constructed in lemma [ formal - solution - w - perturbation ] and used in theorem [ theorem - dimension-1 ] . if @xmath4 is in the row span of @xmath0 then @xmath739 does not depend on @xmath100 and thus it is a convergent series . if @xmath4 is not in the row span of @xmath0 then @xmath740 is gevrey along @xmath92 with index @xmath149 where @xmath741 where @xmath742 . on the other hand , as mentioned in the proof of lemma [ formal - solution - w - perturbation ] since @xmath29 is very generic and @xmath243 is an exponent of @xmath265 with respect to @xmath618 then @xmath243 is associated with a simplex @xmath743 and there is a basis @xmath744 of the kernel of @xmath0 such that for all @xmath613 , @xmath745 for all @xmath746 and @xmath747 . the set @xmath744 is the set of columns of @xmath267 ( if we reorder the variables so that @xmath748 ) and in this case we have that @xmath637 . thus , more explicitly , @xmath749 where @xmath750 is the set of columns of @xmath267 , @xmath751 and @xmath752 . the proof of this formula is technical and follows from standard estimates on gamma functions similar to the ones used in @xcite to compute the index of gevrey solutions for hypergeometric systems . in particular , if @xmath4 is a perturbation of @xmath551 then @xmath483 is close to @xmath320 and if @xmath0 is homogeneous then @xmath753 because @xmath754 for any @xmath658 and hence in both cases the series is convergent . by theorem [ theorem - dimension-1 ] , if both @xmath755 and @xmath21 are very generic then @xmath756 does not have any nonzero solution in @xmath757 for all @xmath147 . this is in contrast with the case of the irregularity of hypergeometric systems along coordinate hyperplanes , where for any slope @xmath149 of @xmath758 along @xmath202 and for very generic @xmath307 one can construct a formal solution @xmath759 of @xmath760 along @xmath66 , such that @xmath761 has gevrey index equal to the slope ( see @xcite and theorem [ hypergeometric - gevrey - solutions ] ) . however , by the comparison theorem for the slopes @xcite and the perversity of the irregularity complex of a holonomic @xmath70-module along a smooth hypersurface @xcite , one knows that for each slope @xmath149 of @xmath762 along @xmath92 at a generic @xmath147 there must exist a formal series @xmath763 with gevrey index @xmath149 such that @xmath764 is convergent at @xmath207 for all @xmath765 . the purpose of this section is to describe gevrey solutions modulo convergent series of the modified system @xmath766 along @xmath92 when @xmath31 and @xmath21 are very generic . to this end , we use the construction of the gevrey solutions at infinity of the hypergeometric system @xmath767 as performed in theorem [ hypergeometric - gevrey - solutions - infinity ] . take @xmath768 with coordinates @xmath769 and @xmath770 so that @xmath771 . denote @xmath772 . we can consider for any @xmath685 the @xmath773linear map @xmath774_{m } t^{-1-\gamma -m } \end{array}\ ] ] where @xmath775 and @xmath776 . [ upsilon - morphism ] notice that @xmath777 is an isomorphism if and only if @xmath778 . in such a case we also have that @xmath779 for all @xmath211 . it is also clear that @xmath780_k \upsilon_{k}(\sum_{m\geq 0 } f_{m}(x ) t^{k + m})$ ] for all @xmath781 . [ formal - modulo - convergent ] assume @xmath31 and @xmath21 to be very generic . if @xmath146 is a slope of @xmath782 along @xmath92 then we can construct @xmath783 gevrey series that are linearly independent solutions of @xmath782 modulo convergent series and whose gevrey index is equal to @xmath149 . here @xmath280 runs over all the facets of @xmath453 such that @xmath784 and @xmath306 . moreover , the classes modulo @xmath785 of these gevrey series form a basis of the solution space of @xmath786 in @xmath787 for points @xmath147 in a relatively open set of @xmath92 . _ proof_. the existence of such facets @xmath280 is given by corollary [ fouriersw2 ] . since @xmath784 and @xmath788 for @xmath789 we have that @xmath790 is a slope of @xmath791 along @xmath462 . thus , by theorem [ hypergeometric - gevrey - solutions - infinity ] we can construct @xmath792 gevrey series along @xmath793 with index @xmath794 . moreover , the classes in @xmath795 are linearly independent solutions of @xmath791 . more precisely , for any @xmath10-simplex @xmath283 the series constructed are of the form @xmath796 for @xmath797 with @xmath260 , @xmath798 ( i. e. @xmath799 ) and @xmath258 for all @xmath800 . using remark [ upsilon - morphism ] we can take @xmath801 as the unique gevrey series along @xmath92 with index @xmath149 verifying @xmath802 for @xmath797 . we conclude by remark [ upsilon - morphism ] that the @xmath803 series @xmath801 constructed are gevrey series with index @xmath804 whose classes modulo @xmath805 are linearly independent . moreover , it can be checked that they are solutions of the modified system modulo @xmath806 by using the fact that their images by the morphism @xmath807 are solutions of @xmath791 . last statement follows from ( [ fouriermgkz = gkz ] ) , @xcite and @xcite . take @xmath808 , @xmath809 and @xmath810 . we have that @xmath811 and @xmath812 . note here that @xmath813 is the binomial ideal generated by the operators in ( [ eq : mtoric ] ) . the unique slope of @xmath814 along @xmath92 is @xmath815 since @xmath816 belongs to the line passing through @xmath817 and @xmath818 is a facet of @xmath819 if and only if @xmath820 . the volume of @xmath178 is one and following the proofs of theorem [ formal - modulo - convergent ] and theorem [ hypergeometric - gevrey - solutions - infinity ] we can take @xmath821 which satisfies the conditions @xmath822 , @xmath823 and @xmath824 . we get that the series @xmath825_{3 m_2 + 5 m_3}}{[\alpha + m_3]_{m_3 } m_2 ! } x_1^{\beta -5 \alpha -3m_2 -5 m_3 } x_2^{m_2 } x_3^{\alpha + m_3 } t^{m_2 + m_3}\ ] ] is a gevrey solution ( modulo convergent series ) of @xmath826 along @xmath92 with index @xmath815 . we assume that the @xmath827-row span of the matrix @xmath0 does not contain the vector @xmath828 , but the one of @xmath450 does , where @xmath450 is the matrix with columns @xmath451 , @xmath829 ( see subsection [ slopes - mhgs ] ) . this case holds if and only if the weight vector @xmath4 is in the image of @xmath830 where @xmath831 is the matrix @xmath832 and it is not in the row span of @xmath0 . in other words , the weight vector @xmath4 lies in the intersection of the set of the secondary cones of @xmath831 ; see ( * ? ? ? solutions of this case can be analyzed by utilizing the borel transformation and the laplace transformation . we review here some basics of the borel summation method which we require in the following ( see @xcite for the details ) . let us consider the formal expression @xmath833\ ] ] where @xmath834 and @xmath835 . solutions constructed in theorem [ theorem - dimension-1 ] are of this form . if its coefficients satisfy @xmath836 with some positive constants @xmath837 , and @xmath838 ( in this last condition will be relaxed ) , then the formal borel transform ( with index @xmath839 ) defined by @xmath840(\tau ) : = \sum_{\ell = 0}^{\infty } \frac{f_{\ell}}{\gamma(1 + ( \ell + \gamma ) /\kappa ) } \tau^{\ell + \gamma}\ ] ] is the product of @xmath841 and a convergent power series at @xmath842 . in addition to , if * the function @xmath843 $ ] can be analytically continued to a sector @xmath844 of infinite radius in a direction @xmath845 with an opening angle @xmath846 , and * the analytic continuation of @xmath843 $ ] satisfies the growth estimate @xmath847(\tau)\big\| \leq c_1 \exp\big[c_2 \big\|\tau\big\|^{\kappa}\big]\ ] ] in @xmath848 with some positive constants @xmath849 , then we say @xmath213 is @xmath839-summable in the direction @xmath850 , and define the @xmath839-sum ( or the borel sum with index @xmath839 ) of @xmath213 by the laplace transformation @xmath851(t ) = \mathcal{l}^{\theta}_{\kappa}\hat{\mathcal{b}}_{\kappa}[f](t ) : = \int^{e^{i\theta}\cdot \infty}_0 e^{-(\tau / t)^\kappa } \hat{\mathcal{b}}_{\kappa}[f](\tau ) d(\tau / t)^{\kappa},\ ] ] where @xmath852 [ remak(a ) ] because of the growth condition ( ii ) of @xmath843 $ ] , the laplace integral converges if @xmath100 satisfies @xmath853 - c_2 \|\tau\|^{\kappa } > 0.\ ] ] since @xmath854 - c_2 \|\tau\|^{\kappa } = \left\|\frac{\tau}{t}\right\|^{\kappa } \big\ { \cos \kappa(\theta - \arg t ) - c_2 \|t\|^{\kappa}\big\}\ ] ] ( note that @xmath855 ) , the laplace integral converges in @xmath856 > c_2 \|t\|^{\kappa}\}.\ ] ] the region has infinitely many connected components . here and in what follows we specify one of them by imposing @xmath857 . since we can vary @xmath858 in slightly , we conclude that @xmath859 $ ] defines a holomorphic function in @xmath860 > c_2 \|t\|^{\kappa},\ , @xmath861 and @xmath862 such that the @xmath859 $ ] is holomorphic in @xmath863 ( cf . * the first paragraph of 2.1 ) ) . [ old - remark-(b ) ] if @xmath213 is @xmath839-summable in a direction @xmath850 , then @xmath213 is a gevrey asymptotic expansion of its borel sum @xmath864(t)$ ] : for any closed subsector @xmath865 of @xmath866 , there exists @xmath867 for which the inequality @xmath868(t ) - \sum_{\ell = 0}^{n-1 } f_{\ell } \ , t^{\ell } \right\| \leq c ' ( k')^n \big\|t\big\|^n \gamma(1 + n/\kappa)\ ] ] holds in @xmath865 for @xmath869 . _ proof_. here we give a sketch of the proof . see ( * ? ? ? 1 ) for the details . it follows from the relation @xmath870)\ ] ] that the remainder of the expansion becomes @xmath871(t ) - \sum_{\ell = 0}^{n-1 } f_{\ell } \ , t^{\ell + \gamma}\label{old - remark-(b)-key - relation}\\ & \notag \qquad = \int^{e^{i\theta}\cdot \infty}_0 e^{-(\tau / t)^\kappa } \left\ { \hat{\mathcal{b}}_{\kappa}[f ] ( \tau ) - \sum_{\ell = 0}^{n-1 } \frac{f_{\ell}}{\gamma(1 + ( \ell + \gamma)/\kappa ) } \ , \tau^{\ell + \gamma } \right\ } \dfrac{\kappa\tau^{\kappa-1}}{t^{\kappa } } d\tau.\end{aligned}\ ] ] since @xmath872 ( \tau ) - \sum_{\ell = 0}^{n-1 } \frac{f_{\ell}}{\gamma(1 + ( \ell + \gamma)/\kappa ) } \ , \tau^{\ell + \gamma } \right\}\ ] ] is holomorphic in some neighborhood of @xmath865 ( which includes the origin ) and satisfies the growth condition in @xmath865 with appropriate constants @xmath873 and @xmath874 , the righthand side of can be estimated by the righthand side of multiplied by @xmath875 . inequality is also known to be equivalent to conditions ( i ) and ( ii ) stated above ( ( * ? ? ? * theorem 1 ( p. 23 ) ) ) . the borel summation method may be also applied to the case when @xmath876 we can define the borel transform @xmath843 $ ] in the same manner as before . in this last case , however , the laplace integral may not converge at @xmath842 . therefore we modify the definition of the borel sum to @xmath877(t ) = \mathcal{l}^{\theta}_{\kappa}\hat{\mathcal{b}}_{\kappa}[f](t ) : = \frac{1}{1 - e^{- 2\pi i \gamma/\kappa } } \int_{\gamma_{\kappa\theta } } e^{-\zeta / t^\kappa } \hat{\mathcal{b}}_{\kappa}[f](\zeta^{1/\kappa } ) \dfrac{d\zeta}{t^{\kappa}}\ ] ] with a path of integration @xmath878 which runs from @xmath879 along @xmath880 to some point near the origin , takes a @xmath881 radian turn along a circle with the center at the origin , and goes back to infinity in the direction @xmath882 . when condition is satisfied and @xmath838 , then coincides with . the borel sum also satisfies the same properties as previously defined . we apply the borel summation method to the gevrey solution constructed in theorem [ theorem - dimension-1 ] . our main result of this section is [ borel - thm ] we assume that the @xmath827-row span of the matrix @xmath0 does not contain the vector @xmath883 but that the one of @xmath450 does . we also assume @xmath29 to be very generic . let @xmath884 be one of the formal solutions of the modified hypergeometric system @xmath885 constructed in theorem [ theorem - dimension-1 ] and @xmath886 be the gevrey index of @xmath887 along @xmath92 . we also assume @xmath888 . then the formal solution @xmath887 is @xmath889-summable ( as a formal power series in @xmath100 ) in all but finitely many directions for each @xmath890 where @xmath891 is a non - empty open set in the @xmath892-space @xmath5 . furthermore its borel sum determines a solution of the modified hypergeometric system @xmath893 . [ remark4 ] under the assumption of theorem [ borel - thm ] , @xmath894 holds for any @xmath895 , where @xmath744 is the basis of @xmath896 given in remark [ remark3 ] . therefore if @xmath897 , then @xmath898 is an integer . the condition that @xmath899 contains @xmath883 is assumed so that the borel transformed series satisfies a regular holonomic system @xcite . hence , the growth condition ( ii ) of the borel summability is satisfied because solutions of regular holonomic systems satisfy a polynomial growth condition . without this assumption , things become more complicated . see also section [ section : borel_transform_revisited ] . _ proof_._(of theorem [ borel - thm])_. first of all , the open set @xmath891 in the theorem can be chosen as follows . there exist constants @xmath900 such that the series @xmath901 , which will be defined in the proof below , converges when @xmath902 belongs to the non empty open set @xmath903 where @xmath904 is defined by the inequalities @xmath905 , for @xmath906 . such constants exist because @xmath901 is a hypergeometric series which satisfies a regular holonomic @xmath0-hypergeometric system ( @xcite , ( * ? ? ? * section 2.5 ) ) ; see also forthcoming lemma [ borel - lemma2 ] . since only non - negative powers of @xmath907 modulo an exponent appear in @xmath908 , we may assume that @xmath909 for @xmath910 . we may choose a non empty domain @xmath911 with compact closure such that @xmath912 for some @xmath153 . to study the analytic properties of @xmath913 $ ] , it is convenient to use @xmath914 since @xmath915 is a formal power series in @xmath916 , @xmath917 does not contain any fractional powers in @xmath918 ( cf . remark [ remark4 ] ) . then we have @xmath919(x , \zeta ) = \sum_{\ell = 0}^\infty \frac{c_{\ell } ( x)}{\gamma(1 + r ( \ell + \gamma ) ) } \zeta^{r ( \ell + \gamma ) } = \hat{\mathcal{b}}_{1/r}[\psi](x , \zeta^{r}).\ ] ] in what follows we simply write @xmath920 ( resp . , @xmath921 ) instead of @xmath922(x , \zeta)$ ] ( resp . , @xmath913(x , \tau)$ ] ) . [ borel - lemma1 ] assume condition holds for @xmath923 . for the power series @xmath55 given in , we have @xmath924 \,\,\ , \mbox { and } \,\,\ , \frac{\partial \varphi_b}{\partial\zeta } = \hat{\mathcal{b}}_1[z^{-1 } \varphi].\ ] ] _ proof_. the first relation follows from @xmath925 = \hat{\mathcal{b}}_1 \left [ z \frac{\partial \varphi}{\partial z } \right].\end{aligned}\ ] ] we also have @xmath926 = \hat{\mathcal{b}}_1 \left [ z^{-1 } \varphi \right ] . \qquad \qed\end{aligned}\ ] ] [ borel - lemma2 ] the formal power series @xmath901 given in formally satisfies the hypergeometric system @xmath927 , where @xmath928 when the matrix @xmath929 contains a rational entry , we regard the @xmath156-module generated by the column vectors as the lattice to define the @xmath0-hypergeometric system . for example , when @xmath930 @xmath931 , the lattice is @xmath932 and the hypergeometric system is nothing but that for @xmath933 and @xmath931 for the lattice @xmath934 . _ proof_. it follows from lemma [ borel - lemma1 ] and relations @xmath935 , @xmath936 that @xmath937 \\ & = \hat{\mathcal{b}}_{1 } \left [ \left . \left ( \sum_{j = 1}^n a_{ij } \theta_j - \beta \right)\psi \right\|_{t = z^r } \right ] = 0\end{aligned}\ ] ] and @xmath938 \\ & = \hat{\mathcal{b}}_{1 } \left [ \left . \left(\sum_{i = 1}^n w_i \theta_i - \theta_t \right)\psi \right\|_{t = z^r } \right ] = 0.\end{aligned}\ ] ] now we take vectors @xmath939 satisfying @xmath940 . by its definition , we obtain @xmath941 without loss of generality we may assume one of @xmath942 and @xmath943 is zero . under this assumption , @xmath944 and @xmath945 are non - negative integers . furthermore @xmath946 holds . therefore @xmath947 hence @xmath948 \\ & \qquad = \hat{\mathcal{b}}_{1 } \left [ z^{-(u_{n + 1 } + v_{n + 1 } ) } \big(\partial_1^{u_1 } \cdots \partial_n^{u_n } z^{v_{n + 1 } } - \partial_1^{v_1 } \cdots \partial_n^{v_n } z^{u_{n + 1}}\big ) \varphi \right ] \\ & \qquad = \hat{\mathcal{b}}_{1 } \left [ z^{-(u_{n + 1 } + v_{n + 1 } ) } \big(\partial_1^{u_1 } \cdots \partial_n^{u_n } t^{v_{n + 1}/r } - \partial_1^{v_1 } \cdots \partial_n^{v_n } t^{u_{n + 1}/r}\big ) \psi\ , \big\|_{t = z^r } \right ] \\ & \qquad = 0.\end{aligned}\ ] ] here we have used the second relation of lemma [ borel - lemma1 ] . this completes the proof . under the assumption of theorem [ borel - thm ] , @xmath927 is regular holonomic @xcite . therefore , up to an non zero constant , @xmath920 is nothing but a gkz series solution of @xmath927 , and @xmath949 converges near @xmath950 if @xmath890 . it also follows that the restriction of @xmath920 to @xmath951 , which is a function of @xmath907 , satisfies some linear ordinary differential equation @xmath952 of fuchsian type . let @xmath953 be the set of singular points of @xmath954 except the origin and infinity and define @xmath955 . for any @xmath845 with @xmath956 we have , * @xmath957 can be analytically continued to a sector @xmath848 with some small @xmath846 since there is no singular points on @xmath958 . * the borel transform @xmath959 has polynomial growth with respect to @xmath907 in @xmath848 since a singular point of @xmath954 is a regular singular point . hence we conclude that @xmath960 is borel summable ( i.e. , 1-summable ) in the direction @xmath850 . since @xmath961 is a finite set for each fixed @xmath962 , @xmath960 is borel summable in all but finitely many directions for each fixed @xmath963 . we can consider a non empty open subset @xmath964 such that the closure of @xmath965 is compact and included in @xmath891 . then the complement in @xmath966 of the union @xmath967 contains an open interval . so for any @xmath850 in this complement we can define @xmath968(x , t)$ ] using @xmath969 as a path of integration for any @xmath970 . we may also avoid if necessary the discriminant of the leading term of the linear ordinary differential operator @xmath971 . we still write @xmath972 . because of the relation , @xmath973 is @xmath889-summable if and only if @xmath55 is @xmath189-summable . therefore @xmath974 is @xmath889-summable in all the directions in an open interval for all points @xmath975 . finally we show that the borel sum of @xmath974 is a solution of the modified hypergeometric system @xmath885 to finish the proof of theorem [ borel - thm ] . because of the relation @xmath976\big\|_{t = z^r } = \mathcal{l}^{\theta}_{1/r}\hat{\mathcal{b}}_{1/r}[\psi]\big\|_{t = z^r } = \mathcal{l}^{\theta / r}_{1}\hat{\mathcal{b}}_{1}[\varphi ] = \mathcal{s}[\varphi],\end{aligned}\ ] ] lemma [ borel - lemma1 ] and lemma [ borel - lemma2 ] , it is enough to prove ( we omit the direction @xmath977 in the following . ) @xmath978 = \mathcal{l}_1 \left[\zeta \frac{\partial\varphi_b}{\partial \zeta}\right ] , \quad z^{-1 } \mathcal{l}_1[\varphi_b ] = \mathcal{l}_1\left[\frac{\partial\varphi_b}{\partial \zeta}\right].\ ] ] _ proof_. to begin with , we give a proof in the case when @xmath979 or @xmath980 . by differentiating under the integral sign we obtain @xmath981 & = \int^{\infty}_0 \dfrac{\zeta}{z } e^{-\zeta / z } \varphi_b(x , \zeta ) \dfrac{d\zeta}{z } - \int^{\infty}_0 e^{-\zeta / z } \varphi_b(x , \zeta ) \dfrac{d\zeta}{z } \\ & = - \int^{\infty}_0 \frac{\partial}{\partial \zeta } ( \zeta e^{-\zeta / z } ) \cdot \varphi_b(x , \zeta ) \dfrac{d\zeta}{z}.\end{aligned}\ ] ] by integral by parts , this equals @xmath982_{\zeta = 0}^{\infty } + \int^{\infty}_0 e^{-\zeta / z } \cdot \zeta \frac{\partial \varphi_b}{\partial \zeta}(x , \zeta ) \dfrac{d\zeta}{z}.\end{aligned}\ ] ] since the boundary terms vanish ( the boundary term coming from infinity vanishes because of the growth estimate of @xmath908 ) , the first relation follows . in a similar manner , we obtain @xmath983 & = \int^{\infty}_0 e^{-\zeta / z } \frac{\partial\varphi_b}{\partial\zeta}(x , \zeta ) \dfrac{d\zeta}{z } \\ & = \left [ e^{-\zeta / z } \frac{\varphi_b(x , \zeta)}{z } \right]_{\zeta = 0}^{\infty } + \frac{1}{z } \int^{\infty}_0 e^{-\zeta / z } \varphi_b(x , \zeta ) \dfrac{d\zeta}{z}.\end{aligned}\ ] ] the growth estimate of @xmath908 at infinity and the behavior of @xmath908 near the origin guarantee that the boundary terms vanish . this proves the second relation . when @xmath984 and @xmath985 satisfies , we use as the definition of the borel sum , and the same argument works . in this case all of the boundary terms come from infinity and they vanish . this finish the proof of theorem [ borel - thm ] . in the preceding proof , we have shown that @xmath986 is of polynomial growth in @xmath280 for @xmath975 . therefore , for an arbitrarily small positive @xmath874 we can find @xmath987 for which holds with @xmath988 , @xmath989 for all @xmath975 after eventually replacing @xmath891 by a smaller open set . therefore , the condition guarantees the following corollary . [ corollary - asymptotic ] the laplace integral of the borel transform @xmath986 converges in @xmath990 if @xmath991 . in particular , we can set @xmath992 in the expression @xmath968(x , t)$ ] of the borel sum of @xmath974 if @xmath993 does not contain @xmath195 . the series @xmath968(x,1)$ ] gives a holomorphic solution of @xmath6 and the formal series @xmath994 also expresses the asymptotic expansion of the solution @xmath968(x,1))$ ] along a curve @xmath995 as @xmath996 for @xmath963 . that is @xmath997(x(t ) , 1)- \psi_n ( x(t),1)\| \leq c k^n \|t\|^n\gamma ( 1 + r n)\ ] ] for all @xmath998 for some constants @xmath999 , where @xmath1000 . a comment on the statement of previous corollary . a priori the constants @xmath1001 of the estimate for @xmath974 and @xmath923 depend on @xmath975 . since @xmath959 is holomorphic at @xmath1002 for @xmath963 , we can find uniform constants @xmath1001 for @xmath892 in a small neighborhood of each @xmath1003 in @xmath891 with @xmath1004 . notice also that , from lemma [ formal - solution - w - perturbation ] , @xmath1005 and that @xmath1006 is a ( possibly gevrey ) solution of @xmath6 . we also have the following formal equalities : @xmath1007 and @xmath968(x , t)={\mathcal s}[\psi](t^wx,1)$ ] . then , fixing @xmath963 , the estimate evaluated at points @xmath1008 proves that the formal expression @xmath1009 is an asymptotic expansion of @xmath1010(t^w x^0,1)$ ] when @xmath1011 . [ remark6 ] since the singular points @xmath1012 of @xmath971 depend on @xmath892 , one of them may meets the path of integration of the borel sum if @xmath892 moves . in that case we obtain the analytic continuation of the borel sum by deforming the path of integration . this is closely related to the stokes phenomenon . [ example12 ] put @xmath84 . then , the image of @xmath830 is @xmath1013 . let us take @xmath1014 . a formal solution of the modified system @xmath607 is @xmath1015_{2m}}{m ! } \left ( \frac{x_2}{x_1 ^ 2 } \right)^m t^m.\ ] ] if @xmath1016 , the gevrey index @xmath1017 of @xmath974 along @xmath91 is @xmath566 . we set @xmath1018 and @xmath1019 . the borel transform @xmath908 is @xmath1020_{2m}}{m ! } \left ( \frac{x_2}{x_1 ^ 2 } \right)^m \frac{\zeta^m}{m ! } = x_1^\beta \cdot { } _ 2f_1(-\beta/2,(-\beta+1)/2,1;4 x_2 \zeta / x_1 ^ 2).\end{aligned}\ ] ] the domain @xmath891 may be defined by @xmath1021 . the series @xmath908 satisfies the @xmath0-hypergeometric system with @xmath1022 and @xmath1023 . the equation @xmath1024 in the proof is @xmath1025 \varphi_b(x , \zeta ) = 0,\ ] ] and @xmath1026 the borel sum of @xmath887 is @xmath1027 and , for each @xmath890 , @xmath55 ( and hence @xmath973 ) is borel summable in all directions except the angle @xmath1028 . the series @xmath1029 can be regarded as an asymptotic expansion of a solution of the original @xmath0-hypergeometric system for @xmath84 and @xmath29 from ( [ eq : asymptotic_expansion ] ) and corollary [ corollary - asymptotic ] . in other words , we have @xmath1030 which is a well - known asymptotic expansion . [ example1356 ] put @xmath1031 . then , the image of @xmath830 is @xmath1032 let us take @xmath1033 in the @xmath1034 . then , the initial ideal @xmath73 is generated by @xmath1035 and @xmath1036 . the rank of this system is @xmath189 and the solution of this system is spanned by @xmath1037 . we extend this solution to a series solution of @xmath6 . the solution can be written as @xmath1038_{u_-}}{[v+u]_{u_+ } } x^{v+u}= x_1^{\beta } \big ( 1 + \frac{\beta(\beta-1)(\beta-2)}{1 ! } x_1^{-3 } x_2 + \cdots \big ) \label{eq : ex1356series2}\ ] ] where @xmath1039 . the corresponding series solution of the modified system is @xmath1040 the gevrey index @xmath1041 of the series @xmath639 with respect to @xmath91 is @xmath1042 if @xmath29 is very generic , see remark [ remark4 ] . apply the borel transformation ( [ borel - tr - of - formal - solution2 ] ) . the transformed series @xmath1043 satisfies the @xmath0-hypergeometric system associated to the matrix @xmath1044 and @xmath1045 . it follows from the condition on the weight vector @xmath4 that this @xmath927 is a regular holonomic system . the series ( [ eq : ex1356seriesb ] ) can be obtained by taking the @xmath1046-initial ideal of @xmath927 with respect to the weight vector @xmath1047 and @xmath1048 as the tie breaking weight vector ( * ? ? ? * chapters 2,3 ) . the kernel element @xmath1049 of @xmath929 as a map from @xmath1050 to @xmath1013 is parametrized as @xmath1051 since we sum the series on @xmath1052 , @xmath1053 and @xmath1054 , these lattice elements @xmath1049 can be parametrized as @xmath1055 @xmath1056 where @xmath1057 or @xmath189 . let us introduce the following hypergeometric series @xmath1058 depending on an integer @xmath1059 . we denote it by @xmath1060 when @xmath1061 and by @xmath1062 when @xmath1063 . then , the series is expressed as @xmath1064 where @xmath1065 it follows from that the series is an asymptotic expansion of @xmath1066 in this section we assume that the @xmath827-row span of the matrix @xmath0 does not contain the vector @xmath1067 and we see that by studying the irregularity of the modified @xmath0-hypergeometric system along @xmath92 we can give an analytic meaning to the gevrey series solutions of the @xmath0-hypergeometric system @xmath1068 along coordinate varieties constructed in @xcite . more precisely , we prove ( remark [ w - in - borel - revisited ] and theorem [ theorem - asymptotic - expansion ] ) that these gevrey solutions of @xmath95 are asymptotic expansions of certain holomorphic solutions of @xmath1068 when some conditions are satisfied . let us start with an observation about the assumption of theorem [ borel - thm ] that the @xmath827-row span of the matrix @xmath0 does not contain the vector @xmath828 , but the one of @xmath450 does . since adding to @xmath4 a linear combination of the rows of @xmath0 does not change the ideal @xmath1069 if we accordingly change @xmath93 , we can assume without loss of generality that @xmath1070 for some non zero @xmath1071 . on the other hand , for @xmath1072 the modified system @xmath1073 is regular along @xmath92 by proposition [ regularity - along - t ] and remark [ regularity - along - t ] . thus , we may further assume that @xmath1074 for some @xmath1075 . for this @xmath4 each of the formal series constructed in the proof of theorem [ theorem - dimension-1 ] is a gevrey series along @xmath92 with order @xmath1076 multiplied by a term @xmath1077 , @xmath685 . this sort of formal series solutions along @xmath92 of the modified system @xmath1073 includes series of the form @xmath1078 where @xmath1079 is a gevrey solution of @xmath247 along a coordinate variety @xmath66 of low dimension which is not gevrey along a coordinate variety of greater dimension ( see the proof of proposition [ gevrey - for - special - weight - vector ] ) . recall that a vector @xmath243 is said to be associated with a simplex @xmath178 if @xmath1080 is such that @xmath612 for all @xmath1081 and @xmath1082 . let us also recall that @xmath1083 is the convex hull of the columns of @xmath0 and the origin while @xmath1084 is the convex hull of the columns of @xmath0 . we have the following [ gevrey - for - special - weight - vector ] take @xmath1085 with @xmath1075 and assume that @xmath29 is very generic . let @xmath178 be a @xmath253-simplex of @xmath0 and @xmath243 a vector associated with @xmath178 . then , up to multiplication by a term @xmath1077 , for some @xmath685 , the series @xmath1086 is a formal power series along @xmath92 ( and in such a case , it is a gevrey solution of @xmath1087 with index @xmath1088 ) if and only if @xmath178 is contained in a facet of @xmath1089 that is not a facet of @xmath1083 . _ proof_. from ( * ? ? ? * theorem 3.11 ) , @xmath241 is a gevrey solution of @xmath1068 along @xmath1090 with order @xmath1091 . in fact , for any simplex @xmath178 of @xmath0 we can construct @xmath1092 linearly independent gevrey solutions as before ( see ( * ? ? ? * remark 3.6 ) ) . the series @xmath1094 is annihilated by @xmath36 because @xmath241 is annihilated by @xmath12 . moreover , monomials appearing in @xmath241 are of the form @xmath1095 for @xmath671 integer vectors in an affine translate of the positive span of the columns of the matrix @xmath267 . thus @xmath1096 is a formal series along @xmath92 ( up to multiplication by @xmath1097 ) if and only if all the exponents of @xmath100 belong to @xmath1098 . this happens if and only if @xmath1099 for all the columns @xmath671 of @xmath267 . for @xmath1100 with @xmath1075 the scalar product of @xmath4 and a column of @xmath267 is given by @xmath1101 for @xmath613 and this product is nonnegative for all @xmath613 if and only if @xmath1102 for all @xmath613 . this is equivalent to say that @xmath178 is a simplex of @xmath0 such that all the columns of @xmath0 are either in the hyperplane @xmath348 passing through the columns of @xmath0 indexed by @xmath178 or in the corresponding single open half space not containing the origin . equivalently , @xmath178 is contained in a facet of @xmath1084 that is not a facet of @xmath1083 . the gevrey series solutions constructed in @xcite are of the form @xmath241 for @xmath243 associated with a suitable simplex @xmath178 of @xmath0 . in order to interpret some of these gevrey series as an asymptotic expansion of a solution of @xmath157 via the modified @xmath0-hypergeometric system , we consider a vector @xmath26 with the following coordinates : @xmath1103 up to addition with a linear combination of the rows of @xmath0 . notice that when @xmath178 is contained in a facet of @xmath1089 that is not a facet of @xmath1083 , this vector @xmath4 verifies the assumptions of theorem [ borel - thm ] . more precisely , @xmath1104 is a linear combination of the rows of @xmath0 . [ two - hyperplanes ] notice that for @xmath4 given by ( [ sigma - weight - vector ] ) all the columns @xmath451 of the matrix @xmath80 except for @xmath1105 are contained in the union of at most two hyperplanes . if we take coordinates @xmath1106 in @xmath1107 , then the hyperplane @xmath1108 contains all the columns @xmath451 such that @xmath1109 and the hyperplane @xmath1110 contains @xmath1111 and all the columns @xmath451 such that @xmath1102 . in particular , the intersection of these two hyperplanes contains all the columns @xmath451 such that @xmath1112 ( for example , all the columns @xmath451 for @xmath492 ) . we also notice that the points @xmath1113 belong to the same hyperplane if and only if @xmath178 is contained in a facet of @xmath1089 that is not a facet of @xmath1083 . notice that if @xmath178 is contained in a facet of @xmath1083 and @xmath243 is associated with @xmath178 , then the corresponding @xmath4 given by ( [ sigma - weight - vector ] ) is @xmath195 and @xmath1114 is convergent . [ gevrey - for - sigma - weight - vector ] let @xmath178 be a @xmath253-simplex of @xmath0 not contained in a facet of @xmath1083 and consider @xmath4 given by ( [ sigma - weight - vector ] ) and @xmath29 very generic . for any vector @xmath243 associated with @xmath178 we have that , up to multiplication by @xmath1097 , the series @xmath1116 is a gevrey solution of @xmath1087 with index @xmath1117 along @xmath92 at any point of @xmath1118 for some @xmath683 , where @xmath1119 . it is clear from ( [ sigma - weight - vector ] ) that @xmath1121 and @xmath1122 for all @xmath1123 . hence , for any exponent @xmath1124 in the series @xmath1114 the corresponding exponent of @xmath100 in the series @xmath1125 is @xmath1126 because @xmath1127 for all @xmath1128 and @xmath658 . we conclude the proof by using remark [ remark3 ] . in analogy with section [ section : borel ] we denote @xmath1129(x , \tau)$ ] , which defines a holomorphic function at any point in @xmath1130 for some @xmath683 by proposition [ gevrey - for - sigma - weight - vector ] . we also denote @xmath1131 and hence @xmath1132(x , \zeta ) = \psi_b ( x , \zeta^{r'})$ ] is convergent at points in the open set @xmath1133 for @xmath4 given by ( [ sigma - weight - vector ] ) the hypergeometric system @xmath927 , can have slopes along @xmath1137 for all @xmath1138 ( see example [ example - slope - infinity-2 ] ) . however , see proposition [ bounded - growth ] where we point out a property of the solution @xmath901 . [ example - slope - infinity-2 ] let us consider the matrix @xmath1139 @xmath1140 , @xmath31 , the simplex @xmath1141 and @xmath1142 given by ( [ sigma - weight - vector ] ) . notice that @xmath1143 and hence @xmath1144 using corollary [ slopes - hypergeometric - at - infinity ] we have that @xmath1145 is a slope of @xmath1146 along @xmath1137 . [ technical - lemma ] assume @xmath1147 where @xmath1148 . let @xmath1149 be a simplex , @xmath1150 a vector associated with @xmath1151 ( i.e. @xmath1152 and @xmath1153 for all @xmath1154 ) . the series @xmath1155 converges at points @xmath1156 and for arbitrarily small @xmath1157 we can choose @xmath1158 such that @xmath1159 . _ proof_. notice that @xmath1163 is a facet of the @xmath1164-umbrella and @xmath1165 is a simplex of @xmath929 contained in @xmath1163 . in particular we know that @xmath796 is convergent in certain open set . let us denote @xmath1166 . if @xmath1167 is the basis of @xmath1168 associated with @xmath1165 then the coordinate sum of @xmath1169 is @xmath1170 let us denote by @xmath1171 the vector given by the first @xmath3 entries of @xmath1169 . since @xmath1172 , we have that @xmath1173 for all @xmath521 and that the series @xmath796 defines a multivalued holomorphic function in the open set @xmath1174 for some @xmath683 . the fact that @xmath1175 guarantees that @xmath1176 for all @xmath1177 and @xmath1178 for all @xmath1179 . thus , if @xmath1180 the last coordinate of @xmath1169 is @xmath1181 while if @xmath1182 the last coordinate of @xmath1169 can be positive . on the other hand , if @xmath1188 for @xmath1189 , it can be shown by using standard estimates on @xmath22-functions ( see e.g. ( * ? ? ? * proposition 1 , section 1.1 ) , ( * ? ? ? * lemma 1 ) and ( * ? ? ? * lemma 3.8 . ) ) , that there exists @xmath1190 such that and for all @xmath1189 we have : @xmath1192 take @xmath1193 where @xmath1194 can be chosen arbitrarily large except when @xmath1195 in which case we take @xmath1196 . then , since @xmath1187 if @xmath1197 , we have the result for for arbitrarily small @xmath1198 if we take @xmath1157 and @xmath1199 big enough . [ bounded - growth ] @xmath1134 has an analytic continuation to an open set of the form @xmath1200 , where @xmath891 is certain open set of @xmath5 and @xmath1201 is a sector with bisecting direction @xmath850 and small enough opening @xmath1202 . moreover , if @xmath1147 then for arbitrarily small @xmath1157 we can chose @xmath1203 such that @xmath1204 for @xmath1205 . notice that @xmath1163 is the set of ( indices of ) columns of @xmath929 belonging to the hyperplane @xmath1206 ( see remark [ two - hyperplanes ] ) and we denote by @xmath1207 the submatrix of @xmath929 consisting of these columns . let @xmath1208 be the cardinality of @xmath250 . recall that @xmath1209 defines a holomorphic function at each point of @xmath1210 ( see proposition [ gevrey - for - sigma - weight - vector ] and ( [ open - set ] ) ) and notice that @xmath1211 for all @xmath1212 . we can write @xmath1213 where @xmath1214 is a holomorphic solution of @xmath1215 , which is regular holonomic because all the columns of @xmath1207 belong to the hyperplane @xmath418 @xcite . let @xmath1216 be the open set such that @xmath1217 ( see ( [ open - set ] ) ) , so that for all @xmath1218 , @xmath1219 is holomorphic in @xmath1220 . recall that the singular locus of a hypergeometric system does not depend on the parameter but only on the matrix ( see @xcite and @xcite ) . in particular , since @xmath1219 is convergent in @xmath1220 for all @xmath648 , we can consider the analytic continuation of all the @xmath1219 along the same path starting at a point in @xmath1220 and avoiding the singular locus of the hypergeometric system associated with @xmath1207 . let @xmath1223 be such that the complex line @xmath1224 ( with coordinate @xmath907 ) intersects @xmath1220 at nonsingular points of @xmath1225 . notice that this intersection is a relative open set in the complex line . let @xmath1226 be the set of points @xmath1227 such that @xmath1228 is a singular point of @xmath1215 . @xmath1226 is a finite set and thus @xmath1229 is also finite . as we vary @xmath496 in a small open set @xmath1230 , @xmath1231 is contained in a finite union of small intervals and we can take @xmath850 such that for @xmath1202 small enough , @xmath1232 for all @xmath1233 . hence we can consider the analytic continuation of each @xmath1219 to an open set containing @xmath1234 . we have extended @xmath908 as a formal solution of @xmath1235 along @xmath272 , which is convergent in some relative open set of @xmath272 . thus , by the constructibility of the solutions of a holonomic system in the sheaf @xmath1236 ( see @xcite ) we have that the formal solution constructed is convergent at any point of @xmath1237 , thus @xmath1209 can be analytically continued to an open set containing @xmath1238 . let us see that the analytic continuation of @xmath1209 satisfies a growth estimate near @xmath1137 . each @xmath1239 has polynomial growth since it is a solution of the regular hypergeometric system @xmath1240 . since @xmath1241 are very generic , each @xmath1239 can be written as a nilsson series that converges in certain open set ( see e.g. ( * ? ? ? * proposition 3.4.4 ) ) which is a linear combination of series of the form @xmath1242 ( for some set of exponents @xmath1243 associated with simplices in certain regular triangulation of the matrix @xmath1207 ) with support @xmath1244 given by integer vectors in @xmath1245 with coordinates sum equal to zero . the open set where the nilsson series converge depends on the regular triangulation of @xmath1207 that the simplices belong to . we need to use a nilsson series expression of @xmath1239 that converges in points @xmath1246 with @xmath1247 for a sufficiently large @xmath683 . it is enough to consider a regular triangulation @xmath92 of @xmath249 and take @xmath1248 as the regular triangulation of @xmath1207 . by properties of regular triangulations , there is one regular triangulation @xmath92 of @xmath249 such that there exists @xmath496 as above so that if @xmath1247 for a sufficiently large @xmath683 then @xmath1249 belongs to the domain of convergence of the series @xmath796 for any vector @xmath1250 associated with @xmath1251 . the series expression of @xmath908 via substitution of each @xmath1219 by its nilsson series expansion is a formal nilsson series ( see e.g. ( * ? ? ? * lemma 6.15 ) ) . we know that this nilsson series converges to @xmath908 at points in @xmath1237 close to @xmath1137 . by lemma [ technical - lemma ] it verifies the desired growth estimate . notice that the nilsson series expansion of @xmath908 near @xmath1137 also provides an analytic continuation to points with @xmath1252 big enough and @xmath892 varying in certain open set of @xmath5 that contains @xmath1253 . we have considered analytic continuations along paths contained in @xmath272 . we can also extend @xmath908 by analytic continuation along paths from a point in @xmath1254 to a point near @xmath1255 avoiding the singular locus of @xmath1256 . if the starting point of the path is close to @xmath1253 the analytic continuation coincides with the nilsson series close to @xmath1137 and thus it satisfies the same growth estimate . finally , the parameter @xmath29 being very generic , the rank of @xmath1146 equals @xmath1257 since the set of exceptional parameters has codimension at least 2 ( * ? ? ? * porism 9.5 ) . so , we can reduce the general case ( when @xmath93 is not necessarily generic ) to the previous one following the ideas of the proof of ( * ? ? ? * theorem 3.5.1 ) ( see also the proof of ( * ? ? ? * theorem 6.2 . ) ) . [ w - in - borel - revisited ] using proposition [ bounded - growth ] , the results in section [ section : borel ] also hold for @xmath4 as in ( [ sigma - weight - vector ] ) instead of @xmath4 satisfying the assumption in theorem [ borel - thm ] if we assume the additional condition @xmath1147 to hold . in particular , we obtain an analogous version of corollary [ corollary - asymptotic ] . let @xmath1258 be very generic and let @xmath241 be a gevrey solution of @xmath265 of index @xmath1259 with respect to some coordinate subspace @xmath635 . let @xmath178 be the simplex which @xmath243 is associated with , so we have that @xmath241 is also a gevrey series of index @xmath1260 with respect to @xmath1261 . note that if we take @xmath4 associated with @xmath178 as in ( [ sigma - weight - vector ] ) then @xmath1262 for all @xmath1263 . by proposition [ gevrey - for - sigma - weight - vector ] we have that @xmath1264 is a gevrey series along @xmath91 of gevrey index @xmath1265 with @xmath1166 . let @xmath1266(x , t)$ ] be the @xmath839-sum of @xmath1125 with respect to @xmath100 in a direction @xmath1267 for @xmath892 in certain open set @xmath891 small enough with compact closure . for any closed subsector @xmath865 of @xmath1268 ( see notations in section [ section : borel ] ) there exist constants @xmath1269 such that the inequality @xmath1270(c , t)-t^{\alpha } \psi_n ( c , t)\| \leq c k^n \gamma ( 1+n/\kappa ) \|t\|^n $ ] holds for @xmath1271 , @xmath1272 and any @xmath1273 . thus , considering parametric curve @xmath1274 then @xmath1275_{u_{-}}}{[v+u]_{u_{+ } } } c^{v+u } t^{w ( v+u ) } = t^{\alpha}\psi_n ( c , t ) \ ] ] and @xmath1266(x(t),1)= t^{\alpha}\mathcal{s}[\psi](c , t)$ ] . note that @xmath892 tends to the point @xmath1276 , with @xmath1277 if @xmath1109 , as @xmath100 tends to @xmath195 . let @xmath1278 be very generic and let @xmath241 be a gevrey solution of @xmath265 of order @xmath1259 with respect to a coordinate hyperplane @xmath1279 . let @xmath178 be the simplex which @xmath243 is associated with and take @xmath4 associated with @xmath178 as well . if @xmath1147 then for @xmath1166 we have that @xmath1280(x , 1)$ ] is a holomorphic solution of @xmath196 and that for each @xmath1281 in certain open set of @xmath1282 , @xmath634 is a gevrey asymptotic expansion of order @xmath211 of @xmath1280(x , 1)$ ] with respect to @xmath1283 in all but finitely many directions . _ proof.- _ assume for simplicity that the hyperplane is @xmath1284 and so @xmath1285 . we have that @xmath1286 for @xmath1287 and @xmath1288 where @xmath1289 is the gevrey index of @xmath241 along @xmath66 . here we write @xmath1292 . for integers of the form @xmath1293 with @xmath1294 we have that @xmath1295 and @xmath1296 where @xmath1297 . then we get the inequality @xmath997(x(t ) , 1)- \psi_n ( x(t),1)\| \leq c ( k^{w_n}/\|c_n\|)^m \gamma ( 1 + m / k ) \|x_n(t)\|^m\ ] ] for all @xmath1298 . this finishes the proof since we can assume by taken a smaller open set @xmath891 that @xmath1299 and that @xmath1300 for some constant @xmath1301 . put @xmath1302 , @xmath557 and @xmath1303 . the vector @xmath1304 is an exponent of the @xmath0-hypergeometric system @xmath265 with respect to a perturbation of @xmath4 and so the series @xmath1305_{(m_1 + 3 m_3 ) /2}}{m_1 ! m_3 ! } x_1^{m_1 } x_2^{(\beta -m_1 - 3 m_3 ) /2 } x_3^{m_3 } t^{m_3 } \ ] ] is one of the series considered in the proof of theorem [ theorem - dimension-1 ] and it is a gevrey solution of the modified system @xmath1306 along @xmath92 with order @xmath1307 . notice that @xmath1307 is the gevrey index of @xmath1308 along @xmath92 if and only if @xmath1309 ( otherwise @xmath1308 is a polynomial ) . following section [ section : borel ] but with our vector @xmath4 ( which does not satisfy the assumptions in section [ section : borel ] but is of the form for @xmath1310 ) we consider the borel transform of @xmath973 with index @xmath1311 : @xmath1312_{(m_1 + 3 m_3 ) /2}}{m_1 ! m_3 ! \gamma ( 1 + m_3 /2 ) } x_1^{m_1 } x_2^{(\beta -m_1 - 3 m_3 ) /2 } x_3^{m_3 } \tau^{m_3}.\ ] ] this series defines a holomorphic function in @xmath1313 for @xmath1314 small enough and it has an analytic continuation with respect to @xmath280 to certain sector @xmath848 . let us see that this analytic continuation has polynomial growth in @xmath280 . if @xmath1315 then its borel transform ( with index @xmath189 ) @xmath1316_{(m_1 + 3 m_3 ) /2}}{m_1 ! m_3 ! \gamma ( 1 + m_3 /2 ) } x_1^{m_1 } x_2^{(\beta -m_1 - 3 m_3 ) /2 } x_3^{m_3 } \zeta^{m_3 / 2 } \ ] ] is a solution of the hypergeometric system associated with @xmath929 and @xmath1317 defined on @xmath1318 with coordinates @xmath1319 . notice that @xmath908 has fractional powers in @xmath907 but defines a multivalued holomorphic function in @xmath1320 for @xmath1314 small enough . it is clear that @xmath1209 is a linear combination of series @xmath1321 with @xmath1322 associated with the simplex @xmath1323 of @xmath929 ( i.e. , @xmath1324 , @xmath1325 for @xmath1326 ) . we have an analytic continuation of @xmath1209 to a point in the open set @xmath1327 for @xmath683 big enough , which must be a linear combination of series @xmath1328 with @xmath1329 associated with the simplex @xmath1330 ( this simplex alone determines a regular triangulation of the of @xmath929 and the set of series @xmath796 with @xmath1250 associated with @xmath1330 generates the space of holomorphic solutions of @xmath1331 at any point in the open set @xmath1332 ) . we have that the columns of @xmath1333 are @xmath1334 and we notice that @xmath1335 . this implies that a series @xmath1336 with @xmath1250 associated with @xmath1330 has polynomial growth as @xmath907 tends to infinity . thus , the analytic continuation of @xmath1209 close to @xmath1137 also does . as a consequence , the borel sum of @xmath973 with index @xmath1337 given by @xmath1338(x , t)= \mathcal{l}^{\theta}_{2}\hat{\mathcal{b}}_{2}[\psi](x , t)=\int^{e^{i\theta}\cdot \infty}_0 e^{-(\tau / t)^2 } \psi_b ( x , \tau ) d(\tau / t)^{2}\ ] ] is a holomorphic solution of @xmath1339 and @xmath1266(x,1)$ ] is a holomorphic solution of @xmath181 which has an asymptotic expansion @xmath1340 that is gevrey of order @xmath1341 along @xmath1342 . this example shows that the hypothesis in proposition [ bounded - growth ] on the umbrella is necessary and that the bound there is sharp . take @xmath1343 where @xmath1344 is a rational number . we can consider @xmath1345 by changing the lattice @xmath934 by the lattice @xmath1346 . then , the weight vector @xmath1347 is associated with @xmath1348 by the formula ( [ sigma - weight - vector ] ) and @xmath1349 . let @xmath94 and @xmath340 be very generic and let @xmath1350 be a vector associated with @xmath178 so that the series @xmath1114 is a gevrey solution of @xmath181 along @xmath1351 with gevrey index @xmath1352 . then for @xmath1353 the borel transform @xmath1354 is convergent in the open set @xmath1355 for some @xmath1314 small enough . moreover , it defines a holomorphic solution of @xmath1331 and then it is a linear combination of the set of series @xmath1356 with @xmath1357 associated with the simplex @xmath1358 . its analytic continuation to points in the open set @xmath1359 for some sufficiently large @xmath683 is a linear combination of the set of series @xmath796 with @xmath1360 associated with the simplex @xmath1361 of @xmath929 . the column vectors of @xmath1362 are @xmath1363 and @xmath1364 . elements in the support @xmath1365 of @xmath796 are of the form @xmath1366 with @xmath1367 . thus any of these series @xmath1368 is convergent in the open set @xmath1369 for some sufficiently large @xmath683 and since the last coordinate of @xmath1370 is @xmath1371 there is a subseries of @xmath796 ( the one with monomials @xmath1372 , @xmath1373 ) such that the set of exponents of @xmath907 in its monomials is contained in @xmath1374 . the fact that @xmath1375 guarantees that the coefficients of this subseries has the same type of growth as @xmath1376 and thus the growth of this series , as @xmath907 tends to @xmath879 in certain sector , is equivalent to the growth of @xmath1377 for some @xmath1378 . notice that for @xmath1344 , we have that @xmath1379 and the hypothesis @xmath1147 ( required in lemma [ technical - lemma ] and proposition [ bounded - growth ] ) is satisfied if and only if @xmath1380 . thus , the bound @xmath1381 but for @xmath1382 we can not choose @xmath874 to be arbitrarily small .	a modified @xmath0-hypergeometric system is a system of differential equations for the function @xmath1 where @xmath2 is a solution of an @xmath0-hypergeometric system in @xmath3 variables and @xmath4 is an @xmath3 dimensional integer vector , which is called the weight vector . we study the irregularity of modified systems by adapting to this case the notion of umbrella introduced by m. schulze and u. walther . especially , we study slopes and gevrey series solutions . we develop some applications of this study . under some conditions we give laplace integral representations of divergent series solutions of the modified system and we show that certain gevrey series solutions of the original @xmath0-hypergeometric system along coordinate varieties are gevrey asymptotic expansions of holomorphic solutions of the @xmath0-hypergeometric system .
in this appendix we provide a derivation of eq . ( [ final ] ) from eqs . ( [ langevin ] ) and ( [ djnl ] ) . substituting @xmath77 and @xmath153 into the two latter equations , we obtain @xmath154 if @xmath155 , eq . ( [ jnu ] ) is trivial and the statistical properties of @xmath156 are determined by eq . ( [ current ] ) with @xmath157 , the same equation as in the case of the linear medium , while the static , time - independent part of the intensity fluctuation @xmath158 is found by solving the stationary langevin equation [ eq . ( [ dinu ] ) with @xmath155 ] . hence , the time - independent part of the speckle pattern remains the same as in the linear medium . if , in contrast , @xmath160 ( as we assume in the main text ) , we divide both sides of eq . ( [ jnu ] ) by @xmath161 , multiply the @xmath162-th cartesian component of the resulting equation by the @xmath163-th cartesian component of a similar equation for @xmath164 , and average the result over disorder . this yields @xmath165 , \label{jjnu}\end{aligned}\ ] ] where @xmath166 denotes the @xmath162-th cartesian component of @xmath167 . after the substitution of eq . ( [ qcorr ] ) for @xmath168 , the time integrations in eq . ( [ jjnu ] ) yield @xmath169 \nonumber \\ & & = 3 \pi d^2 ( c^2/l ) \delta_{i j } \delta({{\mathbf r } } - { { \mathbf r}}_1 ) \nonumber \\ & & \times \left [ \left < i({{\mathbf r}}^{\ , \prime } ) \right > g({{\mathbf r}}^{\ , \prime } , { { \mathbf r}}_1^{\ , \prime } ; \nu ) g^({{\mathbf r}}_1^{\ , \prime } , { { \mathbf r } } ; \nu - \nu^ ) \left < i({{\mathbf r } } ) \right > \right . \nonumber \\ & & + \left . < i({{\mathbf r}}_1^{\ , \prime } ) \right > g^({{\mathbf r}}_1^{\ , \prime } , { { \mathbf r}}^{\ , \prime } ; \nu ) g({{\mathbf r}}^{\ , \prime } , { { \mathbf r } } ; \nu - \nu^ ) \left < i({{\mathbf r } } ) \right > \right . \nonumber \\ & & - \left . \left < i({{\mathbf r}}^{\ , \prime } ) \right > g({{\mathbf r}}^{\ , \prime } , { { \mathbf r } } ; \nu ) \left < i({{\mathbf r}}_1^{\ , \prime } ) \right > g^*({{\mathbf r}}_1^{\ , \prime } , { { \mathbf r } } ; \nu ) \right ] , \label{timeint}\end{aligned}\ ] ] where @xmath170 is the fourier transform of @xmath49 . equation ( [ jjnu ] ) can now be rewritten as @xmath171 where @xmath172 . \label{a}\end{aligned}\ ] ] in the following , we replace both @xmath27 and @xmath173 by their spatial averages @xmath174 and @xmath175 , respectively . this simplifies the further analysis considerably , while can only affect the final result by a numerical factor of order unity , since @xmath27 and @xmath173 do not change significantly as long as the point @xmath176 is far enough from the sample boundaries . we now admit that eq . ( [ jjnu2 ] ) for the correlation function of langevin currents at @xmath160 in a nonlinear medium has a form similar to eq . ( [ current ] ) for langevin currents in a linear medium . this allows us to proceed with analysis of eq . ( [ dinu ] ) in the same way as it was done for eq . ( [ langevin ] ) in the linear medium @xcite . to simplify further calculations , we assume that the disordered sample has open boundaries ( i.e. that the multiple - scattered waves leave the sample when they reach the boundary ) and hence the green s function of eq . ( [ langevin ] ) , @xmath49 , can be approximately written as @xmath177 , where @xmath178 $ ] is the green s function in the infinite medium , @xmath179 describes the leakage of the wave through the sample boundaries , and @xmath85 . we now write the solution of eq . ( [ dinu ] ) as @xmath180 \nonumber \\ & = & \int_v \mathrm{d}^3 { { \mathbf r}}^{\ , \prime } \left [ { { \mathbf \nabla } } g({{\mathbf r } } , { { \mathbf r}}^{\ , \prime } ; \nu ) \cdot { { \mathbf j}}_{\mathrm{ext}}({{\mathbf r}}^{\ , \prime } , \nu ) \right ] , \label{di}\end{aligned}\ ] ] where the second line is obtained as a result of integration by parts , assuming @xmath181 at the boundary of the disordered sample . multiplying eq . ( [ di ] ) by a similar equation for @xmath182 , performing the averaging over disorder using eq . ( [ jjnu2 ] ) , and carrying out the necessary integrations , we obtain @xmath183 . \label{didi}\end{aligned}\ ] ] substituting eq . ( [ didi ] ) into eq . ( [ a ] ) , dividing both sides of the resulting equation by @xmath184 , recalling that @xmath185 , and performing a change of variables @xmath186^{1/2 } { { \mathbf r}}$ ] ( and similarly for @xmath63 and @xmath187 ) , we obtain @xmath188 where @xmath189 is a numerical constant , @xmath56 , and the dimensionless function @xmath190 is defined as @xmath191 \nonumber \\ & & \times \left [ \mathrm{re}{\cal g}({{\mathbf r}}^{\ , \prime } , { { \mathbf r}}_1^{\ , \prime } ; \gamma ) + \frac{\lambda}{\omega } \mathrm{im}{\cal g}({{\mathbf r}}^{\ , \prime } , { { \mathbf r}}_1^{\ , \prime } ; \gamma ) \right ] , \label{eq2}\end{aligned}\ ] ] @xmath192 , \label{g } \\ & & { \cal g}_1({{\mathbf r } } , { { \mathbf r}}_1 ; \gamma_1 ) \nonumber \\ & & = \frac{1}{\left\| { { \mathbf r } } - { { \mathbf r}}_1 \right\| } \exp\left(-2 \gamma_1 \left\| { { \mathbf r } } - { { \mathbf r}}_1 \right\| \right ) , \label{g1 } \\ & & \gamma = \left\ { \left [ 1 + \left ( \frac{\lambda + \omega_d}{\omega } \right)^2 \right]^{1/2 } \right . \nonumber \\ & & + \left . \frac{\lambda + \omega_d}{\omega } \right\}^{1/2 } , \label{gamma } \\ & & \gamma_1 = \left ( \frac{\lambda + \omega_d/2}{\omega } \right)^{1/2}. \label{gamma1}\end{aligned}\ ] ] we now assume that the disordered sample has the shape of a sphere centered at the origin and that @xmath193 provides a good estimation of @xmath190 for the points @xmath194 located far enough from the boundaries . defining @xmath195 and introducing the effective nonlinearity parameter @xmath7 , we rewrite eq . ( [ eq1 ] ) as @xmath196 where @xmath197 is a numerical factor of order unity . since we have already made some approximations that affect the final result by a numerical factor of order unity ( e.g. , we replaced @xmath27 by @xmath174 ) , we omit @xmath197 in eq . ( [ eq3 ] ) and obtain eq . ( [ final ] ) of the main text . the most of integrations in eq . ( [ eq2 ] ) can be performed analytically , while the remaining integrations are easily carried out numerically , allowing us to determine the value of @xmath75 for given @xmath74 , @xmath198 , and @xmath0 from eq . ( [ eq3 ] ) . an important comment is in order in connection with eq . ( [ eq3 ] ) and the analysis it results from . the correlation function of intensity fluctuations @xmath199 entering into eqs . ( [ jjnu ] ) and ( [ a ] ) contains , in principle , not only the long - range contribution given by eq . ( [ didi ] ) , but also a short - range one @xmath200 . the latter contribution has been neglected in our analysis , which is justified for large enough sample size ( @xmath83 ) and moderate frequency @xmath84 ^ 2 $ ] . if one of the above inequalities is violated , the roles played by the short- and long - range contributions to the correlation function of intensity fluctuations in development of the instability become comparable , and the above analysis is no longer valid .	we formulate a langevin description of dynamics of a speckle pattern resulting from the multiple scattering of a coherent wave in a nonlinear disordered medium . the speckle pattern exhibits instability with respect to periodic excitations at frequencies @xmath0 below some @xmath1 , provided that the nonlinearity exceeds some @xmath0-dependent threshold . a transition of the speckle pattern from a stationary state to the chaotic evolution is predicted upon increasing nonlinearity . the shortest typical time scale of chaotic intensity fluctuations is of the order of @xmath2 . propagation of a coherent wave in a disordered medium is diffusive if @xmath3 , where @xmath4 is the wavelength , @xmath5 is the mean free path , and @xmath6 is the size of the medium @xcite . while the wave undergoes multiple scattering and the spatial distribution of the scattered intensity looks quite irregular ( speckle pattern ) , the coherence of the wave is not destroyed and various coherent phenomena can be observed : enhanced backscattering , short- and long - range intensity correlations , universal conductance fluctuations , etc . ( see refs . for reviews ) . available studies of _ nonlinear _ phenomena for diffuse waves include calculations of the enhanced backscattering cone at fundamental @xcite and doubled @xcite frequencies , investigations of optical phase conjugation @xcite , studies of correlations in transmission and reflection coefficients of the second harmonic @xcite and fundamental @xcite waves , an extension of the standard diagrammatic technique to nonlinear disordered media @xcite , and a study of persistent hole burning in multiple - scattering media @xcite . after realizing that the sensitivity of the speckle pattern to changes of the scattering potential diverges for a sufficiently strong nonlinearity @xcite , a new phenomenon , the temporal instability of the multiple - scattering speckle pattern in a disordered medium with cubic nonlinearity , has recently been predicted @xcite . the speckle pattern is expected to become unstable and to exhibit spontaneous fluctuations if the nonlinearity exceeds some critical value . although of primary importance in view of the possible experimental observation of the instability phenomenon , the dynamics of spontaneous intensity fluctuations , their nature and associated characteristic time scales have not yet been studied up to now . in the present paper we formulate the _ dynamic _ langevin description of spontaneous intensity fluctuations in a nonlinear disordered medium . our theoretical method can be viewed as an extension of the _ stationary _ langevin approach introduced in ref . , the latter being inadequate to describe the dynamics of speckles . analysis of the speckle pattern stability with respect to weak periodic excitations shows that if the effective nonlinearity parameter @xmath7 exceeds some critical value @xmath8 ( where @xmath9 is the typical value of the nonlinear correction to the refractive index ) the speckle pattern becomes unstable with respect to periodic excitations at frequencies inside some limited low - frequency interval , and the maximal lyapunov exponent becomes positive . this allows us to describe the chaotic nature of spontaneous intensity fluctuations beyond the absolute instability threshold @xmath10 and to estimate their characteristic time scale . we consider a scalar wave propagating in a nonlinear disordered medium and described by the following wave equation : @xmath11 \right\ } \psi({{\mathbf r } } , t ) \nonumber \\ & & = j({{\mathbf r } } , t ) , \label{weq}\end{aligned}\ ] ] where @xmath12 is a monochromatic source term , @xmath13 denotes the speed of wave in the average medium , @xmath14 is the fractional fluctuation of the dielectric constant at frequency @xmath15 ( possibly slowly varying in time ) , and @xmath16 is a nonlinear constant . equation ( [ weq ] ) describes , e.g. , propagation of optical waves in media with intensity - dependent refractive index @xcite in the scalar approximation and neglecting the generation of the third optical harmonics . the latter assumption is justified in the absence of phase matching @xcite or , more precisely , when @xmath17 , where @xmath18 is the wavenumber at frequency @xmath19 . consider first a linear medium ( @xmath20 ) of typical size @xmath6 and a white - noise gaussian disorder : @xmath21 , where @xmath22 . let the time variations of @xmath14 be random , stationary , and arbitrary slow , so that the time scale of the resulting variations of the amplitude @xmath23 of @xmath24 is much larger than the typical time between two successive scattering events @xmath25 . for @xmath26 and far enough from the boundaries of the disordered sample , the average intensity @xmath27 then obeys the diffusion equation @xcite , while the long - range correlation of intensity fluctuations @xmath28 can be found by solving the langevin equation @xcite : @xmath29 where @xmath30 , @xmath31 is the diffusion constant , and @xmath32 are random external langevin currents : @xmath33 the diagram corresponding to eq . ( [ current ] ) is shown in fig [ fig1](a ) . diagrams contributing to eqs . ( [ current ] ) and ( [ qcorr ] ) . solid lines denote the wave field @xmath34 and the complex conjugated field @xmath35 . dashed lines denote scattering of @xmath34 and @xmath35 on the same heterogeneity . the diagrams ( b ) and ( c ) are obtained by inserting two @xmath36 vertices ( denoted by wavy lines ) to the diagram ( a ) at ( @xmath37 ) and ( @xmath38 ) , respectively.,width=264 ] for a given @xmath39 , the current @xmath40 is a `` fingerprint '' of the disorder @xmath14 . an infinitesimal variation @xmath41 of the dielectric constant will modify @xmath40 by a small amount @xmath42 where the spatial integral is over the volume @xmath43 of the sample , we neglect the terms of the second and higher orders in @xmath41 , and the correlation of random response functions @xmath44 can be found by a functional differentiation of eq . ( [ current ] ) : @xmath45 . \label{qcorr}\end{aligned}\ ] ] here @xmath46 , @xmath47 , @xmath48 is assumed and @xmath49 is the green s function of eq . ( [ langevin ] ) . the diagrams contributing to eq . ( [ qcorr ] ) are shown in fig . [ fig1](b , c ) . in the stationary limit @xmath50 , eqs . ( [ dj ] ) and ( [ qcorr ] ) reduce to eqs . ( 6 ) and ( 7 ) of ref . . the langevin description of intensity fluctuations in disordered media can be extended to the nonlinear case ( @xmath51 ) . to this end , we consider time - independent @xmath52 : @xmath53 , and assume that the diffusion constant @xmath54 and the mean free path @xmath5 are not affected by the nonlinearity . the latter assumption is valid if @xmath55 @xcite , where @xmath56 is the typical value of the nonlinear correction to the refractive index and @xmath57 is the typical value of the average intensity in the medium . we now admit that in a nonlinear medium , the total dielectric constant contains a _ linear _ contribution @xmath58 that we assumed to be time - independent , and a _ nonlinear _ contribution @xmath59 that can vary with time . the variation of the total dielectric constant can be therefore only due its nonlinear part , and we can identify the infinitesimal variation of the dielectric constant @xmath60 in eq . ( [ dj ] ) with @xmath61 , where @xmath62 is the change of the intensity at @xmath63 during some infinitesimal time interval @xmath64 : @xmath65 . substituting @xmath66 into eq . ( [ dj ] ) , noting that @xmath67 , and dividing both sides of the resulting equation by @xmath68 , we obtain the following dynamic equation : @xmath69 where we make use of the fact that @xmath70 and hence @xmath71 $ ] , since @xmath27 is time - independent . equations ( [ langevin ] ) and ( [ djnl ] ) form a self - consistent system of equations : eq . ( [ langevin ] ) governs the spatio - temporal evolution of the intensity fluctuations @xmath72 due to the langevin currents @xmath32 , while eq . ( [ djnl ] ) describes the distributed feedback mechanism , leading to variations of @xmath32 depending on the changes of @xmath72 . note that eq . ( [ djnl ] ) is a linearized equation : only the terms linear in the nonlinear contribution to the dielectric constant @xmath59 are kept , which is justified as long as @xmath73 . in certain circumstances ( see below ) , the linearized nature of eq . ( [ djnl ] ) may result in the exponential growth of its solution with time , and in this sense eqs . ( [ langevin ] ) and ( [ djnl ] ) are analogous to the equations of linear stability analysis commonly used to study the stability of nonlinear systems ( see , e.g. , refs . and for examples of nonlinear optical systems exhibiting instabilities ) . hence , although eqs . ( [ langevin ] ) and ( [ djnl ] ) allow us to study the stability of the speckle pattern and the characteristic time scales of spontaneous intensity fluctuations beyond the instability threshold , they can not be used to determine the amplitude of these fluctuations . surface describing the stability of the multiple - scattering speckle pattern in a nonlinear disordered medium with open boundaries . at given effective nonlinearity parameter @xmath74 and frequency @xmath0 , the surface defines the lyapunov exponent @xmath75 . if @xmath76 , the speckle pattern is unstable with respect to periodic excitations at frequency @xmath0.,width=302 ] consider now an infinitesimal periodic excitation of the static speckle pattern : @xmath77 , where @xmath78 and @xmath79 . such an excitation can be either damped or amplified , depending on the sign of the lyapunov exponent @xmath75 . the value of @xmath75 is determined by two competing processes : on the one hand , diffusion tends to smear the excitation out , while on the other hand , the distributed feedback sustains its existence . the mathematical description of this competition is provided by eqs . ( [ langevin ] ) and ( [ djnl ] ) , that after the substitution of @xmath77 [ and similarly for @xmath40 ] lead to the following equation ( see appendix [ appa ] for the details of derivations ) : @xmath81 here @xmath7 is the effective nonlinearity parameter , the function @xmath82 is shown in fig . [ fig2 ] , and a numerical factor of order unity is omitted . to obtain eq . ( [ final ] ) , we have assumed the disordered sample to have open boundaries ( i.e. the diffusing wave leaves the sample when it reaches a boundary ) and have taken the limits of large sample size ( @xmath83 ) and moderate frequency @xmath84 ^ 2 $ ] , where @xmath85 is the inverse of the typical time needed for a multiple - scattered wave to diffuse through the disordered sample . it follows from fig . [ fig2 ] that for a given frequency @xmath0 the sign of the lyapunov exponent @xmath75 depends on the value of @xmath74 . excitations at frequencies @xmath0 corresponding to @xmath86 are damped exponentially and thus soon disappear . in contrast , excitations at frequencies @xmath0 corresponding to @xmath76 are exponentially amplified , which signifies the instability of the speckle pattern with respect to excitations at such frequencies . noting that @xmath75 is always negative for @xmath87 , we conclude that all excitation are damped in this case and the speckle pattern is absolutely stable . in an experiment , any spontaneous excitation of the static speckle pattern will be suppressed and the speckle pattern will be independent of time : @xmath88 , as in the linear case . when @xmath89 , an interval of frequencies @xmath90 starts to open up with @xmath76 . the speckle pattern thus becomes unstable with respect to excitations at low frequencies . in an experiment , any spontaneous excitation of the static speckle pattern at frequency @xmath91 will be amplified and one will observe a time - varying speckle pattern @xmath72 . main plot : frequency - dependent `` phase diagram '' of the multiple - scattering speckle pattern in a nonlinear disordered medium with open ( solid line ) or reflecting ( dashed line ) boundaries . for a given @xmath0 , @xmath74 should exceed the plotted threshold value for the instability to develop . inset : maximal lyapunov exponent as a function of the effective nonlinearity parameter @xmath74 . the dotted lines show @xmath92 and @xmath93.,width=302 ] the border between stable ( @xmath86 ) and unstable ( @xmath76 ) regimes is shown in fig . [ fig3 ] by a solid line . the instability threshold increases with @xmath0 . at @xmath94 the exact functional dependence of the threshold on @xmath0 is rather sensitive to the peculiarities of the disordered sample ( e.g. , its geometry and conditions on the boundaries ) , since for such slow oscillations the feedback mechanism is ensured by partial waves that have long path lengths @xmath95 and hence `` feel '' the presence of the boundaries and the shape of the sample . using the analytic expression for the function @xmath96 derived in appendix [ appa ] , we find @xmath97 with @xmath98 . this yields @xmath99 , and the shortest typical time scale @xmath100 of spontaneous intensity fluctuations can be estimated as @xmath101 . at high frequencies @xmath102 we find @xmath103 , @xmath104 , and @xmath105 , respectively . the latter results , on the contrary , are weakly sensitive to the peculiarities of the sample , since for the fast oscillations the feedback mechanism is due to relatively short diffusion paths that do not reach the boundaries of the sample . the rise of the instability threshold with @xmath0 can be qualitatively understood by considering the phase difference @xmath106 between two waves traveling through the disordered sample along the same diffusion path but separated in time by @xmath107 . if @xmath14 changes slowly with time , @xmath106 comprises two contributions : @xmath108 , which is the phase difference in a linear medium , and @xmath109 , which is the additional phase difference due to the nonlinearity . the second moment of the latter is @xmath110 where @xmath111 is a curvilinear coordinate of the point @xmath112 , the integrals are along the diffusion path of typical length @xmath113 , and @xmath114 denotes the change of the intensity at @xmath112 during the time @xmath115 . for @xmath94 we can assume that @xmath116 @xcite . taking @xmath117 to be the instability condition for the excitation of the speckle pattern at frequency @xmath0 , and noting that @xmath118 @xcite , we recover @xmath119 as the instability criterion . if @xmath102 , the long - range intensity correlation establishes only for @xmath120 , and the instability condition becomes @xmath121 . the positive sign of the maximal lyapunov exponent @xmath122 for @xmath89 ( solid line in the inset of fig . [ fig3 ] ) , as well as the continuous spectrum @xmath123 of frequencies with @xmath76 , are hallmarks of chaotic behavior @xcite . a sharp transition of the speckle pattern to chaos at @xmath93 is reminiscent of the behavior observed in nonlinear systems with large ( infinite ) number of degrees of freedom ( e.g. , random neural networks with an infinitely large number of nodes @xcite ) and should be contrasted to the `` route to chaos '' through a sequence of bifurcations , characteristic of low - dimensional nonlinear systems @xcite . as one can see from fig . [ fig3 ] , the scaling of @xmath124 with @xmath125 appears to be roughly linear for @xmath126 : @xmath127 with @xmath128 . to demonstrate the sensitivity of the results obtained for @xmath94 to the peculiarities of the disordered sample , we briefly consider a sample with reflecting boundaries ( dashed lines in fig . [ fig3 ] ) . all calculations can be carried out in the same way as for the sample with open boundaries ( see appendix [ appa ] ) , assuming that the green s function of eq . ( [ langevin ] ) is approximately the same as in the infinite medium : @xmath129 . we find that the absolute instability threshold @xmath130 is roughly 2 times lower than in the open geometry , @xmath131 , and @xmath132 for @xmath133 . for an arbitrary sample of disordered nonlinear medium we expect @xmath134 and @xmath135 for @xmath136 and @xmath137 , where @xmath138 , @xmath139 , and @xmath140 . by analogy @xcite with the theory of phase transitions , @xmath124 and @xmath141 can be identified with the order parameter and the critical exponent , respectively . finally , it is worthwhile to note that eqs . ( [ langevin ] ) and ( [ djnl ] ) can also be derived from a time - dependent disordered nonlinear schrdinger equation with a potential @xmath142 . upon the substitutions @xmath143 , @xmath144 , @xmath145 $ ] , and @xmath146 [ where @xmath147 is the energy of the incident schrdinger wave , @xmath148 is the particle mass , @xmath149 is the disordered potential , and @xmath150 is the nonlinear constant ] , our analysis is therefore valid in this case too . the analogy between the wave equation ( [ weq ] ) and the schrdinger equation is known for the stationary case , when the solution @xmath151 can be represented as @xmath152 . however , the dynamic solutions of the two equations differ due to different dispersion relations . although the present paper deals with dynamic speckle patterns , their temporal fluctuations are assumed to be slow and the analogy between the wave and schrdinger equations is recovered within the accuracy of our analysis . the author is grateful to r. maynard and b.a . van tiggelen for fruitful discussions and careful reading of the manuscript . a.yu . zyuzin is acknowledged for a communication explaining some details of ref . .
ultra - high energy ( @xmath6 ev ) cosmic rays ( uhecr ) are still remain a major scientific problem despite being studied worldwide by extensive air shower ( eas ) arrays for good 50 years . their mass composition is still not known exactly , and without this knowledge it is difficult to understand the character of nuclear interactions in this energy region . muons with energy near @xmath7 gev are very important component of eas . they are poorly moderated in the atmosphere , are sensitive to the characteristics of nuclear interactions during development of a shower and to the chemical composition of cosmic rays ( cr ) . due to their yield and properties of lateral distribution they can be effectively registered with widely spaced ground arrays . since 1978 yakutsk eas array has been continuously registering muons with the threshold energy @xmath8 gev . during this period a large amount of experimental data has been accumulated . analysis of this material @xcite has revealed that the development of showers with @xmath9 ev differs significantly from those at lesser energies of cr . it also allowed us to estimate the fraction of primary gamma - quanta in the total cr flux at energies above @xmath10 ev @xcite . m@xmath11 ) ; empty blue circles additional scintillation detectors ( @xmath12 m@xmath11 ) ; filled black squares underground muon detectors of @xmath13 m@xmath11 area and @xmath14 gev threshold ; empty black square underground muon detector of @xmath15 m@xmath11 and @xmath16 gev threshold.,scaledwidth=85.0% ] recently we have shown @xcite that the muon fraction in the total number of charged particles in eass with energies @xmath1 ev changes significantly over periods of time . until 1996 , it fluctuated around a single stable position and then increased significantly . this was accompanied by near simultaneous variations in the energy spectrum and in the global anisotropy of cr within energy range @xmath17 ev @xcite . after 1996 , during the next 7 years , the integral intensity of cr at @xmath18 ev increased by @xmath19% and then started declining . as for the phase of the first harmonic @xmath20 and its amplitude @xmath21 sampled during 1983 - 1994 , they changed to values @xmath22 and @xmath23 during 1998 - 2010 . in recent years , a tendency has been manifested towards the change of these values in the opposite direction . it seems like the effect of some gargantuan explosion which have contributed a significant portion of heavy nuclei to the background . it s difficult to pinpoint exactly what kind of event in the galaxy could led to such result . here we need further studies involving temporal factor of the experimental data . in this work we present the results of analyzes of muon data with @xmath8 gev threshold collected during the period from november 2011 to june 2013 . the geometry of the muon part of the array is shown on fig.[fig:1 ] . the work @xcite describes the technique of their control and calibration . the data from the detector with @xmath24 gev @xcite are currently being accumulated and will be analyzed later . ev and @xmath25 from primary protons ( blue circles ) and iron nuclei ( red squares ) obtained within the framework of the qgsjetii-04 @xcite model.,scaledwidth=85.0% ] further we consider eass with zenith angles @xmath26 and axes fallen within a @xmath27 km radius circle in the center of the array and with the precision of axis detection no less than @xmath28 m. the energy of primary particles was derived from relations : @xmath29 @xmath30 @xmath31 where @xmath32 is the density of charged particles as measured by surface scintillation detectors at @xmath33 m from a shower axis . the precision of @xmath34 estimation in individual showers was no worse than 10% . the relation ( [ eq:1 ] ) unambiguously connects the @xmath35 with @xmath36 at any given cr composition . it is due to the fact that at the distance @xmath37 m from the axis , the lateral distribution functions ( ldf ) of all charged particles inter - cross each other . it is demonstrated on fig.[fig:2 ] where two ldfs are shown , for charged particles in showers with @xmath38 ev and @xmath25 initiated by primary protons ( blue circles ) and iron nuclei ( red squares ) obtained with the use of qgsjetii-04 model @xcite . values for @xmath32 were derived from the modified linsley approximation @xcite : @xmath39 where @xmath40 is the molier radius which depends on air temperature ( @xmath41c ) and pressure ( @xmath42 , mbarn ) : @xmath43 the value for @xmath40 is measured in each individual event ( for yakutsk @xmath44c , @xmath45 m ) . in the expression ( [ eq:4 ] ) @xmath46 is the parameter defined in @xcite : @xmath47 ; * ( b ) * the steepness parameter @xmath48 of muon ldf ( @xmath49 gev ) . on both pictures theoretical predictions for primary protons ( p ) and iron nuclei ( fe ) are represented with red ( qgsjetii-04 ) , blue ( qgsjet01d ) and black ( sibyll-2.1 ) lines . squares experimental values.,scaledwidth=65.0% ] on fig.[fig:3](a ) with black squares are shown muon densities at the distance @xmath50 m from axis of showers within a considered group , with mean values of @xmath36 and @xmath25 . normalization to primary energy @xmath51 gives a representation of muon data that is more descriptive and convenient for further analysis . mean ldfs were obtained within energy bins with logarithmic step @xmath52 which were subsequently shifted towards higher energies by @xmath53 . this procedure was performed for a detailed test of an agreement between the experiment and various hadron interaction models . @xmath54 values were obtained from approximations of mean ldfs . when constructing an ldf , muon densities were multiplied by normalizing ratio @xmath55 and averaged over an energy cut in radial bins @xmath56 . mean muon densities were determined from the expression @xmath57 where @xmath58 and @xmath59 are the numbers of operated muon detectors at axis distances within the interval @xmath60 . the indexes denote whether a detectors had non - zero @xmath61 or zero @xmath62 readings during the registration of event . zero readings are related to cases when a detector hasnt registered any muons while being in a wait state . mean ldfs were approximated according to functions @xcite : @xmath63 with well - known relation by @xcite : @xmath64 where @xmath65 m , @xmath48 is a free parameter . the best fit values of @xmath46 , @xmath32 in ( [ eq:4 ] ) and @xmath48 , @xmath66 in ( [ eq:9 ] ) were determined with the use of @xmath67 minimization . error bars on fig.[fig:3](a ) include the entire combination originated from statistics of events and averaging of the data . lines represent expected values predicted by hadron interaction models qgsjetii-04 ( red ) , qgsjet01d @xcite ( blue ) and sibyll-2.1 @xcite ( black ) . simulations were performed with the use of corsika code @xcite ( version 6.990 in the case of sibyll-2.1 and qgsjet01d and 7.3700 in the case of qgsjetii-04 ) . 200 showers were simulated per each set of initial shower parameters ( mass of primary particle , energy and zenith angle ) . to speed - up the computations , the thin - sampling algorithm was activated in the corsika code with the parameters @xmath68 $ ] and @xmath69 $ ] depending on the primary energy @xcite . the density was calculated directly from total number of particles arrived at a detector of given area . it is clearly seen that the experiment is not consistent with sibyll at neither given composition ; the model predicts significantly less muon yield . other two models agree with our experiment much better and allow to estimate the mass composition of primary particles . to simplify , let us consider a two - component composition , consisting of protons and iron nuclei . in this case the relation @xmath70 gives weighting functions @xmath71 and @xmath72 . within the framework of this hypothesis , according to the qgsjet01d model we have : @xmath73 where @xmath74 are the values obtained in the experiment ( exp ) and in simulation . with red and blue triangles on fig.[fig:4 ] are shown energy dependencies of cr mass composition according to predictions of qgsjetii-04 and qgsjet01d correspondingly . for comparison , on the same figure the data from various eas experiments are shown . purple stars denote the results of kascade obtained during the period from may 1998 to december 1999 @xcite . white diamonds represent the data of tunka-133 obtained from the ldf of cherenkov radiation emitted by eas during two winter observational periods ( 2009 - 2011 ) @xcite . other values were derived from @xmath75 dependence established in experiments and simulations within the framework of qgsjet01 model , with the use of the expression ( [ eq:11 ] ) with substitution @xmath76 . with green triangles are shown hires data related to the observational period from november 1999 to september 2001 @xcite . black circles results from pao obtained between december 2004 and september 2010 @xcite , open circles interpretation of the telescope array data within the framework of qgsjetii model . gev threshold obtained with the use of qgsjet01d model for primary protons with primary energy @xmath77 ev at @xmath25 . the line represents approximation ( [ eq:9 ] ) with @xmath78.,scaledwidth=85.0% ] results presented on fig.[fig:4 ] give evidence that in energy region @xmath79 ev the cr composition probably changes rapidly towards lighter nuclei . our data do not contradict this scenario . it is seen that qgsjetii-04 model better agrees with the experiment than qgsjet01d , especially in view of the works @xcite , which have demonstrated that at @xmath80 ev the mass composition changes rapidly . yet , it is still a bit premature to make a strong conclusion both models predict similar shapes of muon ldfs and agree with experiment , as evidenced by the parameter of ldf s steepness @xmath48 shown on fig.[fig:3](b). experimental values of this parameter confirm the above - mentioned hypothesis about lightening of the cr mass composition with the increase of primary energy . the ldfs obtained in simulations are well - described by the function ( [ eq:8 ] ) in a wide range of distances from a shower axis . it is demonstrated on fig.[fig:5 ] where points represent ldf for muons with @xmath81 gev threshold obtained with the use of qgsjet01d model for primary protons with @xmath38 ev and @xmath25 . line is the approximation of ( [ eq:9 ] ) with steepness parameter @xmath82 found with weighted least square method . similar picture is observed at other values of initial shower parameters ( i.e. energy , zenith angle and composition ) . and muon density ( at @xmath83 ) normalized to primary energy ( @xmath84 ) . predictions of qgsjetii-04 ( red squares ) , qgsjet01d ( blue circles ) and sibyll-2.1 ( black triangles ) for primary protons ( open symbols ) and iron nuclei ( filled symbols ) within the energy range @xmath85 ev at @xmath25 . lines denote linear approximations.,scaledwidth=85.0% ] our estimations of @xmath86 based on the ( [ eq:10 ] ) do not differ drastically from other , more traditional method based on the @xmath87 ( see e.g. , @xcite ) : @xmath88 it is due to some peculiarity of eas muons , which is demonstrated on fig.[fig:6 ] . it shows that within the framework of any shower development model , between the depth of maximum @xmath87 and the logarithm of muon density normalized to @xmath36 , there is a quasi - linear dependence at any composition of primary particles : @xmath89 if one inserts ( [ eq:13 ] ) into ( [ eq:12 ] ) , then ( [ eq:10 ] ) follows . the relations visualized on fig.[fig:6 ] allow one to find @xmath90 from muon density normalized to primary energy . in our case , from the plot on this figure follow the dependencies shown on fig.[fig:7 ] . this method of extracting the depth of maximum from muon data , in our opinion , is comfortably simple and efficient from practical point of view . with sufficient statistics it is essentially a fully functional alternative to a technique of @xmath90 determination from the ldf of cherenkov light radiation . it is possible to use different muon thresholds and distances from shower axis , where muon density is sampled . according to qgsjetii-04 ( red lines ) and qgsjet01d ( blue line ) for primary protons ( p ) and iron nuclei ( fe ) . red stars and blue crosses represent the interpretation of experimental data within the frameworks of the two models correspondingly.,scaledwidth=85.0% ] the comparison of the yakutsk array data to modern ultra - high energy interaction models has demonstrated once again the importance that this component presents for studying of shower development and cr mass composition . the results from fig.[fig:3 ] demonstrate a certain degree of agreement between the experiment and qgsjetii-04 and qgsjet01d models on the whole energy range @xmath91 ev . estimations of the mean cr mass composition from fig.[fig:4 ] obtained within the frameworks of these models are close enough to worldwide data @xcite . the qgsjetii-04 model agrees with them better than qgsjet01d . our data displayed on fig.[fig:4 ] have added to the whole picture that testifies of a rapid change in the cr composition towards lighter nuclei in energy range @xmath92 ev . it can not be ruled out that at @xmath93 there is a peak of local heaviness of cr composition . it is seen from fig.[fig:5 ] that ldfs obtained from simulations are well - described by the approximation ( [ eq:8 ] ) within a wide range of distances from shower axis . this approximation is convenient for comparison between the experimental data and theoretical predictions as seen on fig.[fig:3](b). fig.[fig:6 ] suggests that muon component contains some possibilities yet to discover . in particular , it allows to relatively easy determine the depth of maximum eas development ( see fig.[fig:7 ] ) . we suppose that muons can play a vital role in the energy cross - calibration between world s eas arrays , where there is still no clarity in relation to correctness of any given method of estimating the energy of primary particle . the work was conducted with the financial support from russian academy of science within the framework of research program `` fundamental properties of matter and astrophysics '' and is supported by rfbr grant # 130212036 ofi - m-2013 .	the lateral distribution of muons with threshold energy @xmath0 gev have been studied in showers with energy @xmath1 ev . the data considered in the analysis have been collected from november 2011 to june 2013 . experimental values are compared to predictions obtained with the use of corsika code within the framework of various hadron interaction models . the best agreement between experiment and theory is observed for qgsjetii-04 model . at @xmath2 ev it complies to a mixed cosmic ray composition with the mean atomic number @xmath3 . at @xmath4 ev the composition varies around the value @xmath5 .
let @xmath6 be a connected complex algebraic group . set @xmath7 . then @xmath8 $ ] and @xmath9^{{\mathfrak}q}={{\mathbb c}}[{\mathfrak}q^]^q$ ] . we will call the latter object the _ algebra of symmetric invariants _ of @xmath4 . an important property of @xmath10 is that it is isomorphic to @xmath11 as an algebra by a classical result of m.duflo ( here @xmath11 is the centre of the universal enveloping algebra of @xmath4 ) . let @xmath0 be a reductive lie algebra . then by the chevalley restriction theorem @xmath12 $ ] is a polynomial ring ( in @xmath13 variables ) . a quest for non - reductive lie algebras with a similar property has recently become a trend in invariant theory . here we consider finite - dimensional representations @xmath14 of @xmath0 and the corresponding semi - direct products @xmath2 . the lie bracket on @xmath4 is defined by @xmath15=[\xi,\eta]+\rho(\xi)u-\rho(\eta)v\ ] ] for all @xmath16 , @xmath17 . let @xmath18 be a connected simply connected lie group with @xmath19 . then @xmath7 with @xmath20 . it is easy to see that @xmath21^g\subset { { \mathbb c}}[{\mathfrak}q^]^{{\mathfrak}q}$ ] and therefore @xmath21^g$ ] must be a polynomial ring if @xmath3^{{\mathfrak}q}$ ] is , see ( * ? ? ? * section 3 ) . classification of the representations of complex simple algebraic groups with free algebras of invariants was carried out by g.schwarz @xcite and independently by o.m.adamovich and e.o.golovina @xcite . one such representation is the spin - representation of @xmath22 , which leads to @xmath23 . here @xmath3^{{\mathfrak}q}$ ] is a polynomial ring in @xmath24 variables generated by invariants of bi - degrees @xmath25 , @xmath26 , @xmath27 with respect to the decomposition @xmath28 , see ( * ? ? ? * proposition 3.10 ) . in this paper , we treat another example , @xmath29 , @xmath30 with @xmath31 , @xmath32 , @xmath33 . here @xmath3^{{\mathfrak}q}$ ] is a polynomial ring in exactly the following three cases : * @xmath34 , @xmath35 , and @xmath36 with @xmath37 ; * @xmath38 , @xmath39 ; * @xmath40 and @xmath41 divides @xmath42 . we also briefly discuss semi - direct products arising as @xmath43-contractions of reductive lie algebras . * acknowledgements . * i would like to thank the organisers of the intensive period on perspectives in lie theory " , especially giovanna carnovale and martina lanini , for the invitation to pisa and a very warm welcome . let @xmath7 be an algebraic lie algebra , @xmath6 a connected algebraic group . the index of @xmath4 is defined as @xmath44 where @xmath45 is the stabiliser of @xmath46 in @xmath4 . in view of rosenlicht s theorem , @xmath47 . in case @xmath48 , we have @xmath3^{{\mathfrak}q}={{\mathbb c}}$ ] . for a reductive @xmath0 , @xmath49 . recall that @xmath50 is the dimension of a borel subalgebra of @xmath0 . for @xmath4 , set @xmath51 . let @xmath52 be a basis of @xmath4 and @xmath53)$ ] the structural matrix with entries in @xmath4 . this is a skew - symmetric matrix of rank @xmath54 . let us take pfaffians of the principal minors of @xmath55 of size @xmath56 and let @xmath57 be their greatest common divisor . then @xmath58 is called the _ fundamental semi - invariant _ of @xmath4 . the zero set of @xmath58 is @xmath59 the so called _ singular set _ of @xmath4 . since @xmath60 is clearly a @xmath6-stable subset , @xmath58 is indeed a semi - invariant , @xmath61 . one says that @xmath4 has the codim-2 " property ( satisfies the codim-2 " condition ) , if @xmath62 or equivalently if @xmath63 . suppose that @xmath64 are homogenous algebraically independent polynomials . the _ jacobian locus _ @xmath65 of these polynomials consists of all @xmath66 such that the differentials @xmath67 are linearly dependent . in other words , @xmath68 if and only if @xmath69 . the set @xmath70 is a proper zariski closed subset of @xmath71 . suppose that @xmath65 does not contain divisors . then by the characteristic zero version of a result of skryabin , see ( * ? ? ? * theorem 1.1 ) , @xmath72 $ ] is an algebraically closed subalgebra of @xmath73 , each @xmath74 that is algebraic over @xmath75 is contained in @xmath72 $ ] . [ thm - sum ] suppose that @xmath76 and suppose that @xmath77 are homogeneous algebraically independent polynomials such that @xmath78 and @xmath79 . then @xmath80 $ ] is a polynomial ring in @xmath81 generators . under our assumptions @xmath82 , see ( * ? ? ? * theorem 1.2 ) and ( * ? ? ? * section 2 ) . therefore @xmath83 $ ] is an algebraically closed subalgebra of @xmath73 by ( * ? ? ? * theorem 1.1 ) . since @xmath84 , each symmetric @xmath4-invariant is algebraic over @xmath83 $ ] and hence is contained in it . for semi - direct products , we have some specific approaches to the symmetric invariants . suppose now that @xmath85 is a reductive lie algebra , no non - zero ideal of @xmath0 acts on @xmath1 trivially , @xmath18 is connected , and @xmath2 , where @xmath1 is a finite - dimensional @xmath18-module . the vector space decomposition @xmath86 leads to @xmath87 , where we identify @xmath0 with @xmath88 . each element @xmath89 is considered as a point of @xmath71 that is zero on @xmath0 . we have @xmath90 , where each element of @xmath91 is zero on @xmath1 . note that @xmath92 and @xmath93 is equal to @xmath94 . therefore @xmath95 . the decomposition @xmath86 defines also a bi - grading on @xmath73 and clearly @xmath10 is a bi - homogeneous subalgebra , cf . * lemma 2.12 ) . a statement is true for a generic x " if and only if this statement is true for all points of a non - empty open subset . [ v - inv ] a function @xmath96 $ ] is a @xmath1-invariant if and only if @xmath97 for generic @xmath89 and any @xmath98 . condition of the lemma guaranties that for each @xmath99 , @xmath100 on a non - empty open subset of @xmath71 . hence @xmath101 is a @xmath1-invariant . for @xmath89 , let @xmath102^q\to { { \mathbb c}}[{\mathfrak}g{+}x]^{g_x{\ltimes}\exp(v)}$ ] be the restriction map . by ( * ? * lemma 2.5 ) @xmath103^{g_x{\ltimes}\exp(v)}\cong { { \mathcal s}}({\mathfrak}g_x)^{g_x}$ ] . moreover , if we identify @xmath104 with @xmath0 choosing @xmath105 as the origin , then @xmath106 for any @xmath4-invariant @xmath101 ( * ? ? ? * section 2 ) . under certain assumptions on @xmath18 and @xmath1 the restriction map @xmath107 is surjective , more details will be given shortly . there is a non - empty open subset @xmath108 such that the stabilisers @xmath109 and @xmath110 are conjugate in @xmath18 for any pair of points @xmath111 see e.g. ( * ? ? ? * theorem 7.2 ) . any representative of the conjugacy class @xmath112 is said to be a _ a generic stabliser _ of the @xmath18-action on @xmath113 . there is one easy to handle case , @xmath114 for a generic @xmath89 . here c}}[v^]^{g}$ ] , see e.g. ( ? ? ? * example 3.1 ) , and @xmath115 only if @xmath116 , where @xmath98 , @xmath117 . the case @xmath118 is more involved . [ index-1-sur ] assume that @xmath18 has no proper semi - invariants in @xmath21 $ ] . suppose that @xmath118 , @xmath119 , and the map @xmath107 is surjective for generic @xmath89 . then @xmath3^{{\mathfrak}q}={{\mathbb c}}[v^]^g[f]$ ] , where @xmath101 is a bi - homogeneous preimage of a generator of @xmath120 that is not divisible by any non - constant @xmath18-invariant in @xmath21 $ ] . if we have a lie algebra of index @xmath121 , in our case @xmath122 , then the algebra of its symmetric invariants is a polynomial ring . there are many possible explanations of this fact . one of them is the following . suppose that two non - zero homogeneous polynomials @xmath123 are algebraically dependent . then @xmath124 for some coprime integers @xmath125 and some @xmath126 . if @xmath127 is an invariant , then so is a polynomial function @xmath128{f_1}=\sqrt[ab]{c}\sqrt[a]{f_2}$ ] . since @xmath119 , it is generated by some homogeneous @xmath129 . the group @xmath109 has finitely many connected components , hence @xmath120 is generated by a suitable power of @xmath129 , say @xmath130 . let @xmath96^q$ ] be a preimage of @xmath131 . each its bi - homogeneous component is again a @xmath4-invariant . without loss of generality we may assume that @xmath101 is bi - homogenous . also if @xmath101 is divisible by some non - scalar @xmath132^g$ ] , then we replace @xmath101 with @xmath133 and repeat the process as long as possible . whenever @xmath110 ( with @xmath117 ) is conjugate to @xmath109 and @xmath134 , @xmath135 is a @xmath110-invariant of the same degree as @xmath131 and therefore is a generator of @xmath136 . clearly @xmath137\subset { { \mathbb c}}[{\mathfrak}q^]^q{\otimes}_{{{\mathbb c}}[v^]^g}{{\mathbb c}}(v^)^g=:{\mathcal a}$ ] and @xmath138 . if @xmath139 contains a homogeneous in @xmath0 polynomial @xmath140 that is not proportional ( over @xmath141 ) to a power of @xmath101 , then @xmath142 is not proportional to a power of @xmath143 for generic @xmath144 . but @xmath145 . this implies that @xmath146 $ ] . it remains to notice that @xmath147^g$ ] , since @xmath18 has no proper semi - invariants in @xmath21 $ ] , and by the same reason @xmath137\cap { { \mathbb c}}[{\mathfrak}q]={{\mathbb c}}[v^]^g[f]$ ] in case @xmath101 is not divisible by any non - constant @xmath18-invariant in @xmath21 $ ] . it is time to recall the ras formula @xcite for the index of a semi - direct product : @xmath148 [ v - deg ] suppose that @xmath149 are homogenous polynomials such that @xmath150 with @xmath151 freely generate @xmath152 for generic @xmath89 and @xmath153^g$ ] for @xmath154 ; and suppose that @xmath155 . then @xmath156 if and only if @xmath157 . in view of the assumptions , we have @xmath158 . further , by eq . @xmath159 the result follows . from now on suppose that @xmath18 is semisimple . then both @xmath18 and @xmath6 have only trivial characters and hence can not have proper semi - invariants . in particular , the fundamental semi - invariant is an invariant . we also have @xmath160 . set @xmath78 and let @xmath89 be generic . if @xmath3^q$ ] is a polynomial ring , then there are bi - homogenous generators @xmath161 such that @xmath162 with @xmath163 freely generate @xmath21^g$ ] and the invariants @xmath162 with @xmath151 are _ mixed _ , they have positive degrees in @xmath0 and @xmath1 . [ svob ] suppose that @xmath18 is semisimple and @xmath3^{{\mathfrak}q}$ ] is a polynomial ring with homogeneous generators @xmath164 . then @xmath165 ; * for generic @xmath89 , the restriction map @xmath166^q \to { { \mathbb c}}[{\mathfrak}g{+}x]^{g_x{\ltimes}v}\cong{{\mathcal s}}({\mathfrak}g_x)^{g_x}$ ] is surjective , @xmath152 , and @xmath120 is a polynomial ring in @xmath167 variables . it is worth mentioning that @xmath107 is also surjective for stable actions . an action of @xmath18 on @xmath1 is called _ stable _ if generic @xmath18-orbits in @xmath1 are closed , for more details see ( * ? ? ? * sections 2.4 & 7.5 ) . by ( * theorem 2.8 ) @xmath107 is surjective for generic @xmath89 if the @xmath18-action on @xmath113 is stable . the initial motivation for studying symmetric invariants of semi - direct products was related to a conjecture of d.panyushev on @xmath169-contractions of reductive lie algebras . the results of @xcite , briefly outlined in section [ sec-2 ] , has settled the problem . let @xmath170 be a _ symmetric decomposition _ , i.e. , a @xmath168-grading of @xmath0 . a semi - direct product , @xmath171 , where @xmath172 is an abelian ideal , can be seen as a _ contraction _ , in this case a _ @xmath169-contraction _ , of @xmath0 . for example , starting with a symmetric pair @xmath173 , one arrives at @xmath174 . in @xcite , it was conjectured that @xmath175 is a polynomial ring ( in @xmath13 variables ) . let @xmath176 be a @xmath169-contraction of a reductive lie algebra @xmath0 . then @xmath177 is a polynomial ring ( in @xmath13 variables ) if and only if the restriction homomorphism @xmath178^{{\mathfrak}g}\to { { \mathbb c}}[{\mathfrak}g_1]^{{\mathfrak}g_0}$ ] is surjective . if we are in one of the surjective " cases , then one can describe the generators of @xmath175 . let @xmath164 be suitably chosen homogeneous generators of @xmath179 and let @xmath180 be the bi - homogeneous ( w.r.t . @xmath170 ) component of @xmath162 of the highest @xmath181-degree . then @xmath175 is freely generated by the polynomials @xmath180 ( of course , providing the restriction homomorphism @xmath178^{{\mathfrak}g}\to { { \mathbb c}}[{\mathfrak}g_1]^{{\mathfrak}g_0}$ ] is surjective ) @xcite . unfortunately , this construction of generators can not work if the restriction homomorphism is not surjective , see ( * ? ? * remark 4.3 ) . as was found out by s.helgason @xcite , there are four non - surjective " irreducible symmetric pairs , namely , @xmath182 , @xmath183 , @xmath184 , and @xmath185 . the approach to semi - direct products developed in @xcite showed that panyushev s conjecture does not hold for them . next we outline some ideas of the proof . let @xmath186 be a connected subgroup with @xmath187 . then @xmath188 is reductive , it acts on @xmath189 , and this action is stable . let @xmath190 be a generic element and @xmath191 be its stabiliser in @xmath188 . the groups @xmath191 are reductive and they are known for all symmetric pairs . in particular , @xmath192 is a polynomial ring . it is also known that @xmath193^{g_0}$ ] is a polynomial ring . by @xcite @xmath194 has the codim-2 " property and @xmath195 . making use of the surjectivity of @xmath107 one can show that if @xmath196^{{\tilde{\mathfrak g}}}$ ] is freely generated by some @xmath164 , then necessary @xmath197 for @xmath194 coming from one of the non - surjective " pairs @xcite . in view of some results from @xcite this leads to a contradiction . note that in case of @xmath198 , @xmath199 is simple and @xmath194 is a semi - direct product of @xmath200 and @xmath201 , which , of course , comes from one of the representations in schwarz s list @xcite . form now assume that @xmath202 and @xmath203 with @xmath31 , @xmath32 , @xmath33 . according to @xcite @xmath204^g$ ] is a polynomial ring if either @xmath34 and @xmath35 or @xmath205 , @xmath206 . one finds also the description of the generators of @xmath21^g$ ] and their degrees in @xcite . in this section , we classify all cases , where @xmath3^{{\mathfrak}q}$ ] is a polynomial ring and for each of them give the fundamental semi - invariant . [ ex-0 ] suppose that either @xmath207 or @xmath208 . then @xmath114 for generic @xmath89 and therefore @xmath3^q={{\mathbb c}}[v^]^g$ ] , i.e. , @xmath3^q$ ] is a polynomial ring if and only if @xmath209^g$ ] is . the latter takes place for @xmath210 , for @xmath211 and any @xmath212 , as well as for @xmath208 . non - scalar fundamental semi - invariants appear here only for here the ring of @xmath18-invariants on @xmath113 is generated by @xmath218{vp}},\ ] ] where each @xmath219 is the determinant of the corresponding submatrix of @xmath220 . the generators are algebraically independent if and only if @xmath221 , see also @xcite . we are interested only in @xmath222 that are smaller than @xmath223 . let @xmath224 , where @xmath225 , and let @xmath226 be a subset of cardinality @xmath81 . by choosing the corresponding @xmath81 columns of @xmath227 we get a matrix @xmath228 . set @xmath229 clearly each @xmath230 is an @xmath231-invariant . below we will see that they are also @xmath1-invariants . if @xmath232 , then there is just one invariant , @xmath233 . if @xmath81 is either @xmath121 or @xmath234 , we get @xmath222 invariants . according to lemma [ v - inv ] we have to show that @xmath235 for generic @xmath89 and any @xmath236 . since @xmath237 , there is an open @xmath231-orbit in @xmath113 and we can take @xmath105 as @xmath238 . let @xmath239 be the standard parabolic subalgebra corresponding to the composition @xmath240 and let @xmath241 be the nilpotent radical of the opposite parabolic . each element ( matrix ) @xmath242 is a sum @xmath243 with @xmath244 , @xmath245 . in this notation @xmath246 . let @xmath247 and @xmath248 be @xmath249 and @xmath250-submatrices of @xmath251 standing in the upper left and lower right corner , respectively . then @xmath252 . each column of @xmath253 is a linear combination of columns of @xmath254 and each column of @xmath255 is a linear combination of columns of @xmath256 . therefore @xmath257 notice that @xmath258 and the nilpotent radical of @xmath259 is contained in @xmath122 ( with @xmath260 ) . since @xmath261 ( after the identification @xmath262 ) , @xmath263 for any @xmath264 ; and we have @xmath265 with @xmath266 for this @xmath251 . an easy observation is that @xmath267 hence @xmath268 for all @xmath269 and all @xmath236 . [ th-0 ] suppose that @xmath270 . then @xmath3^q$ ] is a polynomial ring if and only if @xmath221 and @xmath222 divides either @xmath271 , @xmath223 or @xmath272 . under these assumptions on @xmath222 , @xmath76 exactly then , when @xmath222 divides either @xmath271 or @xmath272 . note that the statement is true for @xmath207 by example [ ex-0 ] . assume that @xmath273 . suppose that @xmath274 as above . a generic stabiliser in @xmath0 is @xmath275 . on the group level it is connected . notice that @xmath276 , since @xmath109 has no non - trivial characters . note also that @xmath21^g={{\mathbb c}}$ ] , since @xmath277 . if @xmath3^q$ ] is a polynomial ring , then so is @xmath278^{g_x}$ ] by theorem [ svob](ii ) and either @xmath279 or , arguing by induction , @xmath280 with @xmath37 . next we show that the ring of symmetric invariants is freely generated by the polynomials @xmath230 for the indicated @xmath222 . each element @xmath281 can be presented as @xmath282 , where @xmath283 . each restriction @xmath284 can be regarded as an element of @xmath285 . eq . combined with lemma [ f - v ] and the observation that @xmath286 shows that @xmath284 is either @xmath287 of @xmath122 ( in case @xmath288 , where @xmath289 ) or @xmath230 of @xmath122 . arguing by induction on @xmath223 , we prove that the restrictions @xmath284 freely generate @xmath290 for @xmath260 ( i.e. , for a generic point in @xmath113 ) . notice that @xmath291 . the group @xmath231 acts on @xmath113 with an open orbit @xmath292 . therefore the restriction map @xmath107 is injective . by the inductive hypothesis it is also surjective and therefore is an isomorphism . this proves that the polynomials @xmath230 freely generate @xmath3^q$ ] . suppose that @xmath222 divides either @xmath294 or @xmath295 . then we have @xmath222 different invariants @xmath230 . by induction on @xmath223 , @xmath122 has the codim-2 " property , therefore the sum of @xmath296 is equal to @xmath297 by theorem [ svob](i ) . the sum of @xmath1-degrees is @xmath298 and hence by lemma [ v - deg ] @xmath299 . thus , @xmath4 has the codim-2 " property . here @xmath21^g$ ] is a polynomial ring if and only if @xmath206 ; a generic stabiliser is @xmath300 , and the @xmath18-action on @xmath301 is stable . we assume that @xmath217 for generic @xmath89 and therefore @xmath302 . for an @xmath303-matrix @xmath304 , let @xmath305 with @xmath306 be coefficients of its characteristic polynomial , each @xmath307 being a homogeneous polynomial of degree @xmath308 . let @xmath309 with @xmath310 , @xmath311 , @xmath312 . having these objects we form an @xmath313-matrix @xmath314 and set @xmath315 for each @xmath316 . each @xmath317 is an @xmath318-invariant . unfortunately , these polynomials are not @xmath1-invariants . if we repeat the same construction for @xmath319 with @xmath206 , then @xmath320^{\tilde q}={{\mathbb c}}[v^]^{{{\rm gl}}_n}[\{f_i\mid 2k{+}1{\leqslant}i{\leqslant}n{+}k\}]$ ] and it is a polynomial ring in @xmath321 generators . suppose that @xmath323 . then a generic stabiliser @xmath324 is of index @xmath121 and since the @xmath18-action on @xmath1 is stable , @xmath3^{{\mathfrak}q}$ ] has to be a polynomial ring by ( * ? ? ? * example 3.6 ) . one can show that the unique mixed generator is of the form @xmath325 , where @xmath326 is a certain @xmath318-invariant on @xmath1 of degree @xmath327 and then see that the sum of degrees is @xmath328 . more generally , @xmath4 has the codim-2 " property for all @xmath302 . here each @xmath18-invariant divisor in @xmath113 contains a @xmath18-orbit of maximal dimension , say @xmath329 . set @xmath330 . if @xmath110 is not @xmath331 , then @xmath332 is a semi - direct product with a heisenberg lie algebra . following the proof of ( * theorem 3.3 ) , one has to show that @xmath333 in order to prove that @xmath4 has the codim-2 " property . this is indeed the case , @xmath334 . suppose that @xmath335 and assume that @xmath10 is a polynomial ring . then there are bi - homogeneous generators @xmath336 of @xmath3^q$ ] over @xmath21^g$ ] such that their restrictions to @xmath104 form a generating set of @xmath290 for a generic @xmath105 ( with @xmath337 ) , see theorem [ svob](ii ) . in particular , @xmath338 . take @xmath339 , which is a @xmath169-contraction of @xmath340 . then @xmath4 is a lie subalgebra of @xmath194 . note that @xmath341 acts on @xmath4 via automorphisms and therefore we may assume that the @xmath342-linear span of @xmath343 is @xmath341-stable . by degree considerations , each @xmath344 is an @xmath345-invariant as well . the weyl involution of @xmath231 acts on @xmath1 and has to preserve each line @xmath346 . since this involution interchanges @xmath347 and @xmath348 , each @xmath344 is also a @xmath341-invariant . thus , @xmath349 since @xmath194 is a surjective " @xmath169-contraction , its symmetric invariants are known ( * ? ? ? * theorem 4.5 ) . the generators of @xmath350 are @xmath351 with @xmath352 . here @xmath353 and the generators of ( @xmath354)-degrees @xmath355 are @xmath356 . as the restriction to @xmath357 shows , none of the generators @xmath351 with @xmath358 lies in @xmath73 . this means that @xmath344 can not be equal or even proportional over @xmath21^g$ ] to @xmath359 and hence has a more complicated expression . more precisely , a product @xmath360 necessary appears in @xmath344 with a non - zero coefficient from @xmath21^g$ ] for @xmath361 . since @xmath362 , we have @xmath363 for every @xmath361 . the ring @xmath21^g$ ] is freely generated by @xmath364 polynomials of degree two . therefore , the total sum of degrees over all generators of @xmath10 is greater than or equal to @xmath365 this contradicts theorem [ svob](i ) in view of the fact that @xmath76 . let @xmath368 be the fundamental weights of @xmath5 . we use the standard convention , @xmath369 , @xmath370 . recall that for any @xmath371 , @xmath372 , @xmath373 is irreducible with the highest weight @xmath374 . let @xmath375 be a basis of @xmath347 such that each @xmath376 is a weight vector and @xmath377 is a highest weight vector of @xmath373 . clearly @xmath378 . write @xmath379 with @xmath380 . let @xmath381 be a non - zero @xmath222-linear @xmath18-equivariant map . such a map is unique up to a scalar and one can take @xmath382 with @xmath383 . in case @xmath384 , for any subset @xmath385 with @xmath386 , let @xmath387 be the corresponding ( almost ) canonical map . by the same principle we construct @xmath388 . let us consider the tensor product @xmath389 and its weight subspace @xmath390 . one can easily see that @xmath390 contains a unique up to a scalar non - zero highest weight vector , namely @xmath391 this means that @xmath392 contains a unique copy of @xmath393 , where @xmath393 is an irreducible @xmath5-module with the highest weight @xmath394 . we let @xmath395 denote the representation of @xmath396 on @xmath397 and @xmath398 the representation of @xmath396 on @xmath399 . let @xmath400 be a point in @xmath71 . ( it is assumed that @xmath401 . ) finally let @xmath402 denote a non - zero scalar product between @xmath392 and @xmath403 that is zero on the @xmath5-invariant complement of @xmath393 in @xmath392 . depending on @xmath81 , set @xmath404 for each @xmath405 as above in case @xmath406 . by the constructions the polynomials @xmath407 and @xmath408 are @xmath231-invariants . we restrict @xmath407 and @xmath408 to @xmath409 with @xmath89 generic . changing a basis in @xmath1 if necessary , we may assume that @xmath410 . if @xmath411 , some of the invariants @xmath408 may become linear combinations of such polynomials under the change of basis , but this does not interfere with @xmath1-invariance . now @xmath412 is a vector of weight @xmath413 and @xmath414 of weight @xmath415 . notice that @xmath416 . if @xmath417 and each @xmath418 is one of the @xmath419 , @xmath420 , then in the sequence @xmath421 we must have exactly one @xmath419 for each @xmath422 and @xmath423 copies of each @xmath424 with @xmath425 . hence the only summand of @xmath426 that plays any rle in @xmath407 or @xmath408 is @xmath427 . moreover , in @xmath428 we are interested only in vectors lying in @xmath429 . then up to a non - zero scalar @xmath437 is the determinant of @xmath438 each column of @xmath439 is a linear combination of the columns of @xmath433 , a similar relation exists between @xmath440 and @xmath441 . therefore @xmath442 we have to check that @xmath443 for any @xmath444 and any @xmath98 , see lemma [ v - inv ] . recall that @xmath261 . in case @xmath445 , @xmath433 is zero in each @xmath446 and @xmath431 corresponding to such @xmath251 is a scalar matrix . therefore @xmath447 . the case @xmath411 is more complicated . if @xmath448 , then @xmath449 . let @xmath450 be the corresponding submatrix of @xmath433 and @xmath451 of @xmath432 . one just has to replace @xmath433 by @xmath450 and @xmath432 by @xmath451 in the last polynomial @xmath452 obtaining @xmath453 these are @xmath454 linearly independent invariants in @xmath285 . [ divides ] suppose that @xmath459 and @xmath41 divides @xmath42 , then @xmath118 for generic @xmath89 and @xmath3^q={{\mathbb c}}[v^]^g[{\bf f}]$ ] is a polynomial ring , the fundamental semi - invariant is equal to @xmath460 . a generic stabilser @xmath122 is @xmath461 . its ring of symmetric invariants is generated by @xmath462 , see theorem [ th-0 ] and eq . . we also have @xmath118 . it remains to see that @xmath407 is not divisible by a non - constant @xmath18-invariant polynomial on @xmath113 . by the construction , @xmath407 is also invariant with respect to the action of @xmath463 . the group @xmath464 act on @xmath113 with an open orbit . as long as @xmath465 , @xmath466 , the @xmath467-obit of @xmath468 contains a point @xmath469 , where also @xmath470 . if in addition the upper @xmath471-part of @xmath227 has rank @xmath436 , then the @xmath472 contains @xmath410 . here @xmath407 is non - zero on @xmath473 . since the group @xmath467 is semisimple , the complement of @xmath474 contains no divisors and @xmath407 is not divisible by any non - constant @xmath18-invariant in @xmath21 $ ] . this is enough to conclude that @xmath3^q={{\mathbb c}}[v^]^g[{\bf f}]$ ] , see theorem [ index-1-sur ] . the reason for this misfortune is that @xmath478 for @xmath411 . one could prove that each @xmath408 must be in the set of generators and thereby show that @xmath3^q$ ] is not a polynomial ring . but we present a different argument . assume that the ring of symmetric invariants is polynomial . it is bi - graded and @xmath479 acts on it preserving the bi - grading . since @xmath479 is reductive , we can assume that there is a set @xmath480 of bi - homogeneous mixed generators such that @xmath481^g[h_1,\ldots , h_s]$ ] and the @xmath342-linear span @xmath482 is @xmath479-stable . the polynomiality implies that a generic stabiliser @xmath483 has a free algebra of symmetric invariants , see theorem [ svob](ii ) , and by the same statement @xmath107 is surjective . this means that @xmath81 is either @xmath121 or @xmath484 , see theorem [ th-0 ] , @xmath485 , and @xmath107 is injective on @xmath486 . taking our favourite ( generic ) @xmath410 , we see that there is @xmath487 embedded diagonally into @xmath488 , which acts on @xmath489 as on @xmath490 . the group @xmath487 acts on @xmath486 in the same way . since @xmath41 does not divide @xmath42 , we have @xmath491 . the group @xmath479 can not act on a irreducible module @xmath490 of its non - trivial subgroup @xmath487 , this is especially obvious in our two cases of interest , @xmath492 and @xmath493 . a contradiction . , simple linear lie groups having a free algebra of invariants , _ selecta math _ , * 3*(1984 ) , 183220 ; originally published in _ voprosy teorii grupp i gomologicheskoi algebry _ , yaroslavl , 1979 , 341 ( in russian ) . , invariant theory , in : _ algebraic geometry iv _ , itogi nauki i tekhniki , pages 137314 . nauk sssr vsesoyuz . moscow , 1989 ; english translation : encyclopaedia math . 55 , springer , berlin , 1994 .	let @xmath0 be a complex reductive lie algebra and @xmath1 the underling vector space of a finite - dimensional representation of @xmath0 . then one can consider a new lie algebra @xmath2 , which is a semi - direct product of @xmath0 and an abelian ideal @xmath1 . we outline several results on the algebra @xmath3^{{\mathfrak}q}$ ] of symmetries invariants of @xmath4 and describe all semi - direct products related to the defining representation of @xmath5 with @xmath3^{{\mathfrak}q}$ ] being a free algebra .
the nature of the first generation of stars , and their influence on later epochs of structure formation , is a fundamental issue in modern cosmology . a great deal of theoretical progress has been made ( see reviews by & larson ( 2004 ) and & ferrara ( 2005 ) ) . however , a major question remains : what is the initial mass function of the first stars ? the answer to this question will help explain the contributions that population iii stars made towards reionization and chemical enrichment of the universe , and will also determine if these stars form compact objects ( which may be the progenitors of the supermassive black holes found in the centers of most massive galaxies ) , and the properties of those objects . the nature of the population iii imf has been debated for over half a century ( schwarzschild & spitzer 1953 ; ezer & cameron 1971 ; palla , salpeter , & stahler 1983 ; silk 1983 ; tegmark et al . it has been generally agreed upon that the lack of efficient cooling mechanisms results in very massive stars , with an imf that is extremely top - heavy compared to the galactic imf . this has been supported by recent cosmological simulations of the formation of population iii stars , which argue for massive objects ( abel et al . 2002 ; bromm et al . 2002 ; bromm & loeb 2004 ; yoshida et al . these simulations find that the jeans - unstable clumps which form in the cores of high - redshift halos are on the order of a thousand solar masses out of which a single primordial protostar condenses , and the subsonic contraction of the clump ( even at high densities ) greatly reduces the chance of fragmentation into many low - mass stars . though suggestive , these cosmological simulations lack the necessary physics and resolution to follow the collapse of the primordial cloud core to the point where accretion onto the star has been stopped and the star s final mass has been determined . the issue of accretion shutoff onto pop iii stars has been examined by & palla ( 2003 ) and & mckee ( 2004 ) . omukai & palla perform 1d , spherically symmetric simulations of the shutoff of accretion onto a primordial protostar . they find that there is a critical mass accretion rate , @xmath12 m@xmath7/year , below which stars can grow to essentially arbitrary masses ( certainly @xmath13 m@xmath7 ) , limited only by the mass of available gas and the stellar lifetime . above this critical accretion rate , the maximum mass of the star decreases as @xmath14 increases due to radiative feedback . they find that for a more realistic , time - varying accretion rate taken from ( 2002 ) ( hereafter referred to as abn02 ) , the star can grow to masses of @xmath15 m@xmath7 . tan & mckee present a theoretical model of population iii star formation that combines models of the collapsing cloud core , evolution of the primordial protostar and accretion disk . they show that radiative feedback processes are extremely important once the protostar reaches a mass of @xmath16 m@xmath7 . their models use as the fiducial case an accretion rate onto the disk taken from the simulation of abn02 . as with the omukai & palla calculation , accretion rates onto the forming protostar are critical input parameters , if not the most critical parameter , for determining the final mass of the population iii star . how well have the accretion rates onto population iii protostars been constrained ? recent work has made it clear that these rates must be determined primarily from cosmological simulations . ( 2002 ) do not have the mass and spatial resolution necessary to provide constraints . abn02 , in one of the most highly resolved simulations published to date , simulated the formation of a single population iii protostellar core that has a maximum accretion rate onto the forming protostar of @xmath9 m@xmath7/year . & loeb ( 2004 ) perform a high - resolution calculation of the formation of a primordial protostar and find an accretion rate rate that starts at @xmath17 m@xmath7/year , but declines rapidly . finally , ( 2006 ) present a calculation of the formation of a primordial protostellar cloud that takes into account opacity effects due to molecular hydrogen and also implements a novel particle - splitting scheme . this allows them to accurately follow the evolution of a primordial gas cloud which collapses to a density of @xmath18 @xmath19 . as in abn02 , yoshida et al . do not observe fragmentation of the collapsing cloud core , and observe gas accretion rates that peak at @xmath20 m@xmath7/yr and average around @xmath21 m@xmath7/yr . all of the published work thus far which studies population iii star formation in a cosmological context and at extremely high resolution has been an examination of a single object , and as such can not answer fundamental questions , such as : what is the variance in accretion rates onto primordial protostars ? is there some systematic dependence upon redshift or environment ? and , finally , is the lack of fragmentation in the collapsing halo core universal ? some previous work has studied large numbers of primordial halos ( machacek , bryan , & abel 2001 , 2003 ; yoshida et al . however , these studies lacked the resolution to study properties of the halo cores in detail . in this work we will attempt to address these questions by performing a series of extremely high dynamical range adaptive mesh refinement ( amr ) simulations of the formation of a population iii protostar in a @xmath0cdm cosmology . our mass and spatial resolution is comparable to abn02 and ( 2006 ) . this is the first time that a significant number of such high - resolution calculations have been sampled , allowing us to investigate the robustness of previous results and address questions that they were unable to examine . our sample includes halos which cool and exhibit runaway collapse at their centers over the redshift range @xmath22 . the halo virial masses at collapse range from @xmath23 m@xmath7 and show no systematic trend with redshift . we explore the environmental effects that lead to these variations . we discover that there is a trend towards increasing accretion rates in halos with decreasing collapse redshifts ( later collapse times ) , which is caused by evolving halo virial properties and the physics of molecular hydrogen formation . peak accretion rates vary between @xmath24 and @xmath25 m@xmath7/year and have similar time dependence . finally , we observe that our collapsing protostellar cloud cores do not fragment , consistent with previous results , which suggests that population iii stars which form in halos of mass @xmath11 m@xmath7 always form in isolation . the simulations discussed in this paper study population iii protostars which form in halos of mass @xmath26 m@xmath7 and in the redshift range @xmath27 . these halos are far more common than the extremely rare peaks discussed by ( 2005 ) and ( 2005 ) , and thus represent a more `` typical '' population of primordial stars . additionally , in this paper we do not concern ourselves with halos with virial temperatures above @xmath28 k , where cooling by atomic hydrogen could be important ( oh & haiman 2002 ; bromm & loeb 2003 ) . investigation of this mode of population iii star formation , though potentially very important , is deferred to a later paper . this is the first in a series of papers presenting results of high dynamical range amr cosmological simulations of the formation of population iii stars in a @xmath0cdm cosmological context . as stated previously , this paper discusses issues relating to accretion rates onto evolving primordial protostellar cores . the second paper in this series ( oshea , norman & li 2006 , in preparation ) presents an investigation of the evolution of angular momentum in a collapsing primordial halo core . the third paper in this series ( oshea & norman 2006 , in preparation ) presents an examination of the formation of population iii protostellar halos in the presence of a molecular hydrogen photodissociating uv background . later papers will present more detailed investigations of the evolution of the collapsing protostellar cores and the formation of a primordial protostar embedded in a thick accretion disk . the organization of this paper is as follows : in section [ methodology ] , we provide a summary of enzo , the code used to perform the calculations in this paper and the second and third papers in this series , and of the calculation setup . sections [ repstar ] and [ compare - real ] summarize our results : in section [ repstar ] , we present the detailed evolution of a single , representative population iii protostellar cloud , and compare it to previous work . in section [ compare - real ] , we present results of a series of calculations of population iii protostellar cloud formation with varied large scale structure and simulation volumes , and show some systematic trends in the properties of these clouds . in section [ discuss ] we discuss the results presented in this work , and in section [ summary ] we present a summary of the main results and conclusions . ` enzo ' is a publicly available , extensively tested adaptive mesh refinement cosmology code developed by greg bryan and others ( bryan & norman 1997a , 1997b ; norman & bryan 1999 ; oshea et al . 2004 , 2005 ) . the specifics of the enzo code are described in detail in these papers ( and references therein ) , but we present a brief description here for clarity . the enzo code couples an n - body particle - mesh ( pm ) solver ( efstathiou et al . 1985 ; hockney & eastwood 1988 ) used to follow the evolution of a collisionless dark matter component with an eulerian amr method for ideal gas dynamics by & colella ( 1989 ) , which allows high dynamic range in gravitational physics and hydrodynamics in an expanding universe . this amr method ( referred to as _ structured _ amr ) utilizes an adaptive hierarchy of grid patches at varying levels of resolution . each rectangular grid patch ( referred to as a `` grid '' ) covers some region of space in its _ parent grid _ which requires higher resolution , and can itself become the parent grid to an even more highly resolved _ child grid_. enzo s implementation of structure amr places no fundamental restrictions on the number of grids at a given level of refinement , or on the number of levels of refinement . however , owing to limited computational resources it is practical to institute a maximum level of refinement , @xmath29 . additionally , the enzo amr implementation allows arbitrary integer ratios of parent and child grid resolution , though in general for cosmological simulations ( including the work described in this paper ) a refinement ratio of 2 is used . since the addition of more highly refined grids is adaptive , the conditions for refinement must be specified . in enzo , the criteria for refinement can be set by the user to be a combination of any or all of the following : baryon or dark matter overdensity threshold , minimum resolution of the local jeans length , local density gradients , local pressure gradients , local energy gradients , shocks , and cooling time . a cell reaching any or all of the user - specified criteria will then be flagged for refinement . once all cells of a given level have been flagged , rectangular solid boundaries are determined which minimally encompass them . a refined grid patch is then introduced within each such bounding volume , and the results are interpolated to a higher level of resolution . in enzo , resolution of the equations being solved is adaptive in time as well as in space . the timestep in enzo is satisfied on a level - by - level basis by finding the largest timestep such that the courant condition ( and an analogous condition for the dark matter particles ) is satisfied by every cell on that level . all cells on a given level are advanced using the same timestep . once a level @xmath30 has been advanced in time @xmath31 , all grids at level @xmath32 are advanced , using the same criteria for timestep calculations described above , until they reach the same physical time as the grids at level @xmath30 . at this point grids at level @xmath32 exchange baryon flux information with their parent grids , providing a more accurate solution on level @xmath30 . cells at level @xmath32 are then examined to see if they should be refined or de - refined , and the entire grid hierarchy is rebuilt at that level ( including all more highly refined levels ) . the timestepping and hierarchy rebuilding process is repeated recursively on every level to the maximum existing grid level in the simulation . two different hydrodynamic methods are implemented in enzo : the piecewise parabolic method ( ppm ) ( woodward & colella 1984 ) , which was extended to cosmology by ( 1995 ) , and the hydrodynamic method used in the zeus magnetohydrodynamics code ( stone & norman 1992a , 1992b ) . we direct the interested reader to the papers describing both of these methods for more information , and note that ppm is the preferred choice of hydro method since it is higher - order - accurate and is based on a technique that does not require artificial viscosity , which smoothes shocks and can smear out features in the hydrodynamic flow . the chemical and cooling properties of primordial ( metal - free ) gas are followed using the method of ( 1997 ) and ( 1997 ) , supplemented with reactions for the 3-body formation of @xmath33 documented in abn02 . this method follows the non - equilibrium evolution of a gas of primordial composition with 9 total species : @xmath34 , @xmath35 , @xmath36 , @xmath37 , @xmath38 , @xmath39 , @xmath40 , @xmath33 , and @xmath41 . the code also calculates radiative heating and cooling following atomic line excitation , recombination , collisional excitation , free - free transitions , molecular line excitations , and compton scattering of the cosmic microwave background , as well as any of approximately a dozen different models for a metagalactic uv background that heat the gas via photoionization and photodissociation . we model the cooling processes detailed in ( 1997 ) , but with the molecular hydrogen cooling function of & palla ( 1998 ) . the multispecies rate equations are solved out of equilibrium to properly model situations where , e.g. , the cooling time of the gas is much shorter than the hydrogen recombination time . a total of 9 kinetic equations are solved , including 29 kinetic and radiative processes , for the 9 species mentioned above . the chemical reaction equation network is technically challenging to solve due to the huge range of reaction time scales involved the characteristic creation and destruction time scales of the various species and reactions can differ by many orders of magnitude . as a result , the set of rate equations is extremely stiff , and an explicit scheme for integration of the rate equations can be costly if small enough timestep are taken to keep the network stable . this makes an implicit scheme preferable for such a set of equations , and enzo solves the rate equations using a method based on a backwards differencing formula ( bdf ) in order to provide a stable and accurate solution . it is important to note the regime in which this chemistry model is valid . according to ( 1997 ) and ( 1997 ) , the reaction network is valid for temperatures between @xmath42 k. the original model discussed in these two references is only valid up to n@xmath43 @xmath19 . however , addition of the 3-body h@xmath44 formation process allows correct modeling of the chemistry of the gas up until the point where collisionally - induced emission from molecular hydrogen becomes an important cooling process , which occurs at n@xmath45 @xmath19 . we do not include heating by molecular hydrogen formation , which should take place at densities above @xmath46 @xmath47 , and may affect temperature evolution at these high densities a further concern is that the optically thin approximation for radiative cooling breaks down , which begins at n@xmath48 @xmath19 . beyond this point , modifications to the cooling function that take into account the non - negligible opacity of the gas to line radiation from molecular hydrogen must be made , as discussed by & abel ( 2004 ) . even with these modifications , a more correct description of the cooling of gas of primordial composition at high densities will require some form of radiation transport , which will greatly increase the cost of the simulations . the simulations discussed in this paper are set up as follows . we initialize all calculations at @xmath49 assuming a `` concordance '' cosmological model : @xmath50 , @xmath51 , @xmath52 , @xmath53 , @xmath54 ( in units of 100 km / s / mpc ) , @xmath55 , and using an & hu ( 1999 ) power spectrum with a spectral index of @xmath56 . twelve simulations are generated using a separate random seed for each , meaning that the large - scale structure that forms in each of the simulation volumes is statistically independent of the others . these simulations , detailed in table [ table.siminfo ] , are divided into sets of four simulations in three different box sizes : @xmath57 , and @xmath58 h@xmath59 mpc ( comoving ) . the most massive halo to form in each simulation at @xmath60 ( typically with a mass of @xmath61 m@xmath7 ) is found using a dark matter - only calculation with @xmath62 particles on a @xmath62 root grid with a maximum of 4 levels of adaptive mesh , refining on a dark matter overdensity criterion of @xmath63 . the initial conditions are then regenerated with both dark matter and baryons for each of the simulation volumes such that the lagrangian volume in which the halo formed is now resolved at much higher spatial and mass resolution . these simulations have a @xmath62 root grid and three levels of static nested grids , each with twice the spatial resolution and eight times the mass resolution of the previous grid . this gives an overall effective grid size of @xmath64 in the region where the most massive halo will form . the highest resolution grid in each simulation is @xmath65 grid cells , and corresponds to a volume @xmath66 ) h@xmath59 comoving kpc on a side for the ( 0.3 , 0.45 , 0.6 ) h@xmath59 mpc box . the dark matter particles in the highest resolution grid are ( 1.81 , 6.13 , 14.5 ) h@xmath59 m@xmath7 and the spatial resolution of cells on these grids are ( 293 , 439 , 586 ) h@xmath59 parsecs ( comoving ) . though the simulations have a range of initial spatial resolutions and dark matter masses , we find that the final simulation results are converged the spatial and mass resolution of the 0.3 h@xmath59 mpc volume simulations can be degraded to that of the 0.6 h@xmath59 mpc without significantly changing the results . the simulations with nested grid initial conditions are then started at @xmath49 and allowed to evolve until the collapse of the gas within the center of the most massive halo , which occurs at a range of redshifts ( as shown in section [ compare - real ] , and summarized in tables [ table.meaninfo-1 ] and [ table.meaninfo-2 ] ) . the equations of hydrodynamics are solved using the ppm method with a dual energy formulation ( the results are essentially identical when the zeus hydrodynamic method is used ) . the nonequilibrium chemical evolution and optically thin radiative cooling of the primordial gas is modeled as described in section [ enzocode ] , following 9 separate species including molecular hydrogen ( but excluding deuterium ) . adaptive mesh refinement is enabled within the high resolution region such that cells are refined by factors of two along each axis , to a maximum of 22 total levels of refinement . this corresponds to a maximum resolution of ( 115 , 173 , 230 ) h@xmath59 astronomical units ( comoving ) at the finest level of resolution , with an overall spatial dynamical range of @xmath67 . to avoid effects due to the finite size of the dark matter particles , the dark matter density is smoothed on a comoving scale of @xmath68 pc ( a proper scale of @xmath69 pc at @xmath70 ) . this is reasonable because at that radius in all of our calculations the gravitational potential is completely dominated by the baryons . grid cells are adaptively refined based upon several criteria : baryon and dark matter overdensities in cells of 4.0 and 8.0 , respectively , checks to ensure that the pressure jump and/or energy ratios between adjoining cells never exceeds 5.0 , that the cooling time in a given cell is always longer than the sound crossing time of that cell , and that the jeans length is always resolved by at least 16 cells . this guarantees that the truelove criterion ( truelove et al . 1997 ) , which is an empirical result showing that in order to avoid artificial gravitational fragmentation in numerical simulations the jeans length must be resolved by at least 4 grid cells , is always maintained by a significant margin . simulations which force the jeans length to be resolved by a minimum of 4 and 64 cells produce results which are essentially identical to those presented here . cccccccccc l0_30a & 0.3 & 115.3 & 2.60 & 0.40 & @xmath71 & 0.1964 & 1.873 & 1.153 & 2.557 + l0_30b & 0.3 & 115.3 & 2.60 & 0.40 & @xmath71 & 0.1952 & 2.176 & 1.322 & 2.858 + l0_30c & 0.3 & 115.3 & 2.60 & 0.40 & @xmath71 & 0.1995 & 1.981 & 1.319 & 2.778 + l0_30d & 0.3 & 115.3 & 2.60 & 0.40 & @xmath71 & 0.1922 & 2.727 & 1.702 & 3.022 + + l0_45a & 0.45 & 173.0 & 8.76 & 1.35 & @xmath72 & 0.1879 & 2.669 & 1.873 & 3.615 + l0_45b & 0.45 & 173.0 & 8.76 & 1.35 & @xmath72 & 0.1860 & 2.726 & 1.872 & 3.343 + l0_45c & 0.45 & 173.0 & 8.76 & 1.35 & @xmath72 & 0.1906 & 2.375 & 1.563 & 3.293 + l0_45d & 0.45 & 173.0 & 8.76 & 1.35 & @xmath72 & 0.1861 & 1.959 & 1.191 & 3.827 + + l0_60a & 0.6 & 230.6 & 20.8 & 3.20 & @xmath73 & 0.1815 & 2.458 & 1.790 & 3.154 + l0_60b & 0.6 & 230.6 & 20.8 & 3.20 & @xmath73 & 0.1802 & 3.049 & 1.956 & 3.340 + l0_60c & 0.6 & 230.6 & 20.8 & 3.20 & @xmath73 & 0.1815 & 2.840 & 2.121 & 3.466 + l0_60d & 0.6 & 230.6 & 20.8 & 3.20 & @xmath73 & 0.1959 & 2.700 & 1.814 & 4.147 + [ table.siminfo ] cccccccccc l0_30a & 19.28 & @xmath74 & 118.1 & 1120.0 & @xmath75 & 0.823 & 0.050 & 0.022 & 55.1 + l0_30b & 22.19 & @xmath76 & 91.7 & 1009.6 & @xmath77 & 0.753 & 0.066 & 0.029 & 3.3 + l0_30c & 20.31 & @xmath78 & 132.9 & 1646.8 & @xmath79 & 0.819 & 0.053 & 0.034 & 17.3 + l0_30d & 24.74 & @xmath80 & 64.0 & 671.4 & @xmath81 & 0.711 & 0.046 & 0.014 & 12.7 + + l0_45a & 28.70 & @xmath82 & 67.4 & 1131.7 & @xmath83 & 0.631 & 0.108 & 0.056 & 8.4 + l0_45b & 26.46 & @xmath84 & 72.7 & 1052.8 & @xmath85 & 0.706 & 0.019 & 0.019 & 34.0 + l0_45c & 24.74 & @xmath86 & 100.1 & 1645.6 & @xmath87 & 0.750 & 0.054 & 0.015 & 97.4 + l0_45d & 28.67 & @xmath88 & 90.5 & 2060.0 & @xmath89 & 0.727 & 0.059 & 0.071 & 11.4 + + l0_60a & 24.10 & @xmath90 & 93.8 & 1338.0 & @xmath91 & 0.813 & 0.052 & 0.027 & 26.0 + l0_60b & 25.64 & @xmath92 & 87.3 & 1385.6 & @xmath93 & 0.730 & 0.047 & 0.056 & 11.0 + l0_60c & 28.13 & @xmath94 & 60.7 & 876.9 & @xmath81 & 0.575 & 0.027 & 0.016 & 40.0 + l0_60d & 32.70 & @xmath95 & 62.5 & 1440.9 & @xmath96 & 0.542 & 0.049 & 0.033 & 10.0 + [ table.meaninfo-1 ] cccccccc l0_30a & @xmath97 & @xmath98 & 745.8 & 438.3 & @xmath99 & @xmath100 & 1159.4 + l0_30b & @xmath101 & @xmath102 & 645.2 & 468.6 & @xmath103 & @xmath104 & 1768.6 + l0_30c & @xmath105 & @xmath106 & 808.1 & 501.7 & @xmath107 & @xmath108 & 1797.7 + l0_30d & @xmath109 & @xmath110 & 512.9 & 310.7 & @xmath111 & @xmath112 & 814.8 + + l0_45a & @xmath113 & @xmath114 & 858.4 & 276.3 & @xmath115 & @xmath116 & 1120.2 + l0_45b & @xmath117 & @xmath118 & 698.0 & 295.3 & @xmath119 & @xmath119 & 1077.3 + l0_45c & @xmath120 & @xmath121 & 897.8 & 502.7 & @xmath104 & @xmath122 & 1141.0 + l0_45d & @xmath123 & @xmath124 & 878.0 & 354.2 & @xmath125 & @xmath126 & 856.6 + + l0_60a & @xmath127 & @xmath128 & 808.8 & 431.1 & @xmath129 & @xmath130 & 1750.8 + l0_60b & @xmath131 & @xmath114 & 803.3 & 300.1 & @xmath132 & @xmath111 & 1301.2 + l0_60c & @xmath133 & @xmath134 & 753.7 & 420.1 & @xmath135 & @xmath126 & 1395.9 + l0_60d & @xmath136 & @xmath137 & 917.2 & 249.5 & @xmath138 & @xmath139 & 928.5 + [ table.meaninfo-2 ] in this section we describe in detail the collapse of a single primordial protostellar core out of the ensemble discussed in section [ compare - real ] . this simulation , l0_30a , was selected out of the four simulations performed in a @xmath140 h@xmath59 mpc comoving volume . the results described here are qualitatively similar for all of the calculations described in section [ compare - real ] , though there is some scatter in the exact evolution of each halo due to differences in large scale structure and the detailed merger history of the halo . however , since the collapse is essentially controlled by the chemistry of molecular hydrogen formation , the result is general . figures [ fig.rep.proj-dens ] , [ fig.rep.proj-temp ] , and [ fig.rep.proj-level ] zoom in on the central gas core in each halo at the redshift of collapse ( defined as the redshift where the baryon density at the halo center reaches @xmath141 @xmath19 ) by factors of four , showing projections of log baryon density , log baryon temperature , and maximum refinement level , respectively . the largest - scale panel shows a projection of a volume of the universe 1320 proper parsecs across and deep , and zooms in by factors of four to approximately 1.3 pc across . each panel is centered on the collapsing protostar . at large scales it is apparent from figure [ fig.rep.proj-dens ] that the halo in which the first star in the simulation volume forms is at the intersection of two cosmological filaments , a distinctly asymmetrical situation . examination of figure [ fig.rep.proj-temp ] shows that the filaments and majority of the volume of the halo are relatively hot ( @xmath142 kelvin ) , due primarily to accretion shocks formed by gas raining onto the filaments and into the halo . however , as we zoom in towards the center of the halo we can see that the high - density gas is at a much lower temperature ( a few hundred kelvin ) due to cooling by the significant quantity of molecular hydrogen that is formed in the halo . the gas within the halo is not particularly spherical until scales of a few parsecs are reached , where a somewhat warmer core of gas ( t @xmath142 k ) forms with an overall mass of a few thousand solar masses , harboring a fully - molecular protostellar cloud core with a mass of @xmath143 m@xmath7 . the central core is generally spheroidal due to gas pressure and is not rotationally supported at any point up to the time the simulation is stopped . note that at later times , when the core has collapsed further , it is likely that a rotationally - supported accretion disk will form see & mckee ( 2004 ) for more details . figure [ fig.rep.proj-level ] shows how the adaptive mesh refinement is used to resolve the cosmological structure by concentrating refinement only where it is needed . this method is extremely effective at conserving computational resources - the level 16 grids , which are the highest level of resolution shown in figure [ fig.rep.proj-level ] , only encompass @xmath144 of the overall volume ! ) , the maximum density was @xmath145 @xmath19 with a cosmic mean density of @xmath146 @xmath19 , for an overall density increase of 15 orders of magnitude . top left : view is 1320 pc across . top center : 330 pc . top right : 82.5 pc . bottom left : 20.6 pc . bottom center : 5.2 pc . bottom right : 1.29 pc . note that all sizes are in proper parsecs at @xmath147 . in all panels yellow represents high densities and blue represents low density , with the color table relative in each frame . , scaledwidth=90.0% ] and the simulation and spatial sizes of each panel are the same as in figure [ fig.rep.proj-dens ] . in all panels white represents high temperatures and dark colors represent low temperatures . the color table is relative in each frame . , scaledwidth=90.0% ] . the maximum projected level in each panel is as follows . top left : level 6 . top center : level 8 . top right : level 10 . bottom left : level 12 . bottom center : level 14 . bottom right : level 16 . in each panel the highest level grid is represented is in red , second highest in yellow , and so on . the highest level of resolution at this time is @xmath148 . , scaledwidth=90.0% ] figures [ fig.rep.panel1 ] through [ fig.rep.panel3 ] show the evolution of radial profiles of several spherically averaged , mass - weighted baryon quantities of a representative primordial protostar from approximately the onset of halo collapse until the formation of a fully molecular core . the halo begins its collapse at @xmath149 ( approximately @xmath150 years after the big bang ) and ends its collapse @xmath151 years later , at @xmath152 . figure [ fig.rep.panel1 ] shows the spherically - averaged baryon number density , temperature , and enclosed mass as a function of radius , and the specific angular momentum of the baryon gas as a function of enclosed mass . figure [ fig.rep.panel2 ] shows the molecular hydrogen fraction , electron fraction , @xmath39 fraction and circular velocity as a function of radius . figure [ fig.rep.panel3 ] shows the ratios of gas cooling time to sound crossing time , cooling time to system dynamical time , sound crossing time to dynamical time , and radial velocity as a function of radius . the lines in all of these plots are chosen such that the same line thickness and type corresponds to the same time in each panel . we begin to follow the evolution of the halo at @xmath153 , when the central hydrogen number density has reached @xmath154 particles per cubic centimeter ( black solid line in all plots ) . this corresponds to a core with a radius of @xmath143 parsec and a mass of a few thousand solar masses , which is accreting gas at a significant rate . the molecular hydrogen fraction within this core is slightly less than @xmath155 but is still enough to rapidly cool the center of the halo to @xmath156 kelvin at a cooling rate proportional to the square of the gas density . the gas can not cool below this temperature because of the sharp decrease in the cooling rate of molecular hydrogen below @xmath157 kelvin . this core is the high - redshift equivalent of a molecular cloud core . the halo `` loiters '' for approximately six million years as the amount of molecular hydrogen is slowly built up to a mass fraction of a few times @xmath155 and the central density increases . as can be seen from figure [ fig.rep.panel3 ] , the cooling time in the halo core is significantly larger than both the dynamical time and sound crossing time at this epoch , so the core can be quite reasonably said to be quasistatically evolving . it is useful to compare the late - time evolution of a population iii minihalo to the ( 1977 ) isothermal sphere model of cloud collapse , which is most commonly applied to quasistatic cores in galactic molecular clouds . as the gas density passes roughly @xmath158 @xmath19 the ro - vibrational levels of h@xmath44 are populated at their thermodynamic equilibrium values and the cooling rate becomes independent of density , which corresponds to an increase in gas temperature with increasing density ( as can be seen by the blue and green solid lines in the temperature vs. radius plot in figure [ fig.rep.panel1 ] ) . as the temperature increases the cooling rate again begins to rise , leading to an increase in the inflow velocities of gas . this is easily explained by the shu model , as the infall of gas onto the quasistatic halo is moderated by the sound speed , and the accretion rate @xmath14 scales as the sound speed cubed . examination of the plot of enclosed mass vs. radius in figure [ fig.rep.panel1 ] shows that at this point the enclosed gas mass has exceeded the bonnor - ebert critical mass , which is defined as @xmath159 , where @xmath160 is the local sound speed and @xmath161 is newton s constant . this is the critical mass at which an isothermal sphere of gas with an external pressure @xmath162 becomes gravitationally unstable and undergoes collapse . this occurs in this halo at a mass scale of @xmath142 m@xmath7 . when the central density of the cloud core becomes sufficiently large ( @xmath163 @xmath19 ) the three - body h@xmath44 formation process takes over , resulting in a rapid increase in the molecular hydrogen fraction from a few times @xmath155 to essentially unity . the halo densities also cause a rapid drop in the @xmath39 and electron fractions , though the dominance of the 3-body channel makes the @xmath39 channel for molecular hydrogen formation unimportant . this causes a huge increase in the cooling rate , which results in a rapid drop in temperature of the center of the halo , allowing it to contract and causing an increase in central density of @xmath164 @xmath19 in only @xmath165 years . the infall velocity of gas to the center of the halo peaks at the same radius as the temperature , and drops within that radius ( r @xmath166 pc ) . at a mass scale of @xmath143 m@xmath7 a protostellar cloud core forms which is completely molecular and has gas accreting onto it supersonically . at this point the optical depth of the halo core becomes close to unity to molecular hydrogen ro - vibrational line emission , so we terminate the simulation because the assumption of optically thin radiative cooling used in our code is no longer correct . note also the change from supersonic back to subsonic accretion at very small radii in figure [ fig.rep.panel3 ] ( d ) is not an accretion shock onto a protostar rather , it is a numerical effect due to the flow converging to a single point in a grid cell with finite resolution . panel ( d ) in figures [ fig.rep.panel1 ] through [ fig.rep.panel3 ] shows the dynamical evolution of the collapsing halo . these plots show the spherically - averaged , mass - weighted , specific angular momentum as a function of enclosed baryon mass and the spherically - averaged , mass - weighted , baryon radial velocity and circular velocity as a function of radius . though the angular momentum of the gas within the halo core ( m@xmath167 m@xmath7 ) is conserved overall , there is clear evidence for transport of angular momentum outward . this effect can be shown to be physical ( as opposed to a numerical effect ) , and is due to a combination of turbulent transport of high angular momentum gas outward and a `` segregation effect , '' where low angular momentum gas tends to move inward during halo collapse . a detailed discussion of this issue is deferred to the second paper in this series ( oshea , norman & li 2006 , in preparation ) . the thin dot - dashed lines in the plots of spherically - averaged baryon infall velocity and cylindrically - averaged ( in shells along the axis of rotation ) baryon circular velocity ( the bottom right panel in figure [ fig.rep.panel2 ] correspond to the locally - determined sound speed and keplerian orbital velocities at the final timestep , respectively . the infall of gas is locally subsonic until very late times , when the core has reached densities of @xmath168 @xmath19 and an h@xmath44 fraction approaching unity . the infall velocities then become locally supersonic , but decrease rapidly within this radius because a protostellar cloud core has formed which the gas is accreting onto , and has formed a standing shock . at no point in the simulated evolution is the gas within the halo core rotationally - supported : the initial distribution of gas velocities is such that gas in the core has very little angular momentum , and what little exists is efficiently transported outward . it is useful to examine the relevant chemical , thermal and dynamical timescales of the collapsing halo . the ratios of cooling time to sound crossing time ( calculated as a mass - weighted mean in spherically averaged radial shells ) as a function of radius , cooling time to dynamical time ( calculated assuming spherical symmetry ) , and sound crossing time to dynamical time are plotted in figure [ fig.rep.panel3 ] . within the core of the halo ( @xmath169 parsec ) the sound crossing time ( @xmath170 ) is slightly less than the dynamical time ( @xmath171 ) for the majority of the evolution time of the halo , while the cooling time ( @xmath172 ) is somewhat longer than both of these timescales , but generally only by a factor of a few . if @xmath173 the halo is stable against collapse because the halo can easily equilibrate its pressure to compensate for collapsing gas . if @xmath174 , the system can not come into equilibrium and is in free - fall . in this case , @xmath175 , and the system is undergoing a quasistatic collapse . this can also be seen by examining the evolution of the radial infall velocity as a function of radius in figure [ fig.rep.panel3 ] , where the radial infall velocity is subsonic until the very last portion of the core s evolution , when it becomes locally supersonic . this is triggered by a dramatic increase in the molecular hydrogen fraction , and a corresponding rapid decrease in the cooling time . in the center of the halo at the last few data outputs , the cooling time becomes shorter than both the dynamical time and sound crossing time , creating a situation where gas is essentially free - falling onto the central protostar . as in abn02 , we carefully examine the collapsing protostellar cloud core for signs of fragmentation . this might be expected due to chemothermal instabilities caused by the rapid formation of molecular hydrogen via the 3-body process and the resulting rapid increase in cooling rate ( silk 1983 ) . however , the sound crossing time within the core is less than the h@xmath44 formation timescale until the last output time , allowing mixing to take place and preventing the formation of large density contrasts . by the time that the h@xmath44 formation timescale becomes shorter than the sound crossing time , the core is fully molecular and therefore stable against this chemothermal instability . for an alternate explanation , see & yoshii ( 2003 ) . ( approximately @xmath176 years after the big bang ) . thick dotted line : the state of the halo @xmath177 years after that . thick short - dashed line : @xmath178 years later . thick long - dashed line : @xmath179 years later . thick dot - dashed line : @xmath180 years later . thin solid line : @xmath181 years later . thin dotted line : @xmath182 years later . thin short - dashed line : @xmath183 years later . thin long - dashed line : @xmath184 years later , at a final output redshift of @xmath185 . the total time spanned by the lines in these panels is @xmath151 years . the black dot - dashed line in the bottom left panel is the bonnor - ebert critical mass calculated at the last timestep.,scaledwidth=90.0% ] fraction ( c ) and baryon circular velocity ( d ) as a function of radius of a representative primordial protostar . the lines correspond to the same times as in figure [ fig.rep.panel1 ] and are of the same simulation . the black dot - dashed line in the plot of baryon circular velocity vs. time is the newtonian circular velocity computed from the radius and enclosed baryon mass at the last timestep . , scaledwidth=90.0% ] and are of the same simulation . the black dot - dashed line in the plot of radial velocity as a function of radius is the sound speed calculated using the local baryon temperature in each radial bin at the last simulation timestep . , scaledwidth=90.0% ] as discussed previously , at the time when the simulation is stopped due to a breakdown in the assumption of optically thin radiative cooling at the center of the protostellar cloud , a fully - molecular core with a mass of @xmath143 m@xmath7 has formed and is accreting gas supersonically . the spherically - averaged accretion time at the last output timestep , plotted as a function of enclosed gas mass , is shown as the thick solid line in figure [ fig.rep.accrete ] . the accretion time is defined as @xmath186 , where @xmath187 is the enclosed baryon mass and @xmath188 , with @xmath189 and @xmath190 being the spherically - averaged , mass - weighted baryon density and velocity as a function of radius , and @xmath190 defined as being positive towards the center of the halo . the thick dashed line is the accretion time as determined by taking the local accretion rate from the shu isothermal collapse model , @xmath191 , where @xmath192 is a dimensionless constant of order unity ( we use @xmath193 , as derived by shu ) , @xmath160 is the sound speed , and @xmath161 is the gravitational constant . this value of @xmath14 is calculated in each bin and the accretion time is plotted as @xmath194 . the dot - long dashed line is the kelvin - helmholtz time for a population iii star with a mass identical to the enclosed mass calculated assuming a star with a main sequence radius and luminosity as given by ( 2002 ) . the short dash - long dashed line is the lifetime of a massive primordial star as calculated by schaerer , assuming no mass loss . note that this fit only covers the mass range @xmath195 m@xmath7 , but the lifetime of massive stars with a mass greater than @xmath196 m@xmath7 is roughly constant . the dot - short dashed line is the baryon accretion time for the result in abn02 . we can see that the accretion rates in our model and in abn02 are well - approximated by the shu model for @xmath197 m@xmath7 , which is the bonner - ebert mass scale of the collapsing core . all three curves predict that @xmath198 m@xmath7 of gas will accrete within the kelvin - helmholtz time of the population iii star , implying that it will be at least this massive . as we will see in the next section , the similarity of these curves is somewhat fortuitous , as we find substantial variation of accretion times within our suite of models . , where @xmath187 is the enclosed baryon mass and @xmath188 , with @xmath189 and @xmath190 being the baryon density and velocity as a function of radius , and @xmath190 defined as being positive towards the center of the halo . the thick solid line is the baryon accretion time for this simulation . the thick dashed line is the accretion time as determined by taking the accretion rate from the shu isothermal collapse model , @xmath191 , where @xmath192 is a dimensionless constant of order unity , @xmath160 is the sound speed , and @xmath161 is the gravitational constant . this value of @xmath14 is calculated in each bin and the accretion time is plotted as @xmath194 . the thin line with the label `` kh '' is the kelvin - helmholtz time for a population iii star with a mass identical to the enclosed mass , as calculated from the results given by schaerer . the thin line labeled `` @xmath199 '' is the lifetime of a massive primordial star assuming no significant mass loss , as calculated by schaerer . note that this fit only covers the mass range @xmath195 m@xmath7 , but @xmath200 constant for @xmath201 m@xmath7 . the thin dot - short dashed line labeled `` abn02 '' is the baryon accretion time for the result in abn02 . the plot here corresponds to the last output dataset , corresponding to the thin , long - dashed line in figures [ fig.rep.panel1 ] through [ fig.rep.panel3].,scaledwidth=70.0% ] another important issue relating to studies of the formation of population iii stars in a cosmological context is the consistency of the results over a range of simulation parameters . as discussed in the introduction to this article , previously published high dynamical range calculations of pop iii star formation have concentrated upon a single cosmological realization . while this is an important first step , it neglects possible systematic effects and also allows for error due to small number statistics . in this section we attempt to address some of these issues . twelve simulations are set up as described in section [ simsetup ] . each simulation has the same cosmological parameters but a different cosmological realization ( i.e. large scale structure ) . four simulations in each of three box sizes ( @xmath57 , and @xmath58 h@xmath59 comoving mpc ) are performed , with the results shown in figures [ fig.comp.panel1 ] through [ fig.accrete.panel3 ] . figures [ fig.comp.panel1 ] - [ fig.comp.panel3 ] display several mean properties of the halos . in each panel information for each simulation is plotted as an open shape which assigned according to box size as described in the figure captions . the filled - in shapes correspond to mean values for all simulations of a given box size ( with shapes again corresponding to the box size ) , and the open circle corresponds to the mean of all twelve of the simulations together . all quantities discussed in this section can be found in tables [ table.meaninfo-1 ] and [ table.meaninfo-2 ] . panel ( a ) of figure [ fig.comp.panel1 ] shows the virial mass of each halo calculated using the virial density fitting function of & norman ( 1998 ) at the time of protostellar cloud collapse plotted against the redshift of collapse . though there is a large amount of scatter in virial mass overall , with the smallest halo having a mass of @xmath80 m@xmath7 and the largest @xmath78 m@xmath7 , the average virial mass in each box size is virtually identical . the mean virial mass of all twelve of the halos is @xmath202 m@xmath7 , which is significantly lower than the halo mass of @xmath203 m@xmath7 in abn02 and the minimum mass of @xmath204 h@xmath59 m@xmath7 found by ( 2003 ) ( note that the value found by abn02 is from a different cosmology with much higher values of @xmath205 and @xmath206 , yet is still within the scatter of halo virial masses found in this work ) . in contrast to the virial mass , there is an apparent trend in earlier collapse time ( large collapse redshift ) as a function of box size , with the 0.45 and 0.6 h@xmath59 mpc boxes collapsing at a mean redshift of @xmath207 and the 0.3 h@xmath59 mpc boxes collapsing at a mean redshift of @xmath208 . this can be understood as a result of the larger boxes sampling a much larger volume of the universe . the first halo to collapse in the larger volume tends to be a statistically rarer density fluctuation that collapses at a higher redshift . a consequence of this can be seen in panel ( b ) of figure [ fig.comp.panel1 ] , which shows the mean baryon temperature in each halo as a function of collapse redshift . as with the plot of virial mass vs. collapse redshift , there is a significant amount of scatter in the results , but a clear trend of increased overall halo temperature with collapse redshift is apparent . this is explainable in terms of virial properties of the halo . the virial temperature scales as @xmath209 , so halos with a constant virial mass will be hotter at higher redshifts , as observed in this plot . vs. halo virial mass . panel ( d ) : halo mean baryon temperature vs. halo virial temperature . in each plot , open triangles , squares and hexagons correspond to simulations with 0.3 h@xmath59 mpc , 0.45 h@xmath59 mpc and 0.6 h@xmath59 mpc comoving box size , respectively . the filled shapes corresponds to the average value for simulations with that box size . the open circle corresponds to the average for all simulations together . the upper and lower dashed lines in panels ( c ) and ( d ) correspond to the scaling relation one would expect if @xmath210 = @xmath211 and @xmath210 = 0.5 @xmath211 , respectively . , scaledwidth=90.0% ] the virial scaling of the mean halo baryon temperature vs. halo virial mass ( @xmath212 vs. @xmath213 ) is plotted in panel ( c ) of figure [ fig.comp.panel1 ] , and the mean halo temperature versus the halo virial temperature is plotted in panel ( d ) . the upper and lower dashed lines in both panels correspond to the scaling relation one would expect if @xmath210 = @xmath211 and @xmath210 = 0.5 @xmath211 , respectively . since the actual observed mean halo temperatures are significantly lower than the virial temperature , as shown in panel ( d ) , one can infer that radiative cooling plays a significant role in the condensation of gas in these halos despite the generally poor cooling properties of primordial gas at low temperature . if radiative cooling were completely unimportant the mean halo baryon temperature would be approximately the virial temperature for all halos . examination of panels ( c ) and ( d ) together show that the temperature evolution of the halos is consistent with virial scaling , given the importance of radiative cooling . figure [ fig.comp.panel2 ] shows the relationship of the angular momentum in the halos with various quantities . the angular momentum of a cosmological halo can be described as a function of the dimensionless spin parameter , @xmath214 , where j is angular momentum , e is the total energy , g is the gravitational constant and m is the halo mass . this is roughly equivalent to the ratio of the angular momentum of the halo to the angular momentum needed for the halo to be rotationally supported . mpc , 0.45 h@xmath59 mpc and 0.6 h@xmath59 mpc comoving box sizes , respectively . the filled shapes corresponds to the average value for simulations with that box size . the open circle corresponds to the average for all simulations together . , scaledwidth=90.0% ] panel ( a ) of figure [ fig.comp.panel2 ] shows the gas and dark matter spin parameters plotted against each other for the halo in each simulation that forms the primordial protostellar core , at the epoch of collapse . the mean value of the dark matter spin parameter is approximately 0.05 . the gas spin parameter tends to have somewhat lower values , with a mean of @xmath215 , but , considering the number of halos examined , is consistent with the dark matter spin parameter . there appears to be some overall positive correlation between the dark matter and baryon spin parameters ( e.g. halos with higher overall dark matter angular momentum tend to have higher overall baryon angular momentum ) but there is considerable scatter . in all cases the spin parameters are much less than one , which suggests that the halos have little overall angular momentum . the measured range and distribution of the spin parameters agree well with previous analytical and numerical results which predict typical values for the spin parameter of @xmath216 , with a mean value of @xmath217 ( barnes & efstathiou 1987 ; gardner 2001 ; van den bosch et al . 2002 ; yoshida et al . 2003 ) . panel ( b ) of figure [ fig.comp.panel2 ] plots dark matter spin parameter vs. collapse redshift of the halo . there is no evidence that the collapse redshift systematically depends on the spin parameter . panels ( c ) and ( d ) of figure [ fig.comp.panel2 ] plot the baryon and dark matter spin parameters against the halo virial mass . as with the other quantities examined , there is considerable scatter in the distributions , but no evidence for a clear relationship between halo virial mass and either gas or dark matter spin parameter . in all of the panels of this figure there is no evidence for any systematic effect due to box size or collapse redshift . figure [ fig.comp.panel3 ] plots the angle between the overall dark matter and baryon angular momentum vectors ( @xmath218 ) versus several different quantities . panel ( a ) plots @xmath218 vs. halo virial mass at the time of formation of the pop iii protostar in each halo . overall , the average value for @xmath218 is approximately 25 degrees , which is consistent with recent numerical simulations ( van den bosch et al . 2002 ; yoshida et al . there is a great deal of scatter in @xmath218 , which is also consistent . there is no evidence for correlation between @xmath218 and halo virial mass . panel ( b ) plots @xmath218 vs. collapse redshift for each simulation , and panels ( c ) and ( d ) plot the gas and dark matter spin parameters vs. @xmath218 , respectively . there appears to be no correlation between @xmath218 and collapse redshift or the gas or dark matter spin parameters , and no evidence of there being any systematic effect due to box size . mpc , 0.45 h@xmath59 mpc and 0.6 h@xmath59 mpc comoving box sizes , respectively . the filled shapes corresponds to the average value for simulations with that box size . the open circle corresponds to the average for all simulations together.,scaledwidth=90.0% ] in addition to plots of mean halo properties , it is very useful to look at more detailed information about each halo . figures [ fig.comp.panel4 ] through [ fig.comp.panel8 ] show spherically - averaged , mass - weighted radial profiles of several baryon quantities in eleven of the twelve simulations ( one simulation crashed and could not be restarted before reaching a high enough density ) . since the cores of the most massive halo in each simulation collapse at a range of redshifts , we choose to compare them at a fixed point in the halo s evolution , as measured by the peak central baryon density in the protostellar cloud , which is roughly analogous to a constant point in the evolution of the protostar . in each of the figures discussed below , panel ( a ) shows radial profiles for all of the simulations plotted together . panel ( b ) shows only the results for the 0.3 h@xmath59 mpc box , panel ( c ) shows only results for the 0.45 h@xmath59 mpc box , and panel ( d ) shows only results for the 0.6 h@xmath59 mpc box . line of a given line type and weight correspond to the same simulation in all figures . figure [ fig.comp.panel4 ] shows the plots of number density as a function of radius for eleven simulations , shown at approximately the same point in their evolution . there is remarkably little scatter in the density profiles for all of the simulations ( less than 1 dex at any given radius ) , and the density profiles all tend towards @xmath219 . it was shown by & sweigart ( 1968 ) that for a cloud of gas that is nearly isothermal and slowly rotating and which has negligible support from a magnetic field , the subsonic evolution of the gas will tend to produce a @xmath220 density distribution as long as the thermal pressure remains approximately in balance with the gravitational field . in particular , ( 1939 ) showed that a molecular cloud core which forms at subsonic speeds will tend towards the density distribution of a singular isothermal sphere , @xmath221 where @xmath222 is the isothermal sound speed , t , k , and m are the temperature , boltzmann s constant , and mean molecular weight of the gas , respectively , and and r is the radius . since the halos generally have low angular momentum ( as seen in figure [ fig.comp.panel2 ] ) and magnetic fields are completely neglected in these simulations , it is reasonable that the density should go as @xmath223 in all of the simulations , with the deviation from @xmath224 being a result of the effective equation of state of the primordial gas ( omukai & nishi ( 1998 ) ) . the overall normalization of the density profiles also agrees very well . this can be understood as a result of the cooling properties of hydrogen gas . each of the halos examined in this figure has the same composition , and therefore is cooled by the same mechanism . only a small amount of molecular hydrogen is needed to cool the gas relatively efficiently , suggesting that in a halo that is in a somewhat stable equilibrium the gas temperature at low densities should be approximately constant for different halos , independent of the molecular hydrogen fraction . at densities above @xmath225 @xmath19 the cooling rate becomes independent of density and the halo enters a `` loitering '' phase , where a slow buildup of mass occurs . once a bonnor - ebert mass of gas accumulates , the halo core becomes gravitationally unstable and begins to evolve very rapidly , so small differences in the initial molecular hydrogen fraction become magnified ( as discussed later in this section ) . the scatter in densities at a given radius is related to the scatter in virial masses , and there is no systematic effect with collapse redshift . mpc , ( c ) 0.45 h@xmath59 mpc , ( d ) 0.6 h@xmath59 mpc ( comoving ) . one of the 0.45 h@xmath59 mpc simulations has been omitted since the simulation crashed before reaching a comparable density value . line thicknesses and types correspond to individual simulations . l0_3a : thick solid line . l0_3b : thick dotted line . l0_3c : thick short - dashed line . l0_3d : thick long - dashed line . l0_45a : thick dot - short dashed line . l0_45b : thick dot - long dashed line . l0_45c : thin solid line . l0_45d : thin dotted line . l0_6a : thin short - dashed line . l0_6b : thin long - dashed line . l0_6c : thin dot - short dashed line . l0_6d : thin dot - long dashed line . , scaledwidth=90.0% ] figure [ fig.comp.panel5 ] shows the baryon temperature as a function of radius . at radii outside of @xmath226 parsec , the temperature profiles are similar between all of the simulations , though halos forming at higher redshifts tend to have a higher overall temperature . at smaller radii there is significant scatter in core temperature of the simulations ( for a fixed density ) , with a systematic trend towards halos forming in larger boxes having a lower overall core temperature . examination of figure [ fig.comp.panel6 ] plots molecular hydrogen mass fraction as a function of radius , and shows that halos which form at higher redshift have systematically larger h@xmath44 mass fractions , though this effect is much more pronounced in the core of the halo than in the envelope . this difference in molecular hydrogen fraction can be understood as a result of the overall halo temperature . the rate at which molecular hydrogen is produced at low densities is limited by the availability of free electrons , as described by ( 1997 ) . the mean fraction of free electrons available in the primordial gas is a function of baryon temperature , with larger temperatures corresponding to larger electron fractions . on the other hand , the rate at which molecular hydrogen forms via the @xmath39 channel declines at high temperatures . since the limiting reaction in the formation of molecular hydrogen via the @xmath39 channel is the formation of free electrons , this reaction dominates , and it can be shown using a simple one - zone calculation following the nonequilibrium primordial chemistry that molecular hydrogen production is maximized at @xmath227k . halos with higher overall baryon temperatures will have systematically higher molecular hydrogen fractions . once the core of the halo begins to collapse to very high densities small differences in the molecular hydrogen fraction are amplified , resulting in a general trend towards halos with higher overall baryon temperatures having higher h@xmath44 fractions in their cores , and thus lower central temperatures . 0.3 h@xmath59 mpc , ( c ) @xmath228 0.45 h@xmath59 mpc , and ( d ) @xmath228 0.6 h@xmath59 mpc ( comoving ) . line thicknesses and types are identical to those shown in figure [ fig.comp.panel4 ] . , scaledwidth=90.0% ] 0.3 h@xmath59 mpc , ( c ) @xmath228 0.45 h@xmath59 mpc , and ( d ) @xmath228 0.6 h@xmath59 mpc ( comoving ) . line thicknesses and types are identical to those shown in figure [ fig.comp.panel4 ] . , scaledwidth=90.0% ] figure [ fig.comp.panel7 ] plots the mean baryon radial velocity as a function of radius . this figure shows that there is a clear systematic effect present , where halos forming in simulations with larger boxes having a significantly lower overall radial velocity at small radii . this translates directly to a lower overall accretion rate onto protostellar forming in halos in larger simulation volumes , which is discussed more thoroughly in section [ sec : halo - accretion ] . as discussed in section [ repstar ] , this can be understood using the shu isothermal sphere model , where subsonic collapse of gas onto the core of the sphere occurs at a rate controlled by the sound speed . since the core temperatures are lower overall in the large simulation volumes , this translates to a lower sound speed and overall lower radial infall velocity . a lower infall velocity implies a lower accretion rate , which may have profound implications for the population iii imf and are discussed elsewhere in this work . 0.3 h@xmath59 mpc , ( c ) @xmath228 0.45 h@xmath59 mpc , and ( d ) @xmath228 0.6 h@xmath59 mpc ( comoving ) . line thicknesses and types are the same as in figure [ fig.comp.panel4 ] . , scaledwidth=90.0% ] figure [ fig.comp.panel8 ] shows the baron circular velocity in each halo as a function of radius . there is a great deal of scatter in this relationship , though there is no clear trend with simulation box size . in all cases the overall circular velocity is significantly less than the keplerian orbital velocity , which agrees with our previous observation that the halos have generally low angular momentum . 0.3 h@xmath59 mpc , ( c ) @xmath228 0.45 h@xmath59 mpc , and ( d ) @xmath228 0.6 h@xmath59 mpc ( comoving ) . line thicknesses and types are the same as in figure [ fig.comp.panel4 ] . , scaledwidth=90.0% ] figures [ fig.accrete.panel1 ] [ fig.accrete.panel4 ] show baryon properties relating to the accretion rate onto the forming protostellar core . many of these properties are also detailed in tables [ table.meaninfo-1 ] and [ table.meaninfo-2 ] . figure [ fig.accrete.panel1 ] shows the mass - weighted , spherically averaged accretion time as a function of enclosed gas mass for 11 of the 12 cosmological simulations . there is a 2 order of magnitude scatter in accretion rates , with a trend towards lower accretion rates ( longer accretion times ) in simulations that have larger box sizes . figure [ fig.accrete.panel1a ] shows the mass - weighted , spherically averaged accretion rate onto the protostellar core as a function of time for all twelve cosmological simulations . the time on the x - axis is calculated by summing up @xmath229 on spherical shells , where @xmath230 is the amount of baryon gas and @xmath14 is the instantaneous gas accretion rate at that shell , starting at the center of the halo and assumes no feedback . although this is only a crude estimate of the true accretion rate , we see that all curves rise quickly to a maximum accretion rate at @xmath231 year , and decline thereafter , in agreement with the results of & loeb ( 2004 ) and ( 2006 ) . mpc box simulations . panel ( c ) : 0.45 h@xmath59 mpc box simulations . panel ( d ) : 0.6 h@xmath59 mpc box simulations . individual simulations are identified by line thicknesses and types identical to those in figure [ fig.comp.panel4 ] . the baryon accretion time is defined as @xmath232 , where @xmath187 is the enclosed baryon mass and @xmath188 , with @xmath189 and @xmath190 being the spherically - averaged baryon density and velocity as calculated on spherical shells , and @xmath190 defined as being positive towards the center of the halo . the dot - long dashed line in each panel is the kelvin - helmholtz time for a population iii star with a mass identical to the enclosed mass , as calculated from the results given by schaerer . the long dash - short dashed line is the lifetime of a massive primordial star assuming no significant mass loss , as calculated by schaerer . note that this fit only covers the mass range @xmath195 m@xmath7 , by @xmath200 constant for @xmath201 m@xmath7 . the dot - short dashed line in each panel is the baryon accretion time for the result in abn02 . the upper and lower diagonal thin solid lines are the accretion times for constant accretion rates of @xmath155 and @xmath25 m@xmath7/yr , respectively . , scaledwidth=70.0% ] mpc box simulations . panel ( c ) : 0.45 h@xmath59 mpc box simulations . panel ( d ) : 0.6 h@xmath59 mpc box simulations . line thicknesses and types are the same as in figure [ fig.comp.panel4 ] . the accretion rate is defined as @xmath188 , with @xmath189 and @xmath190 being the spherically - averaged baryon density and velocity as calculated on spherical shells , and @xmath190 defined as being positive towards the center of the halo . the time is calculated by summing @xmath229 in each shell starting at the center of the halo and working outward , and plotting vs. @xmath14 . , scaledwidth=90.0% ] figure [ fig.accrete.panel2 ] shows the mass accretion rate vs. halo virial mass , molecular hydrogen fraction vs. halo virial mass , mass accretion rate vs. molecular hydrogen fraction , and mass accretion rate vs. halo collapse redshift . in all panels in this figure the molecular hydrogen fraction is taken as an average over all gas within the virial radius , and accretion rate values are taken instantaneously at spherically - averaged radial shells corresponding to an enclosed baryon mass of 100 m@xmath7 ( note that the trends here do not change significantly if this mass shell is varied by a factor of 10 in either direction , so the results are quite robust ) . this figure shows that there is a general trend of higher mass accretion rates with increasing halo mass . though there is a great deal of scatter , and the variance is virial mass is not that large . there is no evidence for a comparable trend in the molecular hydrogen fraction with virial mass . the bottom panels show that there is a strong relationship between the mass accretion rate and the mean molecular hydrogen fraction of the halo and the redshift of collapse . the accretion rate steadily decreases with increasing h@xmath44 fraction , and increases with decreasing redshift ( increasing time ) . ) vs. halo virial mass . panel ( b ) : molecular hydrogen fraction vs. halo virial mass . panel ( c ) : mass accretion rate vs. molecular hydrogen fraction . panel ( d ) : mass accretion rate ( @xmath14 ) vs. halo collapse redshift . the @xmath33 fraction is taken as an average over all gas within the virial radius . all values of accretion rate are taken instantaneously at spherically - averaged radial shells corresponding to an enclosed gas mass of 100 @xmath233 . in each plot , open triangles , squares and hexagons correspond to simulations with 0.3 h@xmath59 mpc , 0.45 h@xmath59 mpc and 0.6 h@xmath59 mpc comoving box sizes , respectively . , scaledwidth=90.0% ] the plots shown in figure [ fig.accrete.panel3 ] help to explain these trends . this figure shows the mean halo temperature scaled by virial mass , core molecular hydrogen fraction , and core baryon temperature as a function of collapse redshift , and the core temperature as a function of molecular hydrogen fraction . all `` core '' quantities are measured at the epoch of collapse at a proper radius of 0.1 pc , though changing this value does not change the results significantly . there is a clear relationship between collapse redshift and mean temperature , with temperature decreasing with decreasing collapse redshift ( increasing time ) . this is explainable , as discussed previously , as a simple virial scaling : halo masses at the time of protostellar core formation stay approximately constant with redshift , and since the virial temperature scales as @xmath209 , higher redshift halos have systematically higher temperatures . the limiting reaction in the formation of molecular hydrogen at low ( @xmath234 @xmath19 ) densities is the formation of h@xmath235 , which scales as @xmath236 . this suggests that halos whose gas is overall warmer ( up to @xmath142 k , where high temperatures begin to collisionally dissociate h@xmath44 ) more molecular hydrogen will be produced . this is borne out in the plot of core molecular hydrogen fraction vs. collapse redshift , where halos that form at later times ( and in overall cooler objects ) have less molecular hydrogen . the results of this are shown quite plainly in the plot of core temperature vs. collapse redshift , where objects that form later have higher core temperatures and hence higher accretion rates , since @xmath237 . this relationship is shown in a different way by the plot of core temperature vs. core h@xmath44 fraction in figure [ fig.accrete.panel3 ] , where there is a clear correlation between decreasing temperature with increasing h@xmath44 fraction . vs. @xmath238 ( @xmath239 is the halo collapse redshift ) . panel ( b ) : core @xmath33 fraction vs. collapse redshift . panel ( c ) : core baryon temperature vs. collapse redshift . panel ( d ) : core baryon temperature vs. @xmath33 fraction . the dashed line in the top left panel presents the expected virial scaling of temperature with mass and redshift , arbitrarily scaled . halo mean temperature is a mass - weighted average of all gas within the virial radius . all `` core '' quantities are measured at the epoch of collapse at a proper radius of 0.1 pc . in each plot , open triangles , squares and hexagons correspond to simulations with 0.3 h@xmath59 mpc , 0.45 h@xmath59 mpc and 0.6 h@xmath59 mpc comoving box sizes , respectively . , scaledwidth=90.0% ] figure [ fig.accrete.panel4 ] further explores the correlations shown in figure [ fig.accrete.panel3 ] . panel ( a ) of figure [ fig.accrete.panel4 ] plots the mean , mass - weighted halo temperature as a function of instantaneous halo mass accretion rate . the mass accretion rate is taken by calculating the difference in virial mass between the halo at the time of protostellar core collapse and the virial mass of the most massive progenitor at the last output time and dividing by the change in redshift . there is a clear positive correlation between mean halo temperature and merger rate . this is due to the hydrodynamic heating of gas by mergers , as discussed by ( 2003 ) during halo mergers the gas is heated by shocks to approximately the virial temperature , and the gas in rapidly - growing halos does not have time to cool to low temperatures before halo core collapse . panel ( b ) shows the halo mass accretion rate as a function of halo virial mass at the epoch of core collapse . halos which are undergoing rapid mergers show a strong tendency to be more massive , suggesting that a higher mass threshold must be reached before the central density is high enough , and the @xmath33 formation and cooling rates are fast enough , for the core to collapse . we note that the apparent growth rate as a function of halo mass that we observe is higher than that of ( 2003 ) . this may be due to our significantly higher mass and spatial resolution . panel ( c ) of figure [ fig.accrete.panel4 ] shows the instantaneous accretion rate onto the protostellar core at the epoch of collapse as a function of @xmath240 , which is defined as the mean overdensity of the region within 20 virial radii of the collapsing halo . @xmath240 is a proxy for local environment , as used by ( 2005 ) regions with values of @xmath241 are denser than an average region of the universe , with larger @xmath240 values corresponding to rarer regions . there is a clear relationship between protostellar core accretion rate and environment denser regions have lower accretion rates . this is explained by panel ( d ) , which plots the mean halo molecular hydrogen fraction as a function of @xmath240 . regions with higher values of @xmath240 tend to have higher molecular hydrogen fractions and thus lower temperatures , which lead to lower accretion rates onto the protostellar core ( as shown by figure [ fig.accrete.panel2 ] c ) . . panel ( d ) : halo mean molecular hydrogen fraction vs. @xmath240 . the `` core '' accretion rate in panel ( c ) is measured at the epoch of collapse at a proper radius of 0.1 pc . in panels ( c ) and ( d ) , @xmath240 is defined as the mean density of the region within 20 virial radii of the halo ( at the time of protostellar cloud collapse ) in units of @xmath242 . in each plot , open triangles , squares and hexagons correspond to simulations with 0.3 h@xmath59 mpc , 0.45 h@xmath59 mpc and 0.6 h@xmath59 mpc comoving box sizes , respectively . , scaledwidth=90.0% ] in this paper we have explored several aspects of the formation of population iii stars in a @xmath0cdm universe . this section summarizes some of the processes neglected in our calculations and also attempts to put some of the results in context . the results presented in section [ compare - real ] demonstrate that there is a great deal of scatter between the bulk halo properties such as overall virial mass , collapse redshift , and mean halo temperature among the twelve simulations shown . however , the final state of the density profile is extremely similar . this is due to the chemical and cooling properties of the primordial gas . the minimum temperature of the gas , which is determined by its chemical composition , creates a density profile that goes roughly as @xmath224 for any gas cloud which is only supported by thermal pressure . this is for the gas contained in the halos out of which population iii stars form , so it is reasonable to expect consistent density profiles on halo scales . a detailed examination of the gas properties which may contribute significantly to the final pop iii star mass , such as the core baryon temperature and accretion rate onto the forming primordial protostar , show a significant amount of scatter . this scatter can be attributed to variations in the molecular hydrogen content of the halo on large scales brought on by differences in halo temperature as a result of varied merger rates between simulations , as suggested by figures [ fig.accrete.panel3 ] and [ fig.accrete.panel4 ] . panel ( a ) of figure [ fig.accrete.panel3 ] shows a clear correlation between halo accretion rate and temperature , suggesting that high halo merger rates cause an increase in halo temperature , which contributes to the scatter in mean temperatures for halos of a given mass . there is also a systematic relationship between halo collapse redshift and the mean temperature , with halos that collapse at higher redshift having higher overall halo temperatures and lower accretion rates onto the protostellar core . the higher temperatures due to redshift or merger rate result in somewhat larger molecular hydrogen mass fractions , which cause the halo core to cool more rapidly during collapse . since the accretion onto the primordial protostar is primarily subsonic it is primarily controlled by the local sound speed , with lower core temperatures directly resulting in lower accretion rates . we expect that as the gas collapses to higher densities ( @xmath243 ) the solutions will converge to the results found by & nishi ( 1998 ) , who find that the late - time evolution of the gas can be approximated by a larson - penston similarity solution ( 1969 ; 1969 ) . however , the lower - density cloud of gas around the protostar will not converge to a single solution , so the star will still experience accretion with a range of histories and peak values . one possible source of uncertainty in our understanding of the effect of redshift and environment on the merger rates onto protostellar cores may come from the way in which halos are selected for simulation in our calculations . in each realization , we first run a dark matter - only calculation , and choose the most massive halo at @xmath60 for re - simulation with higher mass and spatial resolution and baryonic physics . selection of the most massive halo in each calculation may result in a bias towards unusual halo formation environments and may skew our results . still , examination of figure [ fig.accrete.panel2 ] ( d ) shows significant scatter in accretion rates for halos which collapse at approximately the same redshift , and suggests that both redshift and environment effects are important . this issue could possibly be clarified by performing an ensemble of simulations using halos chosen in a different manner . after the onset of collapse , the evolution of the core of the halo , which has a mass of one thousand solar masses , becomes effectively decoupled from the halo envelope since the time scales become much shorter within the halo core . this tells us that while the formation of the initial primordial protostellar cloud is strongly coupled to the time scales associated with cosmological structure formation , once the cloud has collapsed we can treat the core of the halo separately from the rest of the calculation . this decoupling will become highly useful when more detailed calculations of the evolution of population iii protostars , including more complicated physics such as radiative transfer , magnetohydrodynamics , and protostellar accretion models , are performed , and will save us significant computational cost . in section [ repstar ] , we compare accretion times for a representative protostellar core using accretion rates estimated with two different methods : a simple calculation of instantaneous mass flux through shells using baryon density and radial velocity , and an estimate based solely on baryon temperature using the shu isothermal collapse model . these two values agree within a factor of two over a large range of mass scales ( @xmath244 m@xmath245 m@xmath7 ) . the reason for this bears examination . as shown by ( 1977 ) , as long as the densities in a condensing molecular cloud core span several orders of magnitude before a stage of dynamic instability is reached , the subsequent collapse properties of the cloud should resemble those of an isothermal sphere . the lack of characteristic time and length scales results in a self - similar wave of infalling gas which propagates outward at the speed of sound , resulting in an accretion rate that scales as the cube of the sound speed . this accretion rate can be derived in a more intuitive way by considering the properties of a cloud of gas with radius r and mass @xmath246 which is marginally unstable . the accretion rate of this gas must be given ( as an order of magnitude estimate ) by @xmath247 , where @xmath248 , where a is the characteristic velocity associated with the cloud ( the virial velocity ) . if this cloud was originally marginally supported against its self - gravity , then @xmath249 ( where g is the gravitational constant ) , which can be substituted into the expression for @xmath14 to give @xmath250 , independent of r. in the case of this quasi - statically collapsing cloud , the virial speed is comparable to the sound speed @xmath160 , giving @xmath251 . while the shu model assumes that the entire cloud is of a constant temperature , our calculations have a varying temperature as a function of radius , and a radially - varying accretion rate based on this temperature is an excellent fit . this is reasonable because the isothermal collapse model assumes that the infall wave propagates at the local sound speed , assuming that the cloud is not supported by any other means . in this calculation we completely neglect the effects of magnetic fields , and it can be seen from figure [ fig.rep.panel3 ] that rotation contributes an insignificant amount of support to our representative protostellar cloud ( as well as in the other 11 simulations discussed in this paper ) , resulting in gas pressure being the sole means of support of the cloud . one slight difference between the shu model and our calculations is the presence of the dark matter halo , which results in a slightly steeper density profile ( @xmath252 ) , though the accretion rate result , being a function of sound speed , is not changed significantly . the observation that the rate of accretion onto the primordial protostar varies systematically as a function of redshift , with halos that collapse at higher redshift having an overall lower accretion rate , has significant implications for both reionization and metal enrichment of the early universe . there is no trend in the bottom right panel of figure [ fig.accrete.panel2 ] that suggests an accretion rate `` floor '' exists at high redshifts . however , one can make an estimate of the minimum possible accretion rate by observing that molecular hydrogen is only effective at cooling the primordial gas down to approximately 200 kelvin , which gives us an accretion rate using the shu model of @xmath253 m@xmath7/year , which is reasonably close to the accretion rates observed in the highest - redshift halos which have collapsed . though this implies convergence , it would be prudent to perform another suite of calculations at an even larger box size to be sure . an absolute minimum accretion rate is set by the minimum temperature that primordial gas can attain , which is effectively that of the cosmic microwave background , which scales as t@xmath254 . this gives a minimum accretion rate using the shu model of @xmath255 how reasonable is this estimate ? as shown in figure [ fig.rep.accrete ] , the shu model does not agree perfectly with the accretion time results derived using baryon densities and infall velocities , but does appear to provide a reasonable approximation for this one simulation . examination of our entire suite of simulations does not support this , however . figure [ fig.disc.mdotmshu ] plots the ratio of the accretion rate calculated from baryon density and radial velocity with that estimated using the shu model . lines are shown for only 8 of the 12 simulations for clarity . this plot shows that simulated accretion rates depart significantly from the shu model , which can be too high or low by a factor of several . this suggests that the shu model , while useful for a qualitative understanding of the physical system , should not be considered to be quantitatively accurate for the purposes of estimating accretion rates . , though for readability only simulations from the 0.3 and 0.6 @xmath256 mpc box are plotted . , scaledwidth=80.0% ] the final masses of the stars in the calculations discussed in this work remains unclear . these simulations lack the necessary radiative feedback processes which might halt accretion onto the protostar , making it impossible to accurately determine the main - sequence mass of the star . however , rough bounds on the mass range of these objects can be determined by examining figure [ fig.accrete.panel1 ] and applying arguments similar to those used by abn02 . a one solar mass object evolves far too slowly to halt accretion , particularly considering the high rates at which mass is falling onto the star ( @xmath257 m@xmath7/year ) . this suggests that the star will have a mass which is significantly higher than 1 m@xmath7 . as discussed by ( 2006 ) , low - mass protostars evolve on a timescale short compared to the kelvin - helmholtz timescale , whereas the opposite is true for high - mass stars . thus , the kelvin - helmholtz timescale ( plotted in figure [ fig.accrete.panel1 ] using values of stellar radius and luminosity from ( 2002 ) ) is a reasonable minimum mass scale for these objects , giving a lower bound to the stellar masses determined from these calculations of @xmath258 m@xmath7 , depending on the simulation . massive stars can move onto the main sequence while still accreting rapidly . however , at high masses these objects have very short lifetimes ( @xmath259 years for a several hundred solar mass star ) and this provides a reasonable upper limit to the mass of the star . comparing the stellar lifetime of these objects ( as given by ( 2002 ) , we derive an upper mass limit of @xmath260 m@xmath7 , with individual stars having mass ranges that vary between @xmath261 m@xmath7 and @xmath262 m@xmath7 , using this simple mass range estimate , with the smaller mass objects forming systematically at higher redshifts . this simple estimate fails to take into account essential physics . as discussed by & palla ( 2003 ) and & mckee ( 2004 ) , radiative feedback from the evolving protostar will strongly affect the dynamics of the accreting gas . & mckee ( 2004 ) and mckee & tan ( 2006 , in prep . ) use a combination of analytic and semi - analytic models of the evolving system , and show that feedback does not limit accretion until the star has reached @xmath16 m@xmath7 , and suggest that ionization creates an hii region that limits growth of stars to masses less than @xmath263 m@xmath7 . they assume a very high ( and time varying ) accretion rate compared to the mean values observed in our suite of simulations , and it is unclear how varying this would affect the final mass of the object . & palla ( 2003 ) perform 1d , spherically - symmetric calculations of the evolution of a population iii protostar undergoing rapid accretion . they include radiative feedback effects , and find that there is a critical mass accretion rate , @xmath12 m@xmath7/year , below which objects can accrete to masses much greater than 100 m@xmath7 ( essentially up to the limit of gas available in the halo core and the main sequence lifetime of the forming object ) , and below which the maximum mass decreases as the accretion rate increases . they find that if they assume a time - varying accretion rate that stars at high values and decreases rapidly with time , the evolution of the star follows the @xmath264 path , and accretion can continue until the end of the star s main sequence lifetime . as can be seen from this discussion , it seems that the arguments for mass range given by abn02 are overly simplistic . however , the models discussed above do not completely resolve the issue of population iii imf tan & mckee use a single accretion rate for their work ( which includes multi - dimensional effects ) , and omukai & palla perform spherically - symmetric calculations which may miss essential accretion shutoff physics . clearly , the relationship between accretion rates and final stellar mass can not be stated with any certainty , and will have to be addressed using multi - dimensional numerical simulations which include protostellar evolution and radiative feedback . however , if one assumes that the main - sequence lifetime of an accreting object is a reasonable proxy for the maximum stellar mass , it seems that population iii stars that form at higher redshifts should generally have smaller main - sequence masses . if in fact a lower overall accretion rate results in a less massive population of stars , these objects will be much less effective at ionizing the intergalactic medium ( since they produce overall fewer uv photons per baryon ) and will produce a completely different nucleosynthetic signature . this may be in accord with the wmap year iii measurements of the polarization of the cosmic microwave background , which suggests that the universe was significantly reionized by @xmath265 ( page et al . models show that the observed level of polarization does not require a large contribution to reionization from population iii stars . additionally , ( 2006 ) uses a semianalytical model constrained by the metallicity distributions of population ii stars in the galactic halo , relative abundances of ultra metal poor halo stars , and estimates of the ionizing photon budget required to reionize the igm , and suggests that population iii stars have an imf which has a mean mass of @xmath266 m@xmath7 . one factor that has been completely ignored in this work is the effect of a soft uv background . massive primordial stars are copious emitters of light in the lyman - werner band ( @xmath267 ev ) . atomic hydrogen is transparent to these photons , which photodissociate molecular hydrogen . this effect , explored by ( 2001 ) and others , increases the threshold mass of halos which form pop iii stars and may also result in an increase in temperatures in the halo cores , causing an increase in the overall accretion rate onto the forming protostars . we explore this issue in detail in a forthcoming paper ( oshea & norman 2006 , in preparation ) . a semianalytical examination of this effect was made by & abel ( 2005 ) , who use press - schechter models of population iii star formation to predict a slowly rising lyman - werner background which provides some support to this idea . this suggests that further calculations including larger simulation volumes as well as a soft uv background will be necessary to make a definitive statement about the most common accretion rates . additionally , these calculations completely neglect the mode of primordial star formation that takes place in halos whose virial temperatures are above @xmath28 k. cooling in these systems is dominated by atomic hydrogen line emission and , particularly in the presence of a strong soft uv background , may result in a much larger amount of cold gas distributed in a different manner than in the systems simulated in this work , which have mean virial masses of a few times @xmath268 m@xmath7 and virial temperatures of around 1000 k. these systems have been examined numerically by & loeb ( 2003 ) , but further work is necessary to explore the properties of these objects . figure [ fig.rep.panel3 ] in section [ repstar ] shows clear signs of angular momentum evolution in the collapsing protostellar core . angular momentum is less of an issue in population iii star formation than in galactic star formation - the collapsing cosmological halo out of which the protostellar core forms has very little angular momentum from the outset ( with a spin parameter of @xmath269 ) , and violent relaxation during virialization results in an angular momentum distribution @xmath270 , so circular velocities are fairly constant throughout the halo . the @xmath142 m@xmath7 halo core out of which the protostar forms is not rotationally supported the gas is essentially completely held up by thermal pressure , and the small amount of angular momentum that is actually transported is not a critical factor in the cloud core s collapse . this angular momentum redistribution , as discussed in section [ repstar ] , appears to be due to a combination of angular momentum segregation and turbulent transport . similar results were reported by ( 2006 ) . a detailed analysis of angular momentum transport issues will be discussed in the second paper in this series ( oshea , norman & li 2006 , in preparation ) . though the collapsing protostellar cloud is not rotationally supported by the time the central density reaches @xmath271 @xmath19 , it is inevitable that at some point a protostar with a rotationally - supported accretion disk will form , as discussed in & mckee ( 2004 ) . unfortunately , this takes place at high densities , where the optically - thin assumption used in the enzo cooling algorithms breaks down , which necessitates either an extension of the cooling model ( ripamonti & abel 2004 ) or the use of a radiation hydrodynamics code such as zeus - mp ( hayes et al . 2005 ) . this issue will be approached in later work . the primordial chemistry model used in these calculations ignores the effects of deuterium , lithium , and the various molecules they form between these elements and ordinary hydrogen . deuterium and lithium have been shown to be unimportant in the temperature and density regimes that we have examined in this work ( galli & palla 1998 ) . however , it is possible that they may be relevant in other situations of importance to population iii star formation in particular , regions which have been ionized to very high electron fractions may experience significant cooling from the hd molecule , which has a low excitation energy and forms preferentially compared to @xmath33 due to chemical fractionation . this is particular true at low temperatures ( @xmath272 k ) where essentially all deuterium is converted to hd . cooling due to the hd molecule has the potential to cool gas down to approximately the temperature of the cosmic microwave background , which scales with redshift as @xmath273 k ( flower et al . 2000 ; galli & palla 2002 ; lipovka , nez - lpez , & avila - reese 2005 ) . this gives a minimum baryon temperature of approximately @xmath274 kelvin at @xmath10 and could further reduce the minimum accretion rate onto a primordial protostar . lithium , while in principle an effective coolant when in lih , is safely ignored , since only a tiny fraction of lithium is converted into molecular form ( mizusawa et al . 2005 ) . a final chemical process which is omitted from our calculation is heating caused by molecular hydrogen formation at high densities ( @xmath275 @xmath19 ) , the inclusion of which which may result in differences in the temperature evolution of the gas in the highest density regimes considered here . this does not affect the main conclusions of our paper , as significant differences between halos can be observed at densities much lower than @xmath276 . the effects of magnetic fields are completely ignored in the simulations discussed in this work . we can justify this by examining the magnetic field necessary to delay the collapse of the halo core . if one assumes that the halo core can be represented reasonably well by an isothermal sphere of constant density ( which is reasonable at the onset of halo collapse ) , we can use the virial theorem to estimate the strength of the magnetic field which is necessary to support the collapse of the halo against gravity . assuming flux freezing and a uniform magnetic field , a magnetically critical isothermal sphere has a mass - to - flux ratio of @xmath277 where m@xmath278 is the mass of the halo , @xmath279 is the magnetic flux in the cloud ( with r@xmath278 and b@xmath278 being the cloud radius and magnetic field strength , respectively ) , and g is the gravitational constant . reasonable values for m@xmath278 and r@xmath278 are @xmath280 m@xmath7 and 4 parsecs , respectively , which gives us a value of the magnetic field of b@xmath281 g. the mean density of the cloud is n@xmath282 @xmath19 and the mean density of the universe at @xmath283 ( the redshift that our cloud collapses ) is @xmath146 @xmath19 , so if we assume a spherical collapse from the mean cosmic density assuming flux freezing , we see that the ratio of the magnetic field in the cloud to the mean universal magnetic field is @xmath284 this gives us a mean magnetic field of @xmath285 g at @xmath286 . since there are no known objects that may produce magnetic fields between recombination ( @xmath287 ) and the epoch of pop iii star formation , and the strength of the magnetic field scales with the expansion of the universe as @xmath288 , we estimate that in order for magnetic fields to be dynamically important in the formation of population iii stars the magnetic field strength at recombination must be @xmath289 g. the current observational upper limit to magnetic field strength at recombination ( albeit at large scales ) is b @xmath290 3 x 10@xmath291 g as measured at the present epoch ( jedamzik , katalini , & olinto 2000 ) , which corresponds to a magnetic field at recombination of approximately @xmath292 g. this is three orders of magnitude higher than needed to be dynamically relevant for population iii star formation ! the only known mechanism that could generate coherent magnetic fields with strengths of this scale is a discontinuous qcd or electroweak phase transition . however , the mechanisms involved are highly speculative and predictions of the possible magnetic field strengths are unreliable ( sigl , olinto , & jedamzik 1997 ) . currently , the most plausible mechanisms for creating magnetic fields at recombination suggest that field strengths are on the order of @xmath293 g at recombination ( matarrese et al . 2005 ) . given the observational uncertainty , it seems reasonable to ignore this effect , though future simulations will certainly include magnetic fields with a variety of field strengths and physical scales .. assuming that the magnetic field at that epoch was strong enough to be dynamically important , we can calculate the effect that this has on the collapse of the star . due to the relatively high electron fraction in the halo core ( @xmath294 at @xmath295 pc ; see figure [ fig.rep.panel2 ] ( b ) ) the assumption of flux freezing in the magnetic field at scales of @xmath143 pc is valid ( maki & susa 2004 ) . however , the electron fraction rapidly declines at smaller radii , reaching @xmath296 at @xmath156 au , and this assumption may no longer be relevant . magnetic fields couple to charged particles ( electrons and ions ) , and these charged particles interact with the neutral medium . at high levels of ionization , collisions between charged and neutral particles are frequent , implying that the magnetic field is strongly coupled to the gas . however , at low levels of ionization there are few charged particles , and the coupling with the neutral gas is weak . in an object that is subject to a gravitational acceleration this produces a relative drift of charged and neutral particles which allows the neutral gas to decouple from the magnetic field . this effect is known as `` ambipolar diffusion , '' and is believed to be an extremely important process in galactic star formation ( mouschovias & morton 1991 ) . the retardation effect that ambipolar diffusion may have on the collapse of the halo core can be estimated by examining the relative timescales of ambipolar diffusion and halo collapse . the ambipolar diffusion timescale can be estimated as @xmath297 where l and @xmath298 are a characteristic length scale and the neutral - ion relative drift velocity , respectively , and @xmath299 is the overall ionization fraction . a proxy for the halo collapse time scale is the free fall time , which for a spherical system is @xmath300 where n is the particle number density in particles per cubic centimeter and g is the gravitational constant . the relevance of ambipolar diffusion can be estimated by taking the ratio of these two quantities , which is known as the `` collapse retardation time , '' @xmath301 . substituting in equations [ eqn - ambipolar ] and [ eqn - dyntime ] , we see that @xmath302 examination of figures [ fig.rep.panel1 ] and [ fig.rep.panel2 ] show that at the final timestep in the calculation , the number density can be fitted by a power law and is roughly @xmath303 @xmath19 while the ionization fraction scales roughly as @xmath304 . plugging these into equation [ eqn - collretard ] shows that @xmath305 is constant with radius . this is only a crude approximation , since the free fall time really should depend on the mean number density instead of the number density at a given radius . considering the rapid falloff of density , n(r ) is a reasonable approximation of n strictly speaking , for a cloud with a density profile that scales as @xmath224 over many orders of magnitude in radius , the mean density is equal to @xmath306 , so our estimate of the free fall time is too high by a factor of @xmath307 . plugging this in to the equation , we get that @xmath308 everywhere , which indicates significant delay in collapse with respect to the free fall time . however , the relevant time scale in this case is more appropriately the quasistatic collapse time , which is approximated as @xmath309 . figure [ fig.rep.panel3 ] shows that the mean radial velocity at the scales of interest ( @xmath310 parsecs ) is roughly 0.5 km / s . this corresponds to @xmath311 and scales linearly with the radius . comparison with the ambipolar diffusion time scale shows that @xmath312 and @xmath313 are within a factor of two of each other , which suggests that the presence of a magnetic field would not significantly impede the collapse of the halo core for the quasistatic collapse case . it is interesting to note that & pineau des forts ( 2003 ) reach a similar conclusion using a significantly different method . there are several effects which we have neglected in these calculations which may affect our results . the box sizes of the calculations presented in this paper are relatively small ( 300 600 kpc / h comoving ) . & loeb ( 2004 ) show that finite box sizes can have significant effects on statistical properties of halo formation , resulting in an overall bias towards under - sampling of the mass function and late halo formation times . however , our calculations are not intended to provide a statistically accurate sampling of the mass function of dark matter halos . rather , the intent is to gain an understanding of the first halo to undergo baryonic core collapse in a simulation volume . an obvious extension of this work would be to simulate a much larger simulation volume so that cosmic variance and possible environmental effects are considered ( as discussed by barkana & loeb ) , and then simulate multiple halos within this volume . an example of this technique is presented by ( 2005 ) , who simulate the formation of the first star - forming halo in a simulation volume which is @xmath314 h@xmath59 mpc on a side . they find that the first halo of mass @xmath315 h@xmath59 m@xmath7 forms by @xmath316 , and would be able to collapse by @xmath317 , when the halo has reached a mass of @xmath318 h@xmath59 m@xmath7 , which agrees reasonably well with estimates for a minimum halo mass to collapse and form a population iii star , as determined in this work . the virial temperature of this halo would be a factor of 2.4 higher than that of a halo which virializes at @xmath10 , and the @xmath33 production rate would be approximately 2.2 times higher . compared to a halo that forms at @xmath319 the highest redshift halo we examine the differences in virial temperature and @xmath33 formation rate are 1.45 and 1.39 , respectively . though somewhat simplistic , these arguments suggest that the range of halo properties explored in our set of simulations may encompass most of the expected range of behavior . though the reed et al . simulations do not include baryonic physics , such a calculation is technically possible using the enzo code , and is a logical extension of our work . another important consideration is the generation of cosmological initial conditions . ( 2006 ) explore the evolution of the halo mass function at high redshifts , and conclude that the overall mass function is suppressed in simulations whose initial conditions are generated at too late of a redshift . they suggest that a starting simulation redshift of @xmath320 would be appropriate for the small boxes used in the calculations discussed in this paper . the magnitude of error involved in simulations s tarting later is not quantified . since our focus is only on the most massive halo to form in the simulation volume , the magnitude of the bias in our results by starting at @xmath321 is unclear . a fundamental assumption that we make concerns the validity of the cold dark matter model at mass and spatial scales comparable to , or greater than , the masses of the halos which form population iii stars . reliable measurements of the cosmological power spectrum only exist down to the length scales corresponding to the lyman-@xmath322 forests , which is composed of structures that are much more massive than @xmath61 m@xmath7 ( mcdonald et al . 2005 ; jena et al . 2005 ; spergel et al . if the power spectrum is suppressed at small scales , this could have significant implications for the formation epoch and properties of population iii stars ( oshea & norman 2006 ) . additionally , a running power spectrum or significant changes in @xmath323 or the primordial power spectrum index ( as suggested by the wmap year iii results ) may have significant effects , which should be explored . in this paper we have performed a suite of twelve high dynamical range simulations of the formation of population iii stars in a @xmath0cdm universe , with the intent to further constrain the variation of accretion rates onto population iii protostellar cores , a parameter that is of crucial importance to the mass scale of population iii stars . our simulations span a range of box sizes and are each an independent cosmological realization . our principal results are as follows : \1 . our results are quantitatively and qualitatively similar to those of abn02 , which was a single realization performed in an @xmath324 scdm universe . this suggests that the scenario presented in that work is generally correct . we note that ( 2006 ) have come to the same conclusion using different numerical techniques . we observe no signs of fragmentation in the collapsing protostellar core in any of our 12 simulations , suggesting that the paradigm of a massive population iii star forming in isolation in a halo of mass @xmath325 m@xmath7 is robust . the virial masses of halos which form population iii protostars do not evolve significantly with redshift in calculations without a soft uv background , and a rough threshold mass of @xmath326 m@xmath7 below which protostellar cores do not form is observed . other bulk halo properties , such as spin parameter and angle of separation between gas and dark matter angular momentum vectors , are consistent with previous numerical simulations of much larger cosmological halos , and do not appear to evolve with redshift . our suite of calculations suggest a clear relationship between halo collapse redshift and environment and the observed accretion rate onto the forming protostellar core , with lower accretion rates seen at higher redshift and in halos which form in significantly overdense regions of the universe . this is explained by cooler gas in the cores of these halos , and is the result of higher molecular hydrogen fractions from warmer overall gas due to the virial scaling of halo temperature with redshift or due to halo s merger history .	we perform 12 extremely high resolution adaptive mesh refinement ( amr ) cosmological hydrodynamic simulations of population iii star formation in a @xmath0cdm universe , varying the box size and large - scale structure , to understand systematic effects in the formation of primordial protostellar cores . we find results that are qualitatively similar to those of , bryan , & norman ( 2002 ) , , coppi , & larson ( 2002 ) , and ( 2006 ) . our calculations indicate that the halos out of which these stars form have similar spin parameters to galaxy - mass halos which form at @xmath1 as well as with previous simulations of high - redshift halos . we observe that , in the absence of a photodissociating ultraviolet background , the threshold halo mass for formation of a population iii protostar does not evolve significantly with time in the redshift range studied ( @xmath2 ) but exhibits substantial scatter ( @xmath3 m@xmath4 m@xmath5 ) due to different halo assembly histories : halos which assembled more slowly develop cooling cores at lower mass than those that assemble more rapidly , in agreement with ( 2003 ) . we do , however , observe significant evolution in the accretion rates of population iii protostars with redshift , with objects that form later having higher maximum accretion rates ( @xmath6 m@xmath7/yr at @xmath8 and @xmath9 m@xmath7/yr at @xmath10 ) . this can be explained by considering the evolving virial properties of the halos with redshift and the physics of molecular hydrogen formation at low densities . our result implies that the mass distribution of population iii stars ( as inferred from their accretion rates ) may be broader than previously thought , and may evolve with redshift . finally , we observe that our collapsing protostellar cloud cores do not fragment , consistent with previous results , which suggests that population iii stars which form in halos of mass @xmath11 m@xmath7 always form in isolation .
feature extraction and model selection are important for sparse high dimensional data analysis in many research areas such as genomics , genetics and machine learning . motivated by the need of robust and efficient high dimensional model selection method , we introduce a new penalized quasi - likelihood estimation for linear model with high dimensionality of parameter space . consider the estimation of the unknown parameter @xmath1 in the linear regression model @xmath2 where @xmath3 is an @xmath4-vector of response , @xmath5 is an @xmath6 matrix of independent variables with @xmath7 being its @xmath8-th row , @xmath9 is a @xmath10-vector of unknown parameters and @xmath11 is an @xmath4-vector of i.i.d . random errors with mean zero , independent of @xmath12 . when the dimension @xmath10 is high it is commonly assumed that only a small number of predictors actually contribute to the response vector @xmath13 , which leads to the sparsity pattern in the unknown parameters and thus makes variable selection crucial . in many applications such as genetic association studies and disease classifications using high - throughput data such as microarrays with gene - gene interactions , the number of variables @xmath10 can be much larger than the sample size @xmath4 . we will refer to such problem as ultrahigh - dimensional problem and model it by assuming @xmath14 for some @xmath15 . following @xcite , we will refer to @xmath10 as a non - polynomial order or np - dimensionality for short . popular approaches such as lasso @xcite , scad @xcite , adaptive lasso @xcite and elastic - net @xcite use penalized least - square regression : @xmath16 where @xmath17 is a specific penalty function . the quadratic loss is popular for its mathematical beauty but is not robust to non - normal errors and presence of outliers . robust regressions such as the least absolute deviation and quantile regressions have recently been used in variable selection techniques when @xmath10 is finite @xcite . other possible choices of robust loss functions include huber s loss @xcite , tukey s bisquare , hampel s psi , among others . each of these loss functions performs well under a certain class of error distributions : quadratic loss is suitable for normal distributions , least absolute deviation is suitable for heavy - tail distributions and is the most efficient for double exponential distributions , huber s loss performs well for contaminated normal distributions . however , none of them is universally better than all others . how to construct an adaptive loss function that is applicable to a large collection of error distributions ? we propose a simple and yet effective quasi - likelihood function , which replaces the quadratic loss by a weighted linear combination of convex loss functions : @xmath18 where @xmath19 are convex loss functions and @xmath20 are positive constants chosen to minimize the asymptotic variance of the resulting estimator . from the point of view of nonparametric statistics , the functions @xmath21 can be viewed as a set of basis functions , not necessarily orthogonal , used to approximate the unknown log - likelihood function of the error distribution . when the set of loss functions is large , the quasi - likelihood function can well approximate the log - likelihood function and therefore yield a nearly efficient method . this kind of ideas appeared already in traditional statistical inference with finite dimensionality @xcite . we will extend it to the sparse statistical inference with np - dimensionality . the quasi - likelihood function can be directly used together with any penalty function such as @xmath22-penalty with @xmath23 @xcite , lasso i.e. @xmath0-penalty @xcite , scad @xcite , hierarchical penalty @xcite , resulting in the penalized composite quasi - likelihood problem : @xmath24 instead of using folded - concave penalty functions , we use the weighted @xmath0- penalty of the form @xmath25 for some function @xmath26 and initial estimator @xmath27 , to ameliorate the bias in @xmath0-penalization @xcite and to maintain the convexity of the problem . this leads to the following convex optimization problem : @xmath28 when @xmath29 , the derivative of the penalty function , ( [ eq5 ] ) can be regarded as the local linear approximation to problem ( [ eq4 ] ) @xcite . in particular , lasso @xcite corresponds to @xmath30 , scad reduces to @xcite @xmath31 and adaptive lasso @xcite takes @xmath32 where @xmath33 . there is a rich literature in establishing the oracle property for penalized regression methods , mostly for large but fixed @xmath10 @xcite . one of the early papers on diverging @xmath10 is the work by @xcite under conditions of @xmath34 . more recent works of the similar kind include @xcite , @xcite , @xcite , which assume that the number of non - sparse elements @xmath35 is finite . when the dimensionality @xmath10 is of polynomial order , @xcite recently gave the conditions under which the scad estimator is an oracle estimator . we would like to further address this problem when @xmath14 with @xmath15 and @xmath36 for @xmath37 , that is when the dimensionality is of exponential order . the paper is organized as follows . section [ sec2 ] introduces an easy to implement two - step computation procedure . section [ sec3 ] proves the strong oracle property of the weighted @xmath0-penalized quasi - likelihood approach with discussion on the choice of weights and corrections for convexity . section [ sec4 ] defines two specific instances of the proposed approach and compares their asymptotic efficiencies . section [ sec5 ] provides a comprehensive simulation study as well as a real data example of the snp selection for the down syndrome . section [ sec6 ] is devoted to the discussion . to facilitate the readability , all the proofs are relegated to the appendices a , b & c. we would like to describe the proposed two - step adaptive computation procedure and defer the justification of the appropriate choice of the weight vector @xmath38 to section [ sec3 ] . in the first step , one will get the initial estimate @xmath39 using the lasso procedure , i.e : @xmath40 and estimate the residual vector @xmath41 ( for justification see discussion following condition [ cond2 ] ) . the matrix @xmath42 and vector @xmath43 are calculated as follows : @xmath44 where @xmath45 is a choice of the subgradient of @xmath46 , @xmath47 is the @xmath8-th component of @xmath48 , and @xmath49 should be considered as a consistent estimator of @xmath50 , which is the derivative of @xmath51 at @xmath52 . for example , when @xmath53 , then @xmath54 and @xmath55 . the optimal weight is then determined as @xmath56 in the second step , one calculates the quasi maximum likelihood estimator ( qmle ) using weights @xmath57 as @xmath58 * remark 1 * : note that zero is not an absorbing state in the minimization problem . those elements that are estimated as zero in the initial estimate @xmath27 have a chance to escape from zero , whereas those nonvanishing elements can be estimated as zero in . * remark 2 * : the number of loss functions @xmath59 is typically small or moderate in practice . problem can be easily solved using a quadratic programming algorithm . the resulting vector @xmath57 can have vanishing components , automatically eliminating inefficient loss functions in the second step ( [ eq8 ] ) and hence learning the best approximation of the unknown log - likelihood function . this can lead to considerable computational gains . see section [ sec4 ] for additional details . * remark 3 * : problem ( [ eq8 ] ) is a convex optimization problem when @xmath60 s are all convex and latexmath:[$\gamma_\lambda ( \|\hat{\beta}^{(0)}_j and efficient computational algorithms such as pathwise coordinate optimization @xcite and least angle regression @xcite . one particular example is the combination of @xmath0 and @xmath62 regressions , in which @xmath63 , @xmath64 and @xmath65 . here @xmath66 denotes the median of error distribution @xmath67 . if the error distribution is symmetric , then @xmath68 . if the error distribution is completely unknown , @xmath66 is unknown and can be estimated from the residual vector @xmath69 or being regarded as an additional parameter and optimized together with @xmath1 in ( [ eq8 ] ) . another example is the combination of multiple quantile check functions , that is , @xmath70 where @xmath71 is a preselected quantile and @xmath72 is the @xmath73-quantile of the error distribution . again , when @xmath72 s are unknown , they can be estimated using the sample quantiles @xmath73 of the estimated residuals @xmath48 or along with @xmath1 in ( [ eq8 ] ) . see section [ sec4 ] for additional discussion . in this section , we plan to establish the sampling properties of estimator ( [ eq5 ] ) under the assumption that the number of parameters ( true dimensionality ) @xmath10 and the number of non - vanishing components ( effective dimensionality ) @xmath74 satisfy @xmath14 and @xmath75 for some @xmath15 and @xmath76 . particular focus will be given to the oracle property of @xcite , but we will strengthen it and prove that estimator ( [ eq5 ] ) is an oracle estimator with overwhelming probability . @xcite were among the first to discuss the oracle properties with np dimensionality using the full likelihood function in generalized linear models with a class of folded concave penalties . we work on a quasi - likelihood function and a class of weighted convex penalties . to facilitate presentation , we relegate technical conditions and the details of proofs to the appendix . we consider more generally the weighted @xmath0-penalized estimator with nonnegative weights @xmath77 . let @xmath78 denote the penalized quasi - likelihood function . the estimator in ( [ eq5 ] ) is a particular case of and corresponds to the case with @xmath79 . without loss of generality , assume that parameter @xmath80 can be arranged in the form of @xmath81 , with @xmath82 a vector of non - vanishing elements of @xmath80 . let us call @xmath83 the biased oracle estimator , where @xmath84 is the minimizer of @xmath85 in @xmath86 and @xmath87 is the vector of all zeros in @xmath88 . here , we suppress the dependence of @xmath89 on @xmath90 and @xmath91 . the estimator @xmath92 is called the biased oracle estimator , since the oracle knows the true submodel @xmath93 , but nevertheless applies a penalized method to estimate the non - vanishing regression coefficients . the bias becomes negligible when the weights in the first part are zero or uniformly small ( see theorem [ thm2 ] ) . when the design matrix @xmath94 is non - degenerate , the function @xmath95 is strictly convex and the biased oracle estimator is unique , where @xmath94 is a submatrix of @xmath12 such that @xmath96 $ ] with @xmath94 and @xmath97 being @xmath98 and @xmath99 sub - matrices of @xmath12 , respectively . the following theorem shows that @xmath89 is the unique minimizer of @xmath100 on the whole space @xmath101 with an overwhelming probability . as a consequence , @xmath102 becomes the biased oracle . we establish the following theorem under conditions on the non - stochastic vector @xmath103 ( see condition [ cond2 ] ) . it is also applicable to stochastic penalty weights as in ; see the remark following condition [ cond2 ] . [ thm1 ] under conditions [ cond1]-[cond4 ] , the estimators @xmath104 and @xmath105 exist and are unique on a set with probability tending to one . furthermore , @xmath106 for a positive constant @xmath107 . for the previous theorem to be nontrivial , we need to impose the dimensionality restriction @xmath108 , where @xmath109 controls the rate of growth of the correlation coefficients between the matrices @xmath94 and @xmath97 , the important predictors and unimportant predictors ( see condition [ cond5 ] ) and @xmath110 is a non - negative constant , related to the maximum absolute value of the design matrix [ see condition [ cond4 ] ] . it can be taken as zero and is introduced to deal with the situation where @xmath111 is small or zero so that the result is trivial . the larger @xmath112 is , the more stringent restriction is imposed on the choice of @xmath113 . when the above conditions hold , the penalized composite quasi - likelihood estimator @xmath105 is equal to the biased oracle estimator @xmath89 , with probability tending to one exponentially fast . * remark 4 : * the result of theorem [ thm1 ] is stronger than the oracle property defined in @xcite once the properties of @xmath104 are established ( see theorem [ thm2 ] ) . it was formulated by @xcite for the scad estimator with polynomial dimensionality @xmath10 . it implies not only the model selection consistency and but also sign consistency @xcite : @xmath114 in this way , the result of theorem [ thm1 ] nicely unifies the two approaches in discussing the oracle property in high dimensional spaces . let @xmath115 and @xmath116 be the first @xmath35 components and the remaining @xmath117 components of @xmath118 , respectively . according to theorem [ thm1 ] , we have @xmath119 with probability tending to one . hence , we only need to establish the properties of @xmath120 . [ thm2 ] under conditions [ cond1]-[cond5 ] , the asymptotic bias of non - vanishing component @xmath120 is controlled by @xmath121 with @xmath122 furthermore , when @xmath123 , @xmath120 possesses asymptotic normality : @xmath124 where @xmath125 is a unit vector in @xmath126 and @xmath127}{\left(\sum_{k=1}^kw_ke[\partial \psi_k(\varepsilon)]\right)^2}.\ ] ] since the dimensionality @xmath35 depends on @xmath4 , the asymptotic normality of @xmath128 is not well defined in the conventional probability sense . the arbitrary linear combination @xmath129 is used to overcome the technical difficulty . in particular , any finite component of @xmath128 is asymptotically normal . the result in theorem [ thm2 ] is also equivalent to the asymptotic normality of the linear combination @xmath130 stated in @xcite , where @xmath131 is a @xmath132 matrix , for any given finite number @xmath133 . this theorem relates to the results of @xcite in classical setting ( corresponding to @xmath134 ) where he established asymptotic normality of @xmath135-estimators when the dimensionality is not higher than @xmath136 . the asymptotic normality ( [ eq10 ] ) allows us to do statistical inference for non - vanishing components . this requires an estimate of the asymptotic covariance matrix of @xmath115 . let @xmath137 be the residual and @xmath138 be its @xmath8-th component . a simple substitution estimator of @xmath139 is @xmath140 see also the remark proceeding . consequently , by , the asymptotic variance - covariance matrix of @xmath120 is given by @xmath141 another possible estimator of the variance and covariance matrix is to apply the standard sandwich formula . in section [ sec5 ] , through simulation studies , we show that this formula has good properties for both @xmath10 smaller and larger than @xmath4 ( see tables 3 and 4 and comments at the end of section [ sec5.1 ] ) . note that only the factor @xmath142 in equation depends on the choice of @xmath38 and it is invariant to the scaling of @xmath38 . thus , the optimal choice of weights for maximizing efficiency of the estimator @xmath115 is @xmath143 where @xmath144 and @xmath43 are defined in section [ sec2 ] using an initial estimator , independent of the weighting scheme @xmath90 . * remark 5 : * the quadratic optimization problem does not have a closed form solution , but can easily be solved numerically for a moderate @xmath59 . the above efficiency gain , over the least - squares , could be better understood from the likelihood point of view . let @xmath145 denote the unknown error density . the most efficient loss function is the unknown log - likelihood function , @xmath146 . but since we have no knowledge of it , the set @xmath147 , consisting of convex combinations of @xmath148 given in , could be viewed as a collection of basis functions used to approximate it . the broader the set @xmath147 is , the better it can approximate the log - likelihood function and the more efficient the estimator @xmath149 in becomes . therefore , we refer to @xmath150 as the quasi - likelihood function . the restriction of @xmath151 guarantees the convexity of @xmath150 so that the problem becomes a convex optimization problem . however , this restriction may cause substantial loss of efficiency in estimating @xmath152 ( see table 1 ) . we propose a one - step penalized estimator to overcome this drawback while avoiding non - convex optimization . let @xmath153 be the estimator based on the convex combination of loss functions and @xmath154 be its nonvanishing components . the one - step estimator is defined as @xmath155^{-1 } \phi_{n,{\mathbf{w}}}(\hat{\mbox{\boldmath $ \beta$}}_1 ) , \ \hat{{\mbox{\boldmath $ \beta$}}}^{\footnotesize{\mbox{os}}}_{\mathbf{w}2 } = \mathbf{0},\ ] ] where @xmath156 @xmath157 [ thm3 ] under conditions [ cond1]-[cond5 ] , if @xmath158 , then the one - step estimator @xmath159 enjoys the asymptotic normality : @xmath160 provided that @xmath161 , @xmath162 is lipchitz continous , and @xmath163 , where @xmath164 denote the maximum eigenvalue of a matrix and @xmath165 are defined as in theorem [ thm2 ] . the one - step estimator overcomes the convexity restriction and is always well defined , whereas is not uniquely defined when convexity of @xmath150 is ruined . note that if we remove the constraint of @xmath166 ( @xmath167 ) , the optimal weight vector in is equal to @xmath168 this can be significantly smaller than the optimal variance obtained with convexity constraint , especially for multi - modal distributions ( see table 1 ) . the above discussion prompts a further improvement of the penalized adaptive composite quasi - likelihood in section [ sec2 ] . use to compute the new residuals and new matrix @xmath169 and vector @xmath43 . compute the optimal unconstrained weight @xmath170 and the one - step estimator . in this section , we discuss two specific examples of penalized quasi - likelihood regression . the proposed methods are complementary , in the sense that the first one is computationally easy but loses some general flexibility while the second one is computationally intensive but efficient in a broader class of error distributions . first , we consider the combination of @xmath0 and @xmath62 loss functions , that is , @xmath171 and @xmath172 . the nuisance parameter @xmath66 is the median of the error distribution . let @xmath173 denote the corresponding penalized estimator as the solution to the minimization problem : @xmath174 if the error distribution is symmetric , then @xmath175 and the minimization problem ( [ eq17 ] ) can be recast as a penalized weighted least square regression @xmath176 which can be efficiently solved by pathwise coordinate optimization @xcite or least angle regression @xcite . if @xmath177 , the penalized least - squares problem ( [ eq17 ] ) is somewhat different from ( [ eq5 ] ) since we have an additional parameter @xmath66 . using the same arguments , and treating @xmath66 as an additional parameter for which we solve in , we can show that the conclusions of theorems [ thm2 ] and [ thm3 ] hold with the asymptotic variance equal to @xmath178 where @xmath179 $ ] and @xmath180 is the density of @xmath67 . this will hold when @xmath66 is either known or unknown . explicit optimization of is not trivial and we go through it as follows . since @xmath181 is invariant to the scale of @xmath38 , by setting @xmath182 , we have @xmath183 where @xmath184 and @xmath185 . note that @xmath186 \leq \sigma.\ ] ] hence , @xmath187 and @xmath188 the optimal value of @xmath107 over @xmath189 can be easily computed . if @xmath190 , then the optimal value is obtained at @xmath191 in particular , when @xmath192 , @xmath193 , and the optimal choice is the least - squares estimator . when @xmath194 , if @xmath192 , then the minimizer is @xmath193 . in all other cases , the minimizer is @xmath195 i.e. we are left to use @xmath0 regression alone . the above result shows the limitation of the convex combination , i.e. @xmath196 . in many cases , we are left alone with the least - squares or least absolute deviation regression without improving efficiency . the efficiency can be gained and achieved by allowing negative weights via the one - step technique as in section [ sec3.4 ] . let @xmath197 . the function @xmath198 has a pole at @xmath199 and a unique critical point @xmath200 provided that @xmath201 . consequently , the function @xmath198 can not have any local maximizer ( otherwise , from the local maximizer to the point @xmath199 , there must exist a local minimizer , which is also a critical point ) . hence , the minimum value is attained at @xmath202 . in other words , @xmath203 where @xmath204 since the denominator can be written as @xmath205 , we have @xmath206 , namely , it outperforms the least - squares estimator , unless @xmath207 . similarly , it can be shown that @xmath208 } \leq \frac{1}{4b_\varepsilon^2},\ ] ] namely , it outperforms the least absolute deviation estimation , unless @xmath209 . when error distribution is symmetric unimodal , @xmath210 , according to chapter 5 of @xcite . the worst scenario for the @xmath0-regression in comparison with the @xmath62-regression is the uniform distribution ( see chapter 5 , @xcite ) , which has the relatively efficiency of merely @xmath211 . for such uniform distribution , @xmath212 and @xmath213 , @xmath214 , and @xmath215 . hence , the best @xmath0-@xmath62 is 4 times better than @xmath0 regression alone . more comparisons about the weighted @xmath0-@xmath62 combination with @xmath0 and least - squares are given in table 1(section [ sec4.3 ] ) . the weighted composite quantile regression ( cqr ) was first studied by @xcite in classical statistical inference setting . @xcite used equally weighted cqr ( ecqr ) for penalized model selection with @xmath10 large but fixed . we show that the efficiency of ecqr can be substantially improved by properly weighting and extend the work to the case of @xmath216 . consider @xmath59 different quantiles , @xmath217 . let @xmath218 . the penalized composite quantile regression estimator @xmath219 is defined as the solution to the minimization problem @xmath220 where @xmath221 is the estimator of the nuisance parameter @xmath222 , the @xmath73-th quantile of the error distribution . note that @xmath223 are nuisance parameters and the minimization at is done with respect to them too . after some algebra we can confirm that the conclusions of theorems [ thm2 ] and [ thm3 ] continue to hold with the asymptotic variance as @xmath224 as shown in @xcite and @xcite , when @xmath225 , the optimally weighted cqr ( wcqr ) is as efficient as the maximum likelihood estimator , always more efficient than ecqr . computationally , the minimization problem in equation ( [ eq24 ] ) can be casted as a large scale linear programming problem by expanding the covariate space with new ancillary variables . thus , it is computationally intensive to use too many quantiles . in section [ sec4.3 ] , we can see that usually no more than ten quantiles are adequate for wcqr to approach the efficiency of mle , whereas determining the optimal value of @xmath59 in ecqr seems difficult since the efficiency is not necessarily an increasing function of @xmath59 ( table 2 ) . also , some of the weights in @xmath226 are zero , hence making wcqr method computationally less intensive than ecqr . from our experience in large @xmath10 and small @xmath4 situations , this reduction tends to be significant . the optimal convex combination of quantile regression uses the weight @xmath227 where @xmath228 and @xmath169 is a @xmath229 matrix whose @xmath230-element is @xmath231 . the optimal combination of quantile regression , which is obtained by using the one - step procedure , uses the weight @xmath232 clearly , both combinations improve the efficiency of ecqr and the optimal combination is most efficient among the three ( see table 1 ) . when the error distributions are skewed or multimodal , the improvement can be substantial . in this section , we studied the asymptotic efficiency of proposed estimators under several error distributions . for comparison , we also included @xmath0 regression , @xmath62 regression and ecqr . the error distribution ranges from the symmetric to asymmetric distributions : double exponential ( de ) , @xmath233 distribution with degree of freedoms 4 ( @xmath234 ) , normal @xmath235 , gamma @xmath236 , beta @xmath237 , a scale mixture of normals ( @xmath238 ) @xmath239 and a location mixture of normals ( @xmath240 ) @xmath241 . to keep the comparison fair and to satisfy the first assumption of mean zero error terms , we first centered the error distribution to have mean zero . table 1 shows the asymptotic relative efficiency of each estimator compared to mle . @xmath0-@xmath242 and @xmath0-@xmath62 indicate the optimal convex @xmath0-@xmath62 combination and optimal @xmath0-@xmath62 combination , respectively . while @xmath0 regression can have higher or lower efficiency than @xmath62 regression in different error distributions , @xmath0-@xmath242 and @xmath0-@xmath62 regressions are consistently more efficient than both of them . wcqr@xmath243 denote the optimal convex combination of multiple quantile regressions and wcqr represent the optimal combination . in all quantile regressions , quantiles @xmath244 were used . as shown in table 1 , wcqr@xmath243 and wcqr always outperform ecqr and the differences are more significant in double exponential distribution and asymmetric distributions such as gamma and beta . in de , @xmath234 and @xmath235 , nine quantiles are usually adequate for wcqr@xmath243 and wcqr to achieve full efficiency . in @xmath245 and @xmath237 , they need 29 quantiles to achieve efficiency close to mle while the other estimators are significantly inefficient . this difference is most expressed in multimodal distributions , mn@xmath246 and mn@xmath247 , with wcqr outperforming all . one of the possible problems with ecqr is that the efficiency does not necessarily increase with @xmath59 , making the choice of @xmath59 harder . for example , for the double exponential distribution , the relative efficiency decreases with @xmath59 . this is understandable , as @xmath248 is optimal : putting more and odd number of quantiles dilutes the weights . .asymptotic relative efficiency compared to mle [ cols="<,>,^,^,^,^,^,^,^",options="header " , ] the eqtls are almost all located within 500 kb upstream tss or 500 kb downstream tes ( figure 2 ) and mostly from 100 kb upstream tss to 350 kb downstream tes . [ fig : loci ] in this paper , a robust and efficient penalized quasi - likelihood approach is introduced for model selection with np - dimensionality . it is shown that such an adaptive learning technique has a strong oracle property . as specific examples , two complementary methods of penalized composite @xmath0-@xmath62 regression and weighted composite quantile regression are introduced and they are shown to possess good efficiency and model selection consistency in ultrahigh dimensional space . numerical studies show that our method is adaptive to unknown error distributions and outperforms lasso @xcite and equally weighted composite quantile regression @xcite . the penalized composite quasi - likelihood method can also be used in sure independence screening @xcite or iterated version @xcite , resulting in a robust variable screening and selection . in this case , the marginal regression coefficients or contributions will be ranked and thresholded @xcite . it can also be applied to the aggregation problems of classification @xcite where the usual @xmath62 risk function could be replaced with composite quasi - likelihood function . the idea can also be used to choose the loss functions in machine learning . for example , one can adaptively combine the hinge - loss function in the support vector machine , the exponential loss in the adaboost , and the logistic loss function in logistic regression to yield a more efficient classifier . let @xmath249 be the set of discontinuity points of @xmath45 , which is a subgradient of @xmath60 . assume that the distribution of error terms @xmath250 is smooth enough so that @xmath251 . additional regularity conditions on @xmath252 are needed , as in @xcite . [ cond1 ] the function @xmath252 satisfies @xmath253=a_kc+o(\|c\| ) \mbox { as } \|c\| \to 0 $ ] , for some @xmath254 . for sufficiently small @xmath255 , @xmath256 $ ] exists and is continuous at @xmath257 , where @xmath258 . the error distribution satisfies the following cramr condition : @xmath259 , for some constants @xmath260 and @xmath59 . this condition implies that @xmath261 , which is an unbiased score function of parameter @xmath1 . it also implies that @xmath262 exists . the following two conditions are important for establishing sparsity properties of parameter @xmath105 by controlling the penalty weighting scheme @xmath103 and the regularization parameter @xmath113 . [ cond2 ] assume that @xmath263 and @xmath264 . in addition , @xmath265 the first statement is to ensure that the bias term in theorem [ thm2 ] is negligible . it is needed to control the bias due to the convex penalty . the second requirement is to make sure that the weights @xmath103 in the second part are uniformly large so that the vanishing coefficients are estimated as zero . it can also be regarded as a normalization condition , since the actual weights in the penalty are @xmath266 . the lasso estimator will not satisfy the first requirement of condition [ cond2 ] unless @xmath113 is small and @xmath267 . nevertheless , under the sparse representation condition @xcite , @xcite show that with probability tending to one , the lasso estimator is model selection consistent with @xmath268 , when the minimum signal @xmath269 . they also show that the same result holds for the scad - type estimators under weaker conditions . using one of them as the initial estimator , the weight @xmath270 in ( [ eq8 ] ) would satisfy condition [ cond2 ] , on a set with probability tending to one . this is due to the fact that with @xmath271 given by ( [ eq6 ] ) , for @xmath272 , @xmath273 , whereas for @xmath274 , @xmath275 , as long as @xmath276 . in other words , the results of theorems [ thm1 ] and [ thm2 ] are applicable to the penalized estimator ( [ eq8 ] ) with data driven weights . [ cond3 ] the regularization parameter @xmath277 , where parameter @xmath109 is defined in condition [ cond5 ] and @xmath110 is a constant , bounded by the restriction in condition [ cond4 ] . we use the following notation throughout the proof . let @xmath131 be a matrix . denote by @xmath278 and @xmath279 the minimum and maximum eigenvalue of the matrix @xmath131 when it is a square symmetric matrix . let @xmath280 be the operator norm and @xmath281 the largest absolute value of the elements in @xmath131 . as a result , @xmath282 is the euclidean norm when applied to a vector . define @xmath283 . [ cond4 ] the matrix @xmath284 satisfies @xmath285 for some positive constants @xmath286 . there exists @xmath287 such that @xmath288 where @xmath289 is the @xmath8-th row of @xmath94 . furthermore , assume that the design matrix satisfies @xmath290 and @xmath291 , where @xmath292 is the @xmath293-th column of @xmath12 . [ cond5 ] assume that @xmath294 @xmath295 where @xmath296 is an @xmath35-dimensional ball centered at @xmath297 with radius @xmath298 and @xmath299 is the diagonal matrix with @xmath8-th element equal to @xmath300 . recall that @xmath301 and @xmath302 is the true model . [ lem6.1 ] under conditions [ cond2 ] and [ cond4 ] , the penalized quasi - likelihood @xmath100 defined by has a unique global minimizer @xmath303 , if @xmath304 @xmath305 where @xmath306 , @xmath307 and @xmath308 stand for the subvectors of @xmath103 , consisting of its first @xmath35 elements and the last @xmath117 elements respectively , and @xmath309 and @xmath310 ( the hadamard product ) in are taken coordinatewise . conversely , if @xmath311 is a global minimizer of @xmath100 , then holds and holds with strict inequality replaced with non - strict one . * proof of lemma [ lem6.1 ] : * under conditions [ cond2 ] and [ cond4 ] , @xmath100 is strictly convex . necessary conditions and are direct consequences of the karush - kuhn - tucker conditions of optimality . the sufficient condition follows from similar arguments as those in the proof of theorem 1 in @xcite and the strict convexity of the function @xmath312 . [ lem6.2 ] under conditions [ cond1]-[cond5 ] we have that @xmath313 where @xmath314 is the subvector of @xmath103 , consisting of its first @xmath35 elements . * proof of lemma [ lem6.2 ] : * since @xmath315 , we only need to consider the sub - vector of the first @xmath35 components . let us first show the existence of the biased oracle estimator . we can restrict our attention to the @xmath35-dimensional subspace @xmath316 . our aim is to show that @xmath317 for sufficiently large @xmath318 . here , there is a minimizer inside the ball @xmath319 , with probability tending to one . using the strict convexity of @xmath100 , this minimizer is the unique global minimizer . by the taylor expansion at @xmath320 , we have @xmath321 where @xmath322 where @xmath323 $ ] . by the cauchy - schwarz inequality , @xmath324 note that for all @xmath325 , we have @xmath326 and @xmath327 which is of order @xmath328 by condition [ cond4 ] . hence , @xmath329 uniformly in @xmath330 . finally , we deal with @xmath331 . let @xmath332 . by lemma 3.1 of @xcite , we have @xmath333 for a positive constant @xmath107 . combining all of the above results , we have with probability tending to one that @xmath334 where the right hand side is larger than 0 when @xmath335 for a sufficiently large @xmath336 . since the objective function is strictly convex , there exists a unique minimizer @xmath337 such that @xmath338 [ lem6.3 ] under the conditions of theorem [ thm2 ] , @xmath339^{-1/2}\sum_{i=1}^n \psi_{\mathbf{w}}(\varepsilon_i ) { \mbox{\bf b}}^t{\mbox{\bf s}}_i \stackrel{\mathcal{d}}{\to } \mathcal{n}(0,1)\ ] ] where @xmath340 . * proof of lemma [ lem6.3 ] : * by condition [ cond1 ] , since @xmath341 is independent of @xmath342 , we have @xmath343 , and @xmath344^{-1/2 } \sum_{i=1}^n \psi_{\mathbf{w}}(\varepsilon_i ) \mathbf{b}^t { \mbox{\bf s}}_i \right ] = 1.\end{aligned}\ ] ] to complete proof of the lemma , we only need to check the lyapounov condition . by condition [ cond1 ] , @xmath345 . furthermore , condition [ cond4 ] implies @xmath346 for a positive constant @xmath347 . using these together with the cauchy - schwartz inequality , we have @xmath348^{-1/2 } \psi_{\mathbf{w}}(\varepsilon_i ) \mathbf{b}^t { \mbox{\bf s}}_i \right\|^{2 + \xi}\\ & = & o(1 ) \sum_{i=1}^n \left\| n^{-1/2 } { \mbox{\bf b}}^t { \mbox{\bf s}}_i \right\|^{2 + \xi}. \\ & = & o(1 ) \sum_{i=1}^n \left\| n^{-1/2 } \\|{\mbox{\bf s}}_i\\| \right\|^{2 + \xi},\end{aligned}\ ] ] which tends to zero by condition [ cond4 ] . this completes the proof . the following bernstein s inequality can be found in lemma 2.2.11 of + @xcite . [ lem6.4 ] let @xmath349 be independent random variables with zero mean such that @xmath350 , for every @xmath351 ( and all @xmath8 ) and some constants @xmath135 and @xmath352 . then @xmath353 for @xmath354 . then the following inequality is a consequence of previous bernstein s inequality . let @xmath355 satisfy the condition of lemma [ lem6.4 ] with @xmath356 . for a given sequence @xmath357 , @xmath358 . a direct application of lemma [ lem6.4 ] yields @xmath359 * proof of theorem [ thm1 ] : * we only need to show that @xmath104 is the unique minimizer of @xmath312 in @xmath360 on a set @xmath361 which has a probability tending to one . since @xmath362 already satisfies , we only need to check . we now define the set @xmath361 . let @xmath363 and consider the event @xmath364 with @xmath365 being chosen later . then , by condition [ cond1 ] and bernstein s inequality , it follows directly from that @xmath366 where @xmath367 is the @xmath293-th column of @xmath12 . taking @xmath368 , we have @xmath369 for some positive constant @xmath370 , by condition [ cond4 ] . thus , by using the union bound , we conclude that @xmath371 we now check whether holds on the set @xmath361 . let @xmath372 be the @xmath4-dimensional vector with the @xmath8-th element @xmath373 . then , by condition [ cond2 ] @xmath374 \right \\|_\infty \nonumber \\ & = & o\left ( n^{1/2 } u_n + \left\\| { \mbox{\bf q}}^t \mbox{diag } ( \partial { \mbox{\boldmath $ \psi$}}_{\mathbf{w}}({\mbox{\bf v } } ) ) { \mbox{\bf s}}(\hat { \mbox{\boldmath $ \beta$}}_{1}^o - { \mbox{\boldmath $ \beta$}}_1^ * ) \right\\|_{\infty } \right ) \end{aligned}\ ] ] where @xmath375 lies between @xmath104 and @xmath376 . by condition [ cond5 ] , the second term in is bounded by @xmath377 where the equality follows from lemma [ lem6.2 ] . by the choice of parameters , @xmath378 by taking @xmath379 . hence , by lemma [ lem6.1 ] , @xmath104 is the unique global minimizer . by using , we have @xmath387 or equivalently , @xmath388 note that @xmath389 . we have for any vector @xmath330 , @xmath390 consequently , for any unit vector @xmath125 , @xmath391 by using conditions [ cond4 ] and [ cond5 ] . this shows that the second term in ( [ eqc4 ] ) , when multiplied by the vector @xmath392 is of order @xmath393 by condition [ cond2 ] . therefore , we need to establish the asymptotic normality of the first term in . this term is identical to the situation dealt by @xcite . using his result , the second conclusion of theorem [ thm2 ] follows . this completes the proof . * proof of theorem [ thm3 ] : * first of all , by taylor expansion , @xmath394 where @xmath395 lies between @xmath376 and @xmath154 . consequently , @xmath396 by the definition of the one step estimator and , we have @xmath397 where @xmath398 we first deal with the remainder term . note that @xmath399 and @xmath400 where @xmath401 . by the liptchiz continuity , we have @xmath402 where @xmath403 is the liptchiz coefficient of @xmath404 . let @xmath405 be the identity matrix of order @xmath35 and @xmath406 . by ( [ eqc8 ] ) , we have @xmath407 hence , all of the eigenvalues of the matrix is no larger than @xmath408 . similarly , by ( [ eqc8 ] ) , @xmath409 and all of its eigenvalue should be at least @xmath410 . consequently , @xmath411 by condition [ cond5 ] and the assumption of @xmath154 , it follows from ( [ eqc7 ] ) that @xmath412 thus , for any unit vector @xmath125 , @xmath413 the main term in can be handled by using lemma [ lem6.3 ] and the same method as @xcite . this completes the proof .	in high - dimensional model selection problems , penalized least - square approaches have been extensively used . this paper addresses the question of both robustness and efficiency of penalized model selection methods , and proposes a data - driven weighted linear combination of convex loss functions , together with weighted @xmath0-penalty . it is completely data - adaptive and does not require prior knowledge of the error distribution . the weighted @xmath0-penalty is used both to ensure the convexity of the penalty term and to ameliorate the bias caused by the @xmath0-penalty . in the setting with dimensionality much larger than the sample size , we establish a strong oracle property of the proposed method that possesses both the model selection consistency and estimation efficiency for the true non - zero coefficients . as specific examples , we introduce a robust method of composite l1-l2 , and optimal composite quantile method and evaluate their performance in both simulated and real data examples . _ key words _ : composite qmle , lasso , model selection , np dimensionality , oracle property , robust statistics , scad
the key property of active galactic nuclei ( agn ) is that they emit enormous amount of electromagnetic radiation over a very broad energy range . the agn central location in their host galaxies imply that agn radiation can play a very important role in determining the ionization structure and dynamics of matter not only near the agn but also on larger , galactic and even intergalactic scales ( ciotti & ostriker , 1997 , 2001 , 2007 ; silk & rees 1998 ; king 2003 ; murray et al . 2005 ; sazonov et al . 2005 ; springel et at . 2005 ; hopkins et al . 2005 ; wang et al . 2006 ; fabian et al . 2006 ; thacker et al . 2006 , and references therein ) . in the first paper of this series ( proga 2007 , paper i hereafter ) , we reported on results from our first phase of gas dynamics studies in agn on sub- and parsec - scales . this is a complex problem as it involves many aspects of physics such as multi - dimensional fluid dynamics , radiative and magnetic processes . therefore , our approach was to set up simulations as simple as possible and to start with exploring the effects of the x - ray heating [ important in the so - called preheated accretion , ( e.g. , ostriker et al . 1976 ; park & ostriker 2001 , 2007 ) ] and radiation pressure on gas that is gravitationally captured by a black hole ( bh ) . we adopted the numerical methods developed by proga et al . ( 2000 , psk00 hereafter ) for studying radiation driven disk winds in agn . generally , our simulations cover a relatively unexplored range of the distance from the central bh : we end where models of galaxies begin ( e.g. , ciotti & ostriker 2007 ; springel et at . 2005 ) and we begin where models of bh accretion end ( e.g. , hawley & balbus 2002 ; ohsuga 2007 ) . in paper i , we presented results from axisymmetric time - dependent hydrodynamical ( hd ) calculations of gas flows . the flows were non - rotating and exposed to quasar radiation . we took into account x - ray heating and the radiation force due to electron scattering and spectral lines . to compute the radiation field , we considered an optically thick , geometrically thin , standard accretion disk as a source of uv photons and a spherical central object as a source of x - rays ( a corona ) . the gas temperature , @xmath0 and ionization state in the flow were calculated self - consistently from the photoionization and heating rate of the central object . we found that , for a @xmath1 black hole with an accretion luminosity of 0.6 of the eddington luminosity , the flow settles into a steady state and has two components : ( 1 ) an equatorial inflow and ( 2 ) a bipolar inflow / outflow with the outflow leaving the system along the disk rotational axis . the inflow is a realization of a bondi - like accretion flow . the second component is an example of a non - radial accretion flow which becomes an outflow once it is pushed close to the rotational axis where the radiation pressure accelerate it outward . in some cases the outflow is heated by radiation so that it can be accelerated also by thermal expansion . our main result was that the existence of the above two flow components is robust to the outer boundary conditions and the geometry and spectral energy distribution of the radiation field . however , the flow properties are not robust . in particular , the outflow power and collimation is higher for the radiation dominated by the uv / disk emission than for the radiation dominated by the x - ray / central engine emission . our most intriguing result was that a very narrow outflow driven by radiation pressure on lines can carry more energy and mass than a broad outflow driven by thermal expansion . here , we report on results from simulations that are basically reruns of those presented in paper i , but with inclusion of gas rotation . our goal is to assess how rotation changes the flow solution . in particular , we study how rotation changes the flow pattern , mass and energy fluxes , and temporal behavior . we also present results from a new set of simulations that illustrate how complex and dynamic the flow evolution can be even for relatively simple initial and boundary conditions . we describe our calculations in section 2 . we present our results in section 3 . the paper ends , in section 4 , with discussion and our conclusions . in this paper we extend the work presented in paper i by relaxing some assumptions and simplifications . our numerical hd calculations are in most respects as described by paper i. here we only describe the key elements of the calculations and list the changes we made . we refer a reader to paper i and psk00 for details . we consider an axisymmetric hd flow accreting onto a supermassive bh . the flow is non - spherical because it is irradiated by an accretion disk . the disk radiation flux , @xmath2 is highest along the disk rotational axis and is gradually decreasing with increasing polar angle , @xmath3 : @xmath4 . the flow is also irradiated by a corona . we account for some effects of photoionization . in particular , we calculate the gas temperature assuming that the gas is optically thin to its own cooling radiation . we include the following radiative processes : compton heating / cooling , x - ray photoionization heating , and recombination , bremsstrahlung and line cooling . we also take into account some effects of photoionization on radiation pressure due to lines ( line force ) . namely , we compute the line force using a value of the photoionization parameter , @xmath5 and the analytical formulae due to stevens & kallman ( 1991 ) . this procedure is computationally efficient and gives good estimates for the number and opacity distribution of spectral lines for a given @xmath5 without detail information about the ionization state ( see stevens & kallman 1991 ) . additionally , we take into account the attenuation of the x - ray radiation by computing the x - ray optical depth in the radial direction . to be consistent with our gas heating rates where we include x - ray photoionization but not uv photoionization , we do not account for attenuation of the uv radiation . we assume that the total accretion luminosity , @xmath6 has two components : @xmath7 due to the accretion disk and @xmath8 due to the corona . for simplicity , we assume that the disk emits only uv photons , whereas the corona emits only x - rays i.e. , the system uv luminosity , @xmath9 and the system x - ray luminosity , @xmath10 ( in other words @xmath11 and @xmath12 ) . with the above simplifications , only the corona radiation is responsible for ionizing the flow to a very high ionization state . in our calculations , the corona contributes to the radiation force due to electron scattering but does not contribute to line driving we note that metal lines in the soft x - ray band may have an appreciable contribution to the total radiation force in some cases . on the other hand , the disk radiation contributes to the radiation force due to both electron and line scattering . we perform our calculations in spherical polar coordinates @xmath13 assuming axial symmetry about the rotational axis of the accretion disk ( @xmath14 ) . our computational domain is defined to occupy the angular range @xmath15 and the radial range @xmath16 , where @xmath17 is the inner radius of the disk around a schwarzschild bh with a mass , @xmath18 and radius @xmath19 . the @xmath20 domain is discretized into zones . our numerical resolution in the @xmath21 direction consists of 140 zones . we fix the zone size ratio , @xmath22 ( i.e. , the zone spacing is increasing outward ) . gridding in this manner ensures good spatial resolution close to the inner boundary , @xmath23 . in the @xmath3 direction , our numerical resolution consists of 50 zones . the zone size ratio is always @xmath24 ( i.e. , grid points are equally spaced ) . for the initial condition , in paper i , we assumed spherical symmetry and set all hd variables to constant values everywhere in the computational domain . here we will allow the gas to rotate . in simulations with rotation , we break spherical symmetry of the initial and boundary conditions by introducing a small , latitude - dependent angular momentum . namely , for large radii we assume the specific angular momentum , @xmath25 , depends on the polar angle , @xmath3 , as @xmath26 where @xmath27 on the equator ( @xmath28 ) and monotonically decreases to zero at the poles ( at @xmath29 and @xmath30 ) . the initial distribution of the rotational velocity is : @xmath31 we express the specific angular momentum on the equator as @xmath32 where @xmath33 is the `` circularization radius '' on the equator in units of @xmath34 for the newtonian potential ( i.e. , @xmath35 at @xmath36 ) . we adopt two forms for the function @xmath37 : @xmath38 and @xmath39 we also follow paper i in setting the boundary conditions at the outer radius , @xmath40 : to represent steady conditions at the outer radial boundary . during the evolution of each model , we apply the constraints that in the last zone in the radial direction , all hd quantities , expect the radial velocity , are set to their initial values at all times . the radial velocity is allowed to float . in this paper we set @xmath41 using eq . 2 whereas in paper i , @xmath41 was set to zero consistently with the initial conditions . one of the simplifications made in paper i was an assumption that the radiation temperature , @xmath42 [ or more generally the spectral energy distribution ( sed ) ] does not change with the position in the flow and it was set to the temperature of the isotropic corona radiation , @xmath43 . this assumption finds justification in x - ray observations which show that quasar radiation heats a low - density gas nearly uniformally , on scales comparable to the bondi radius , up to an equilibrium compton temperature of about @xmath44 k ( sazonov et al . 2005 ; allen et al . 2006 and references therein ) . however , within our theoretical framework , the sed and @xmath42 can change with position . for example , we assume that the uv / disk radiation decreases with increasing @xmath3 ( see eq . 6 in paper thus even if the x - ray / central object radiation does not change with @xmath3 ( see eq . 7 in paper i ) the ratio between the x - ray and uv fluxes increases with increasing @xmath3 . consequently , our model predicts that radiation is softer near the poles than near the equator . moreover , we take into account the attenuation of the x - ray radiation but not of the uv radiation . thus , for a given @xmath3 the ratio between the x - ray and uv fluxes can decrease if the x - ray optical depth is high . to test how important these effects are , we allow @xmath42 to vary with position . we introduce position - dependence of @xmath42 in the following way . we start with the standard expression for the net compton heating rate : @xmath45 where @xmath46 is the radiation flux , @xmath47 is the electron number density ( other symbols have their usual meaning ) . as metioned above , we consider two sources of radiation : a disk that emits uv photons with energy between 0 and 50 ev , and a corona that emits photons with energy above 50 ev . each of this sources has its own mean photon energy , @xmath48 or equivalently radiation temperature , ( @xmath49 ) . the radiation temperature of these two sources can be computed from : @xmath50 and @xmath51 we consider these temperatures as additional free parameters . we chose the radiation temperature of the disk to be @xmath52 k and of the corona to be @xmath53 k. having set the radiation temperature of the disk and corona radiation we can compute the radiation temperature of the total radiation field from : @xmath54 where @xmath55 . to obtain the expression above we used the definition of @xmath42 and the formulae for the radiation fluxes from the disk and corona ( eqs . 6 and 15 in paper i ) . practically , to account for the position - dependent radiation temperature , we adopt the same formulae for the radiative heating / cooling rates as in paper i but replace @xmath43 with @xmath56 ( see eqs . 19 and 20 in psk ) . the seventh column in table 1 shows the values of the adopted @xmath42 or its range for cases when we use eq . we finish this section with a note about the gas temperature at the outer radius . this temperature is typically set to the compton temperature assuming that the central x - rays heat a gas up to an equilibrium compton temperature . however , in some of our simulations the optical depth toward the radiation source or the local density in the flow is so high that radiation can not heat a gas at large radii to the compton temperature . in such cases , there is a mistmatch between the gas temperature assumed at @xmath40 ( i.e. , @xmath57 ) and that computed for @xmath21 close to @xmath40 . it is possible that gas at large radii is heated not only by the central source but also by shocks or other sources such as supernovae . to mimic such sources of heating , we introduce an uniform background x - ray radiation , @xmath58 in some of our simulations . we illustrate effects of this radiation by showing results of one model ( run crbgd ) , where we assumed @xmath59 erg @xmath60 s@xmath61 , for which the gas with relatively low density is comptonized , e.g. , for @xmath62 g @xmath63 , @xmath64 ( we assume that @xmath42 of the background radiation is the same as of the central source , that it @xmath65 k. ) as in paper i , we assume the mass of the nonrotating bh , @xmath66 and the disk inner radius , @xmath67 cm throughout this paper . we compute the total accretion luminosity as @xmath68 , where @xmath69 and @xmath70 is the rest mass conversion efficiency and the mass accretion on bh , respectively . we assume a relatively high conversion efficiency appropriate for disk accretion onto a non - rotating bh , i.e. , @xmath71 . we express the accretion luminosity in in units of the eddington luminosity for the schwarzschild bh , i.e. , @xmath72 . we refer to this normalized luminosity as the eddington number , @xmath73 ) . table 1 summarizes the properties of models from paper i ( runs a , b , b1 , b2 , b3 , and c ) and our new models ( the other models listed in the table ) . columns ( 2 ) to ( 11 ) give the input parameters that we varied : the eddington number , @xmath74 , the disk contribution to the total luminosity , @xmath75 , the corona contribution to the total luminosity , @xmath76 , the uv contribution to the total luminosity , @xmath77 , the x - ray contribution to the total luminosity , @xmath78 , the radiation temperature , @xmath42 , the x - ray background radiation flux , @xmath79 , the gas temperature at the outer boundary , @xmath57 , the gas density at the outer boundary , @xmath80 , and the circularization radius , @xmath33 . columns ( 12 ) to ( 17 ) give some of the gross properties of the solutions : the mass inflow rate through the outer boundary @xmath81 , the net mass flux rate through the inner boundary @xmath82 , the mass outflow rate through the outer boundary @xmath83 , the maximum outflow velocity at the outer boundary , @xmath84 , the outflow power carried out through the outer boundary in the form of kinetic energy , @xmath85 , and in the form of thermal energy , @xmath86 . table 1 also explains our convention of labeling our runs . all other model parameters not listed in table 1 are as in paper i. simulations without rotation presented in paper i , show that an infalling gas collimates an outflowing gas and that the collimation increases with increasing radius . in addition , for a given @xmath74 , the collimation increases with decreasing ratio between @xmath78 and @xmath77 . out of three cases explored in paper i , case c with the smallest @xmath87 , shows the strongest collimation and the highest efficiency of turning an inflow into an outflow . fig . 1 in paper i and fig . 1 here show that in run c the gas is siphoned off within a very narrow channel along the pole . in paper i , we argued that the collimation , outflow power and other results will likely change if one would allow for significant gas rotation . one would expect that the gas will converge toward the equator due to the combination of the centrifugal and gravitational forces . this , in turn , will likely broaden and weaken the outflow in the polar region because less gas will be pushed toward the polar region . to test effects of rotation , we rerun models presented in paper i with rotation . we set @xmath88 that is near the maximum value for which the flow will not circularize inside our computational domain . we do not consider higher @xmath89 at the moment because we want to avoid complexities that will result from formation a rotationally supported torus or disk inside the computational domain ( e.g. , hawley , smarr & wilson 1984a ; 1984b ; clarke , karpik & henriksen 1985 ; hawley 1986 ; molteni et al . 1994 ; ryu et al . 1995 ; chen et al . 1997 ; toropin et al . 1999 ; kryukov et al . 2000 ; igumenshchev & narayan 2002 ; proga & begelamn 2003 ; chakrabarti et al . 2004 , and references therein ) . in these exploratory simulations , our choice of high @xmath33 , yielding low @xmath25 , allows us then to study first , relatively simple flows and set a stage for modeling more complex flows with high @xmath25 . we assume that the circularized gas will accrete onto smbh on a viscous time scale . however , as we do not model this part of the flow , we do not consider any details of an actual process / es leading to transport of angular momentum . the transport is most likely due to magnetorotational instability ( balbus & hawley 1991 ) but a contribution from the photon viscosity can be important in the case of high radiation luminosity . we note that our choice of low @xmath25 can be relevant to real qsos because feeding a smbh in qso with gas of high @xmath25 can be problematic . namely , for high @xmath25 , an accretion disk would form at large radii and be self - gravitating ( e.g. , paczynski 1978 ; shlosman & begelman 1987 ) . converting disk material to stars could then starve the smbh and qso would be quenched ( e.g. , goodman 2003 ) . after reviewing variety of possibilities , goodman ( 2003 ) suggested that qso disks do not extend beyond a thousand @xmath90 so that they could be gravitationally stable . if so such the disks must be replenished with gas of small @xmath25 as that we explore here . figure 1 presents the results for runs c , cr , and cr . for run cr , a step function describes the angular distribution of angular momentum on the outer boundary ( we set @xmath91 to @xmath92 in see eq . 4 ) . for @xmath93 , we assume that the specific angular momentum at the outer boundary equals @xmath94 , whereas for @xmath95 , @xmath96 . for run cr , a smooth function describes @xmath25 ( see eq . 5 ) . the figure shows the instantaneous density and temperature distributions , and the poloidal velocity field of the models . in addition , it shows the compton radius corrected for the effects of radiation pressure due to electrons ( see eq . [ 19 ] in paper i ) and the contours of the mach number equal to one , ( @xmath97 , where @xmath98 is the sound speed ) . the detailed calculations confirm our general expectations : compared with the non - rotating case , the outflow is less collimated and weaker in the rotating case . as table 1 shows , in the runs with rotation ( runs cr and cr ) , the outflow power , @xmath83 , and @xmath84 are lower than those in the run without rotation ( run c ; compare also fig . 4 in paper 1 with figs . 2 and 3 here ) . another difference is that in runs cr and cr , gas does not cool as much as in run c , especially at small radii ( i.e. , @xmath99 ) . comparison between runs c , cr , and cr shows that in run cr the solution is an intermediate one between runs c and cr which is not too surprising , given the step function is an intermediate distribution between zero-@xmath25 and the one described by function @xmath100 . in particular , the outflow in run cr is less collimated than in run c and more collimated than in run cr . one of new unexpected features that we found in run cr is that the relatively cold outflow is fragmented and time - variable . although the flow in run cr settles down into a time - averaged steady state , it is not as steady as in run c. an indication of this behavior can be found in fig . 3 that shows three radial mass flow rates as a function of radius : the net rate , @xmath101 , the inflow rate @xmath102 , and the outflow rate @xmath103 ( see eqs . 22 , 23 , 24 in paper i for formal definitions ) . for a perfect steady state , one expects @xmath104const at all radii as in run c ( see fig . 4 in paper i ) . however , in case cr the above equation holds only at small radii , @xmath105 k. we note that contrary to @xmath106 , @xmath107 is a smooth function of radius . thus the unsteadiness of in the flow appears to be caused by the unsteady behavior of the outflow , especially the outflow at large radii where it can cool down . we relate the fragmentation and time - variability of the outflow to line force turning on abruptly when @xmath0 decreases below @xmath108 k and turning off when @xmath0 increases again . 1 shows that @xmath0 decreases in the regions where the inflow sharply turns into the outflow . the density there increases and the gas radiatively cools . the turning on of line force leads to an enhanced acceleration of the outflow , but this alone is not sufficient to fragment the outflow . for example , in runs c and cr the cold outflow is not fragmented and is quite steady . thus there must be another factor / s that may contribute to fragmentation and time - variability . we note that in run c , the outflow is nearly radial hence its inner parts shield the outer parts form the central radiation . consequently , the outflow can not be heated downstream by the central radiation . however , in run cr the outflow is not radial and the flow can be heated up downstream because , as its density decreases during acceleration , it is irradiated by stronger unattenuated x - ray flux . there the outflow orientation with respect to the radiation flux appears to be one of the key factors causing fragmentation and time - variability of the outflow . this conclusion is supported by the fact that even in run cr , the outflow is not fragmented at large radii where the outflow becomes almost radial , and clumps merge with each other . to show the variable solution in more details , in fig . 4 we present a sequence of density maps of the inner part of the flow in run cr at four different times . the left panel shows the flow at a time when a clump brakes from a high density filament at @xmath109 . subsequent panels show how this and other clumps move outward and how they are stretched . 4 shows also formation a new clump ( the second panel from the right ) . we find that clumps form usually at the same location ( i.e. , at @xmath110 and @xmath111 ) every @xmath112 s or so which is of order of a dynamical time scale at radius where the clumps form . generally , despite the time - variability , the instantaneous maps shown in fig . 1 are representative of run cr because they show an example of a large scale inflow and outflow with continuous production of small scale clumps that merge at large radii ( i.e. , beyond @xmath113 ) . 3 and 4 show that in runs cr and cr the outflow power is dominated by the kinetic energy not the thermal energy . however , comparing with run c ( see fig . 4 in paper i ) the dominance is not as strong . we conclude that in case c , rotation reduces the outflow collimation and the outward flux of mass and kinetic energy . rotation also leads to fragmentation and time - variability of the outflow and an increase of the outward flux of the thermal energy . as expected , rotation does not change much the mass inflow rate through the outer boundary . figs . 5 and 6 show results for case b with and without rotation , i.e. , runs b and br ( see also fig . 3 in paper i ) . in this case , rotational effects are almost the same as in case c. the main difference is that in run br , an outflow does fragment and the overall flows settles down into a steady state . this however is understandable : in run br , radiative heating is strong and the gas does not cool therefore line force does not turn on . we return now to case c and consider effects of the position - depenedent @xmath42 . left panels of fig . 7 and fig . 8 show results for run cx . in comparison with run c , the outflow in run cx is broader . the mass outflow rate in run cx is only slightly higher than in run c. however , the outflow rate almost cancels out the inflow rate so that the net rate is two orders of magnitude smaller than the mass flux through the outer boundary . run cx is a good example where agn irradiation can significantly reduce the rate at which the central engine is fed with matter . in run cx , @xmath42 is lower near the poles than near the equator . in addition , @xmath42 near the pole is lower in run cx than @xmath42 in run c. the latter difference explains why the outflow in run cx is so strong : the relatively low @xmath42 in the polar region leads to a lower gas temperature in this region . this in turn leads to more mass being pushed towards the pole . this mass can then be effectively turned into an outflow because in the polar region the radiation flux is sub - eddington and additionally , line - force turns on there because the gas temperature is low enough . in other words , the siphon effect seen in many of our simulations is very strong in run cx . right panels of fig . 7 and fig . 9 show results for run crx that is a rerun of run cx with rotation . comparing these two run , we find that effects of rotation in the simulations with the position - dependent @xmath42 are similar to those in the simulations with constant @xmath43 . namely , rotation decreases the degree of the outflow collimation and decreases the outward flux of mass and kinetic energy . in addition , rotation leads to an increase of the outward flux of the thermal energy . in run crx , the cold outflow is nearly radial and does not fragment as much as in run cr . in this paper , we do not present results for case a with rotation because in this case the flow is dominated by thermal effects and rotation does not change much the solution . however , we present here results for case a with the position - dependent @xmath42 ( run ax ) because they show new effects . fig . 10 compares results for runs a and ax . in run ax , the flow pattern is different that that seen in runs presented in paper i or runs shown here so far . the dramatic difference is that in run ax the equatorial inflow is replaced by an equatorial outflow . generally , in run ax there are equatorial and polar outflows both being fed by an inflow of gas at intermediate polar angles . the equatorial outflow is a simple consequence of the higher radiation temperature near the equator that leads to a high gas temperature and an enhanced thermal expansion . in the polar region , where the gas temperature is lower , an outflow is driven by thermal expansion assisted by radiation pressure on electrons . overall the flow pattern in run ax is dominated by the outflow . this leads to the net mass inflow rate at small radii being one order of magnitude lower than the inflow rate at large radii . we conclude that both rotation and position - dependent @xmath42 lead to qualitative and qualitative changes in the flow . most prominent of them are weaker collimation and fragmentation of the outflow in cases with rotations , and production of an thermal equatorial outflow in cases with position - dependent @xmath42 . a realistic model of accretion flows should include many physical effects . here , we focused on the role of gas rotation and position - dependent @xmath42 . we consider our simulations just as exploratory tests . these tests support the notion that agn radiation can play a very important role in determining the ionization structure and dynamics of matter on sub- and parsec - scales . we finish the presentation of our models with showing results for two runs that illustrate how complex the flow dynamics can be even in a very simple set - up as the one we focus on here . one of our motivations is to understand gas dynamics in broad line regions ( blrs ) and narrow line regions ( nrl ) so characteristic to agn . these regions are thought be made of cold gas clouds moving randomly or nearly randomly and having a small filling factor ( e.g. , krolik 1999 and reference therein ) . formation , evolution and other key aspects of these clouds are not well understood . our simulations show that an accretion flow which is initially smooth and spherical can break into inflows and outflows . in cases with rotation , we have seen outflows fragmented due to line force and x - ray irradiation . these results raise the following question : can we produce many cold clouds with a small filling factor . the answer seems to be yes as we show in figs . 12 and 13 , and 14 . figs . 12 , 13 , and 14 compare results for case c with @xmath114 and @xmath115 g @xmath63 . left panels in fig . 12 show results for run crgd . the density and temperature distribution and as well as some other properties of this run differ a lot from other runs shown here . the main difference is a much larger dynamical range in the temperature and density plus the fact that the flow is far from reaching any steady state . however , close inspection of the results from run crgd show that the flow pattern is similar to that seen in most other runs : there is an equatorial inflow and a bipolar outflow . the latter is not fragmented because it is radial in agreement with our explanation of the outflow fragmentation . the large dynamical range in the density and temperature in run crgd is a simple consequence of the higher density at the outer boundary . a high density gas can cold much faster than a low density gas . in this run the gas is also heated by shocks because for @xmath114 radiation pressure of electrons alone can drive a powerful and broad outflow that collides with an inflowing gas . as we discussed in 2.2 , an accretion flow at large radii does not have to be heated by the central radiation source . the lower left panel in fig . 12 illustrates this point as one can see that near the equator a good fraction of the inflow is relatively cold . one of the prediction of such a model is that for a wide range of the inclination angle , this cold gas should produce absorption lines redshifted with respect to the systematic velocity . however , such lines are not being observed . the problem of large scale accretion of cold gas can be easly overcome by introduction of non - central source of heating . right panels in fig . 12 , show results for run crbgd which is a rerun of model crgd with the x - ray background radiation ( see 2.2 ) . in run crbgd , there are no large regions of shock heated gas . the only region where shock heating is important is in a narrow polar region of a low density . the background radiation heating in this run keeps the gas from rapid cooling which in turn can lead to abrupt turn - on of line driving and strong expulsion of gas as seen in run crgd . comparing runs cr and crbgd , we see that an outflow tends to fragment more if the gas density is higher . the density and temperature maps show that the outflow occupies a relatively large fraction of the computational domain . however , fig . 14 shows that this outflow does not change much the overall mass budget . as in run cr , this is caused by rotation that reduces the amount of gas that is pushed toward the polar region where it can be siphoned off . we have calculated a series of models for rotating flows that are under the influence of supermassive bh gravity and radiation from an accretion disk surrounding the bh . we seek to determine self - consistently what fraction of the flow is gravitationally captured by the bh or what fraction is driven away by thermal expansion and radiation pressure . this work is an extension of the work presented in paper i , where nonrotating flows were studied . here , we consider effects of rotation and of a position - dependent radiation temperature , density at large radii , and an uniform x - ray background radiation . as in the non - rotating case , the rotating flow settles into a configuration with two components ( 1 ) an equatorial inflow and ( 2 ) a bipolar inflow / outflow with the outflow leaving the system along the pole . however , the rotating flow does not reach a steady state . in addition , rotation reduces the outflow collimation and the outward flux of mass and kinetic energy . moreover rotation increases the outward flux of the thermal energy and can lead to fragmentation and time - variability of the outflow . in future , we plan to check whether thermal instability can contribute to fragmentation and time - variability of the outflow . as expected , rotation does not change much the mass inflow rate through the outer boundary . in our model , the radiation comes from an uv emitting disk and x - ray emitting spherical corona . as a result , the radiaton temperature is position - dependent : in the polar region radiation is dominated by a softer disk component whereas near the equator , radiation is dominated by a harder corona component . the two main changes due this position - dependence are : ( i ) an increase in the power of the outflow in the polar region and ( ii ) development of a large scale thermally driven outflow in the equatorial region . overall , we conclude that our exploratory study provides an additional support to the idea that agn radiation can significantly change gas dynamics and photoionzation structute on sub - parsec- and parsec - scales . as it have been discussed and shown in the past , we found that agn radiation can significantly reduce the rate at which the central bh is fed with matter ( e.g. , figs . 8 and 13 ) . this result should not depend on the inner radius of the computional domain because most of the outflow is launched from a radius larger than the inner radius of the computational domain . however , we note when reducing the inner radius of the computional domain one should also consider additional processes , in particular , disk accretion and disk winds and jets . thus our mass inflow rate should be viewed only as an upper limit for the bh accretion rate . our results emphasize also a little appreciated feature , i.e. , agn radiation can drive a non - spherical , multi - temperature and very dynamic flow pattern . this result may have implications for the problem of agn feedback and the problem of the origin , geometry and physics of nlr and blr . this work is supported by nasa through grants hst - ar-10680 and hst - ar-11276 from the space telescope science institute , which is operated by the association of universities for research in astronomy , inc . , under nasa contract nas5 - 26555 .	we study the axisymmetric , time - dependent hydrodynamics of rotating flows that are under the influence of supermassive black hole gravity and radiation from an accretion disk surrounding the black hole . this work is an extension of the earlier work presented by proga , where nonrotating flows were studied . here , we consider effects of rotation , a position - dependent radiation temperature , density at large radii , and uniform x - ray background radiation . as in the non - rotating case , the rotating flow settles into a configuration with two components ( 1 ) an equatorial inflow and ( 2 ) a bipolar inflow / outflow with the outflow leaving the system along the pole . however , with rotation the flow does not always reach a steady state . in addition , rotation reduces the outflow collimation and the outward flux of mass and kinetic energy . moreover rotation increases the outward flux of the thermal energy and can lead to fragmentation and time - variability of the outflow . we also show that a position - dependent radiation temperature can significantly change the flow solution . in particular , the inflow in the equatorial region can be replaced by a thermally driven outflow . generally , as it have been discussed and shown in the past , we find that self - consistently determined preheating / cooling from the quasar radiation can significantly reduce the rate at which the central bh is fed with matter . however , our results emphasize also a little appreciated feature . namely , quasar radiation drives a non - spherical , multi - temperature and very dynamic flow . these effects become dominant for luminosities in excess of 0.01 of the eddington luminosity . = = = 1=1=0pt = 2=2=0pt
supernovae ( sne ) inject kinetic energy , momentum , and enriched ejecta into the interstellar medium ( ism ) and are thought to be one of the main sources of interstellar turbulence and galactic outflows . an important factor in understanding these processes is the evolution of supernova remnants ( snrs ) as their ejecta encounter various phases of the ism . for example , a large fraction of core - collapse sne from high - mass stars may explode promptly and interact with the molecular cloud ( mc ) from which they formed . this hypothesis suggests that the direct interaction between supernova ( sn ) ejecta and mcs , which we refer to as _ snr - mc interactions _ , may occur at a rate significantly higher than expected from the filling fraction of molecular gas in the ism @xcite . while it has been estimated that as many as half of all snrs have ejecta in physical contact with mcs @xcite , this estimate is highly uncertain and relies on a number of assumptions about massive stars , sne , and the relation between sne and mcs . detection of snr - mc interactions has traditionally relied on a number of methods for observing shocked molecular gas directly . strong , narrow emission from the oh @xmath5 maser ( @xmath6 ) is often detected in the vicinity of snrs , especially those with other signatures of snr - mc interactions @xcite . more recent surveys have established that targeted searches for oh maser emission may be a reliable method of searching for snr - mc interactions ( e.g. , * ? ? ? * ; * ? ? ? to date , however , oh masers have only been detected toward @xmath7 of the 294 known snrs in our galaxy @xcite , although a lack of detailed spectroscopic information and localization makes it impossible to associate all of these detections with snrs . this low incidence relative to the expected rate of snr - mc interactions may be due to the unique physical conditions required for maser emission . oh @xmath5 masers are collisionally excited behind slow ( @xmath8 ) c - type shocks in regions with moderate temperatures and densities ( @xmath9 , @xmath10 ) @xcite . studies of the warm , diffuse h@xmath11 regions around interacting snrs indicate that physical conditions are often outside this range @xcite . regions with higher shock velocities could also depopulate the hydroxyl @xmath12 state or dissociate the molecule entirely . convincing evidence of snr - mc interactions must come from other sources in these cases . internal flows within snrs are expected to become highly turbulent where the snr shock front interacts with the dense ism @xcite . molecular lines toward snrs , such as the rotational transitions of @xmath0co , should be velocity - broadened as a result of increased turbulence from a snr - mc interaction . shocked mcs are also known to exhibit asymmetric line profiles where these interactions occur @xcite . as opposed to the presence of oh @xmath5 maser emission , the detection of velocity - broadened , broad molecular line ( bml ) @xmath0co relies on more detailed spectroscopic information , especially in faint line `` wings '' where turbulent broadening is usually most apparent . at the same time , this type of observation offers insight into snr evolution and the physical properties of both sn ejecta and mcs . even if the studies of oh maser emission , bml emission , and other shock indicators are all taken together , only about 60 galactic snr - mc interactions have been identified . however , to date there has been no large - scale systematic survey for shock - broadened molecular lines . instead , the interactions have not been identified in a homogeneous manner , but are generally a result of investigations of individual snrs , notably ic 443 , w28 , w44 , ctb 109 , kes 69 , kes 79 , w51c , and cassiopeia a ( cas a ) @xcite . these sources span a wide range of ages and morphologies , from young , shell - like snrs such as cas a , through middle - aged and older ( @xmath13 ) remnants , and including mixed - morphology remnants and plerions . a systematic and unbiased survey of broadened co emission toward galactic snrs may yield a larger incidence of such interactions . the work described in this paper uses co emission for the first large - scale , systematic search for bml regions toward snrs . to interpret these observations , we developed an algorithm to identify regions of broadened @xmath0co @xmath2 emission where they exist toward each snr - mc system . we justify these identifications using the approximate systemic velocity at the distance inferred to each snr , an estimate of the `` broadened '' co emission , and the radio continuum morphology of the remnant itself . in this way , we have identified bml regions toward nineteen snrs in our sample , including nine newly identified candidate snr - mc interactions . lccccccccccc [email protected] & kepler & @xmath14 & @xmath15 & @xmath16 & 0.17 & 4.0 & & s & & hess ( 0.68 ) & + [email protected] & & @xmath17 & @xmath18 & @xmath16 & 0.08 & 16.3 & d & s & & hess ( 14.6 ) & + [email protected] & & @xmath19 & @xmath20 & @xmath21 & 0.11 & 6.4 & d & s & & hess ( -0.1 ) & + [email protected] & & @xmath22 & @xmath23 & @xmath16 & 0.15 & 14.4 & & s & & hess ( 4.6 ) & + [email protected] & & @xmath24 & @xmath25 & @xmath16 & 0.09 & 4.4 & d & c & y & hess ( 5.5 ) & 1 + [email protected] & & @xmath26 & @xmath27 & @xmath16 & 0.09 & 19.4 & & s & & hess ( 1.7 ) & + [email protected] & & @xmath28 & @xmath29 & @xmath30 & 0.12 & 10.9 & & s & & hess ( 0.7 ) & + [email protected] & & @xmath31 & @xmath32 & @xmath16 & 0.09 & 15.6 & d & s & & hess ( 2.4 ) & + [email protected] & & @xmath33 & @xmath34 & @xmath16 & 0.12 & 15 & & c & & hess ( 4.7 ) & + [email protected] & & @xmath35 & @xmath36 & @xmath16 & 0.11 & 4.8 & & c & y & hess ( 24.3 ) & 2 + [email protected] & & @xmath37 & @xmath38 & @xmath16 & 0.12 & 13.3 & & s & & hess ( 3.3 ) & + [email protected] & & @xmath39 & @xmath40 & @xmath16 & 0.11 & 18.7 & & s & & hess ( 1.8 ) & + [email protected] & & @xmath41 & @xmath42 & @xmath43 & 0.11 & 8.5 & & s & & hess ( 1.4 ) & + [email protected] & & @xmath44 & @xmath45 & @xmath16 & 0.15 & 10 & dp & c & & hess ( 4.7 ) & 3,4,5 + [email protected] & & @xmath46 & @xmath47 & @xmath16 & 0.18 & 18.1 & & s & & hess ( 6.6 ) & + [email protected] & & @xmath48 & @xmath49 & @xmath50 & 0.13 & 5.6 & & s & & hess ( 6.9 ) & + [email protected] & & @xmath51 & @xmath52 & @xmath53 & 0.13 & 13.2 & d & s & & hess ( 2.8 ) & + [email protected] & & @xmath54 & @xmath55 & @xmath16 & 0.09 & 18.9 & & s & & hess ( 1.2 ) & + [email protected] & & @xmath56 & @xmath57 & @xmath16 & 0.07 & 4.8 & & c & y & hess ( 6.4 ) & 7 + [email protected] & & @xmath58 & @xmath59 & @xmath50 & 0.12 & 6.9 & d & ? & & hess ( 1.5 ) & + [email protected] & [email protected] & @xmath60 & @xmath61 & @xmath16 & 0.10 & 8.5 & d & s & y & hess ( 6.0 ) & 3,8 + [email protected] & & @xmath62 & @xmath63 & @xmath50 & 0.16 & 10 & d & s & y & hess ( 4.5 ) & 9 + [email protected] & kes 75 & @xmath64 & @xmath65 & @xmath16 & 0.13 & 6.0 & dp & s & y & hess ( 10.1 ) & 3,10,11 + [email protected] & & @xmath66 & @xmath67 & @xmath68 & 0.14 & 4.7 & & ? & & hess ( 0.9 ) & + [email protected] & & @xmath69 & @xmath70 & @xmath71 & 0.11 & 6.3 & & s & & hess ( 1.9 ) & + [email protected] & 3c 391 & @xmath72 & @xmath73 & @xmath16 & 0.10 & 7.2 & dp & s & & hess ( 0.0 ) & 5,12,13,14 + [email protected] & & @xmath74 & @xmath75 & @xmath76 & 0.10 & 17 & d & s & & hess ( 2.2 ) & + [email protected] & & @xmath77 & @xmath78 & @xmath71 & 0.11 & 5.7 & & s & & hess ( 2.5 ) & + [email protected] & kes 79 & @xmath79 & @xmath80 & @xmath16 & 0.13 & 7.1 & dp & s & y & hess ( 2.7 ) & 3,15,16 + [email protected] & & @xmath81 & @xmath82 & @xmath83 & 0.09 & 8.7 & & s & & hess ( -1.4 ) & + [email protected] & 3c 396 & @xmath84 & @xmath85 & @xmath16 & 0.11 & 6.2 & dp & c & y ? & hess ( 0.5 ) & 3,17,18,19 + [email protected] & 3c 397 & @xmath86 & @xmath87 & @xmath16 & 0.10 & 10.3 & dp & s & & hess ( 3.0 ) & 3,20 + [email protected] & w49b & @xmath88 & @xmath89 & @xmath16 & 0.09 & 2.5 & dp & s & y ? & hess ( 1.0 ) & 3,21,22 + [email protected] & & @xmath90 & @xmath91 & @xmath16 & 0.09 & 6.2 & p & c & y & hess ( 3.8 ) & 3,17,23,24 + [email protected] & 4c-21.53 & @xmath92 & @xmath93 & @xmath21 & 0.19 & 8.2 & & s & & hess ( 0.8 ) & + [email protected] & & @xmath94 & @xmath95 & @xmath96 & 0.11 & 7.3 & & f & & hess ( 1.6 ) & + [email protected] & & @xmath97 & @xmath98 & @xmath16 & 0.12 & 3.8 & & f & & & + [email protected] & ctb 80 & @xmath99 & @xmath100 & @xmath16 & 0.07 & 1.5 & & f & y & magic ( 0.6 ) & 25 + [email protected] & & @xmath101 & @xmath102 & @xmath103 & 0.14 & 7.1 & & s & & & + [email protected] & ctb 87 & @xmath104 & @xmath105 & @xmath106 & 0.08 & 6.1 & & f & y ? & veritas ( 6.2 ) & 17,21 + [email protected] & & @xmath107 & @xmath108 & @xmath16 & 0.10 & 10 & & c & y & & 17 + [email protected] & & @xmath109 & @xmath110 & @xmath16 & 0.15 & 11.9 & & s & & & + [email protected] & & @xmath111 & @xmath112 & @xmath113 & 0.11 & 6.0 & & s & & & + [email protected] & cas a & @xmath114 & @xmath115 & @xmath16 & 0.05 & 3.4 & dp & s & y & magic ( 5.2 ) & 26,27,28,29 + [email protected] & ctb 1 & @xmath116 & @xmath117 & see & 0.09 & 3.1 & & s & y & & 30 + [email protected] & tycho & @xmath118 & @xmath119 & @xmath16 & 0.10 & 2.5 & & s & & veritas ( 5.8 ) & + [email protected] & 3c 58 & @xmath120 & @xmath121 & @xmath16 & 0.08 & 2.0 & & f & y & magic ( 5.7 ) & 31 + [email protected] & hb 3 & @xmath122 & @xmath123 & see & 0.11 & 2.0 & dp & s & y & & 32,33 + [email protected] & crab & @xmath124 & @xmath125 & @xmath16 & 0.10 & 2.0 & & f & y & hess ( 129 ) & 34 + [email protected] & ic 443 & @xmath126 & @xmath127 & @xmath128 & 0.09 & 1.5 & dp & c & y ? & veritas ( 8.3 ) & 35,36 the snrs in our survey were selected from those accessible with the heinrich hertz submillimeter telescope ( smt ) on mt . graham , arizona , @xmath129 . in general , all snrs with declination greater than @xmath130 were considered , yielding a sample between galactic longitude @xmath131 and @xmath132 degrees . there are currently 160 known snrs in this volume @xcite . from this larger sample , we selected snrs with an apparent size less than @xmath133 in diameter along their largest axis . the targets we observed are listed in . some targets listed in were selected for comparison to our larger sample . these snrs are well - studied and large ( diameter @xmath134 ) and in these cases we mapped a smaller subregion of the entire remnant . these targets include ctb 80 , ctb 1 , hb 3 , and ic 443 . ic 443 is a well - known snr - mc system , while hb 3 is suspected to have snr - mc interactions @xcite . reliable distances are critical when characterizing snrs . in order to draw an association between broad - line mcs and the snrs themselves , a variety of distance indicators exist such as proper motion estimates , radial velocity , absorption , and ob associations , each with varying degrees of uncertainty . where the properties of snrs and their environments differ , distance indicators can yield estimates with significant disagreement . for example , the distance to [email protected] ( kepler / sn 1604 ) has historically been controversial , with estimates ranging from around @xmath135 kpc @xcite to @xmath136 kpc @xcite , although proper motion estimates now favor the closer distance . this lack of agreement between various methods underscores the systematic uncertainties present in all distance indicators , from discrepant optical absorption @xcite to disagreement in the maximum optical brightnesses of historical sne @xcite . for these reasons , we carefully chose distance indicators from among the literature , favoring those determined from proper motion estimates , kinematic estimates from absorption and co associations , and line - of - sight x - ray absorption . the values we chose are given in and described in . however , for nineteen of the remnants in our survey , we were unable to find reliable distance indicators in the literature . for these objects , we resort to the surface brightness - size or @xmath137 relation @xcite . this empirical relationship describes snr evolution with time , whereby the radio surface brightness ( @xmath138 ) decreases monotonically while the physical diameter ( @xmath139 ) increases monotonically . thus , a power law relationship can be fit such that @xmath140 @xmath141 and @xmath142 are calibrated using snrs whose distances are known from other methods . measuring the radio surface brightness and angular size of the snr then yields an angular diameter distance . @xcite fit this relationship and derive distances to 225 galactic snrs . the authors find the best - fit slope to be @xmath143 . this method involves significant systematic uncertainties and should be used with caution . distance measurements for some of the snrs in our sample are taken from this study where we adopt the authors most probable distance ( i.e. , the orthogonally fit value from table 4 in @xcite ) . these include sixteen out of the 50 snrs we observed . we provide our own estimates for the remaining three snrs , which we describe in . we mapped the vicinity of 50 snrs ( , ) with the 10 m smt using the alma - type sideband separating 1.3 mm receiver between 2013 april 21 and 2015 may 8 . apart from the size and average rms noise of each map as given in , the observations and reduction for each remnant were identical to those described in @xcite . as discussed in , there were several targets for which we did not observe the entire remnant . rather , we targeted either a region we knew to have broadened lines or an x - ray ( i.e. , @xmath144 ) bright region toward any known molecular emission . these objects were used either for comparison to other targets ( hb 3 , ic 443 ) or as a targeted search for new bml regions toward large remnants ( ctb 80 , ctb 1 ) . in particular , we targeted two regions of x - ray bright emission from ctb 1 , centered at @xmath145 with dimensions ( @xmath146 ) @xmath147 and at @xmath148 with dimensions @xmath149 . similarly , for hb 3 we targeted two regions centered at @xmath150 with dimensions @xmath147 and at @xmath151 with dimensions @xmath16 . these regions were centered on local x - ray enhancements as reported by @xcite . for the other large targets , we used single fields as reported in . we analyzed these four remnants and we indicate whether bml regions were detected in each case . however , given the fact that we did not map the entirety of each remnant and some were selected on the basis of previous evidence for snr - mc interactions , we discount all four in our discussion of snr - mc interactions ( except in ) as a possible source of bias . in our analysis , we used the @xmath152 resolution data , which has 512 channels for a velocity range of @xmath153 at @xmath154 . for a receiver tuned to the @xmath0co @xmath2 and centered at a systemic velocity of @xmath155 , we can be reasonably certain that all galactic molecular emission along the line - of - sight is observed . the expected velocity range of the co spectrum at each distance is determined by assuming a @xmath156 uncertainty on the distances we assume to each snr , which agrees with typical fractional errors for distances calculated using the @xmath157 relation @xcite . comparison with the clemens rotation curve @xcite allows us to limit the velocity range over which mcs might exist that are interacting with a snr . we extend this range by @xmath158 at either end of the range in order to account for systematic errors in finding bml regions associated with snrs ( see , e.g. , ) . demonstrates this method for the known bml region toward 3c 391 . the distance to this snr is taken to be @xmath159 kpc , which at galactic coordinates @xmath160 corresponds to a systemic velocity range of @xmath161 to @xmath162 or @xmath163 to @xmath164 with the added velocity range . within the range of permitted velocities for each snr , we imposed criteria on the line width and profile of observed molecular emission which we describe and justify in . in summary , these criteria are : * we required co emission to be detected at @xmath165 ( the root mean square of emission in the line - free spectrum at each pixel as in ) in every spectral channel we used to calculate the line profile . * we used only five spectral channels for the width determination . * we required that the line profile not include secondary maxima . * we required the fwhm of observed @xmath0co lines @xmath166 . the purpose of these criteria was to identify regions of molecular emission that correspond with disturbed molecular gas where a snr - mc interaction has occurred . at the same time , we attempted to minimize false positives , e.g. due to molecular emission overlapping in velocity - space , and false negatives , e.g. where interactions are subtle and exhibit small velocity - widths , the co line is weak , or strongly asymmetric line profiles are present . for clouds with sizes on the order of a few parsecs , quiescent @xmath0co line widths are typically @xmath167 @xcite . therefore , we adopted a threshold of a @xmath168 fwhm to indicate bml regions . we evaluate and discuss the validity of this threshold in . furthermore , bml regions can have asymmetric line profiles where the molecular gas has been shocked and accelerated . in addition to large line widths ( i.e. , @xmath169 ) at negative or positive velocities relative to the peak , the line profile is characterized by faint wings where turbulent acceleration of molecular gas is apparent . this line profile can be confused with multiple mcs along the same line - of - sight ( e.g. , toward the galactic center ) resulting in superimposed emission in the spectrum of a small region or even a single pixel . thus , identifying bml regions ideally relies on as small a region as possible with as few spectral channels as possible . to account for line widths , asymmetry , and molecular emission from multiple sources along a line - of - sight , we developed an algorithm for identifying bml regions in @xmath0co @xmath2 . initially , we constructed a list of individual spectral channels by examining the spectrum at each pixel . where the @xmath0co emission in individual spectral channels exceeded @xmath170 and represented a local maximum , we flagged that channel for analysis . these flagged spectral features span the entire data cube and represent the line peaks of individual emission features in our data . at @xmath154 and for a spectrum with @xmath152 resolution , the @xmath168 velocity - width threshold we adopted corresponds to a feature whose fwhm is @xmath171 or roughly five spectral channels . thus , to measure the line width of a broad feature , we required at least five channels or four channels in addition to the flagged local maximum of a spectral feature . confusion where a single spectral channel contains emission from multiple mcs can be problematic . in part , we minimized confusion by limiting the number of channels we used to measure line widths to the absolute minimum . however , this method does not preclude the scenario in which two molecular features with narrow widths ( i.e. , @xmath172 ) are separated by less than their half widths in velocity - space . therefore , in addition to the requirements described above , our algorithm stipulates that the emission on either side of the line peak must decrease monotonically in each channel and that the emission in each channel exceeds @xmath165 . finally , using the five spectral channels around each flagged line peak , we measured the line widths using a method that accounts for possible asymmetries in the line profile . we took the spectrum at a particular pixel @xmath173 and channel @xmath174 corresponding to one of the flagged local maxima in our data cube . here , we use the antenna temperature at a particular pixel , @xmath175 . next , we measured the velocity centroid @xmath176 of the line peak around @xmath174 by calculating the first moment of the spectrum @xmath177 we then evaluated the second moment of the spectrum in this range , @xmath178 this statistic is the velocity dispersion of the spectrum around the line peak . for a gaussian line profile , the fwhm @xmath179 . for comparison to other line indicators , we used the velocity - width statistic @xmath180 . if this value exceeds @xmath168 , the peak of the corresponding line profile is flagged in our data cube as a bml . for snrs with newly identified bml regions , we give the total number of pixels with bml identifications in parentheses in . lcccc [email protected] & 7 ( 10 ) & + 2.6 & 7.3 & 16.4 ( 16.3 ) + [email protected] & 8 ( 10 ) & + 31 & 7.1 & 4.0 ( 6.4 ) + [email protected] & 15 ( 21 ) & + 32 & 8.1 & 7.2 ( 4.4 ) + [email protected] & 2 ( 3 ) & + 50 & 6.9 & 11.8 ( 15.6 ) + [email protected] & 12 ( 15 ) & + 42 & 8.5 & 12.7 ( 13.2 ) + [email protected] & 7 ( 11 ) & + 91 & 6.7 & 5.6 ( 6.9 ) + [email protected] & 12 ( 20 ) & + 100 & 9.0 & 9.0 ( 12.8 ) + [email protected] & 5 ( 5 ) & + 94 & 10.0 & 9.0 ( 10 ) + [email protected] & 3 ( 4 ) & + 43 & 7.5 & 11.8 ( 18.5 ) as described in , local maxima in the spectrum at each pixel are flagged if the line profile calculated at that point matches the outlined criteria . in this way , we can construct a separate `` flagged '' data cube with dimensions equal to the original data cube , that is , the antenna temperature @xmath181 . this `` flagged '' data cube has @xmath182 if the spectrum at position @xmath183 and in the velocity channel @xmath184 meets our criteria , and @xmath185 otherwise . we can further break down the `` flagged '' data cube into regions by using a group - finding algorithm simultaneously in spatial coordinates and velocity - space . for the purposes of this analysis , the definition of bml regions will ultimately depend on the physical significance of these groups . as described in , our @xmath0co @xmath2 observations were performed with a 10 m diameter telescope at @xmath154 with a fwhm beam of @xmath186 . each data cube was constructed by taking observations spaced @xmath187 apart . in our analysis , detection of pixels with broad lines whose centroids lie within each others fwhm and within a radius of 2 pixels represents a single bml region . for bml regions that appear extended on the sky , we can define groups that span several pixels , provided the broad - line detection is approximately at the same velocity ( i.e. , within a window spanning @xmath188 measured for each bml ) . a velocity centroid is measured for each bml detection representing a local maximum in the @xmath0co spectrum . however , each of these points is biased by the way in which we construct the original data cube @xmath189 , and the determination of a group of bml detections must be physically significant . in the same way , the velocity centroid of each local maximum must be tied in some way to the actual velocity centroid for the entire bml region with which it is associated . therefore , we utilized a signal - weighted average to measure this value , for which the bml region velocity centroid ( @xmath190 ) is defined by the number of local maxima it includes ( @xmath191 ) , the gaussian amplitude of each local maximum ( @xmath192 ) , and the velocity centroid calculated ( @xmath193 as described in ) @xmath194 for newly discovered bml regions toward snrs , we report in the number of bml pixels we detect toward each remnant , the signal - weighted velocity centroid as described above , and the average velocity - width of all bml detections toward the snr . finally , we recalculated the distance to each remnant using the signal - weighted velocity centroid as a prior for the clemens rotation curve and we report this distance in . in order to resolve the distance ambiguity , we chose the distance closest to our original estimate from . the agreement between the distances reported in is variable , although nearly always within the @xmath156 uncertainty on the original estimate given the prior we used to find associated co emission ( ) . in some cases , these values are very close to the original distances , especially for [email protected] , [email protected] , [email protected] , and [email protected] , where the distances were estimated from the @xmath157 relation for the first three objects and x - ray column density for the latter . we applied our algorithm to two test cases involving known snr - mc systems : 3c 391 ( ) and cas a ( ) . in this way , we demonstrate the types of bml regions selected by our algorithm and the accuracy of the measurements . these two systems present unique problems that must be addressed in identifying bml regions in a large sample of @xmath0co @xmath2 data . here , we discuss previously detected signatures of bml regions toward each remnant and the underlying snr - mc interactions they may trace . from these examples , we examine the extent to which our algorithm can be optimized . several observations of @xmath0co targeted on 3c 391 have revealed the presence of broadened molecular features from @xmath195 to @xmath196 @xcite . the emission exhibits significant broadening relative to the ambient gas , with @xmath0co @xmath2 line widths exceeding @xmath197 in some regions ( ) . this emission has consistently been interpreted as a signal of postshock gas . the molecular line profiles also exhibit a bright , narrow component ( @xmath198 ) at the same velocities and peaking in the same region . this observation suggests an association between molecules of significantly different temperatures and densities and presents a challenge when identifying regions of shocked molecular gas . one hypothesis for the presence of the narrow component is that it arises from cooler , less turbulent gas reforming behind a dissociative shock @xcite . that is , while the narrow component may be associated with the snr shock , in some regions it could be superimposed upon and dominate the line profile from the broadened emission . under this hypothesis , we might only detect bml regions where the snr - mc interaction is ongoing or recent . thus , detection of recently shocked gas in @xmath0co may be extremely sensitive to the criteria for identifying individual line components ( i.e. , local maxima ) in the spectrum . in our analysis , we selected all the local maxima in velocity - space whose signal exceeds @xmath170 . in this way , our criteria are unbiased by the coincidence of narrow and bright line components with broad and faint components as in 3c 391 . one caveat is that false detections can occur where the local maxima are close together in velocity - space ; for example , we might report no detection if the narrow component was within @xmath199 spectral channels of the broad component . in this case , additional information on the line profile would be necessary to calculate an accurate line width . while this scenario does not occur in 3c 391 , one could imagine dissociated molecular gas reforming behind a shock whose narrow emission peaks at roughly the same systemic velocity as the undissociated , turbulently broadened component . as a follow - up to our initial analysis , we modified our requirement for the five channels to include scenarios where the line profile does not decrease monotonically for two channels on either side of the local maxima . while we still required five velocity channels , we extended the channel window on either side of the line profile to avoid contamination from other emitting regions in velocity - space . as we demonstrate in ( left ) , the well - studied , shocked mc toward 3c 391 is located to the south of the remnant along a region of bright radio continuum . the @xmath0co spectrum in this region is significantly broadened in a feature centered around a systemic velocity of @xmath200 as seen in ( right ) . our data is of sufficient spatial and spectral resolution to resolve and confirm the narrow and broad spectral features toward this cloud . another important point that arises when examining similar bml regions is that the @xmath200 line appears to lie outside the velocity range inferred for an object at @xmath201 and @xmath202 , and the broad profile itself is at least @xmath158 from the maximum velocity allowed within this range . the remnant appears to be near the tangent point of the galactic rotation curve at this velocity ( @xmath162 ) , but the mc is slightly redshifted with respect to this velocity . given that the shock interaction with the mc has been confirmed in multiple wavelengths and is generally assumed to be associated with the remnant , one hypothesis for the velocity discrepancy is that remnants may accelerate mcs in an ongoing interaction . this point underscores the need for care in treating snr - mc interactions whose distances are near the tangent point of the rotation curve along a particular line - of - sight . as we describe in , we extend the velocity window used as a prior in our co spectra by @xmath158 at either end to account for this potential problem toward other snrs . cas a exhibits subtle indications of a snr - mc interaction in several ways . warm co gas ( t @xmath203 ) directly to the west of the snr implies some source of energy to heat the cloud @xcite . there is also a correlation between the brightest portions of the radio shell at @xmath204 , nonthermal x - ray emission , and the radio spectral indices in the same region suggesting that the ejecta are encountering a denser medium @xcite . in the near - infrared k@xmath205 band , strong nonthermal emission occurs in the same regions @xcite . these observations are consistent with an interaction scenario , although they do not conclusively point to the presence of a snr - mc interaction . recent results indicate that hard x - ray emission above @xmath206 toward cas a is dominated by the western half of the remnant @xcite . knots of bright emission in nonthermal emission are consistent with steep spectral indices in the x - ray and some change in the physical environment . these knots are coincident with bright radio emission and the large mc to the west of cas a. further evidence of an ongoing interaction scenario may come from changes in this and other nonthermal emission on short timescales . in the mc itself , h@xmath11co absorption studies suggest that certain species are broadened to as much as @xmath207 @xcite . similarly , slight broadening in @xmath0co @xmath208 up to @xmath209 is observed @xcite , while line widths in @xmath0co @xmath2 range from @xmath210 @xcite . thus , evidence for a snr - mc interaction toward cas a is strongly correlated in the x - ray , radio , near - infrared , and sub - millimeter , but the overall result in @xmath0co @xmath2 is subtle . in order to detect bml regions toward similar snr - mc interactions , we must rely on a precise constraint in the velocity - width . optimizing this constraint might be achieved by comparison with narrow molecular emission in the same field . line widths of narrow components in the cas a field range from @xmath211 , or about half of the `` broadened '' line widths . therefore , one criterion for `` broadened '' emission might be : can we divide these emission lines into distinct `` narrow '' and `` broad '' categories ? for example , how sensitive is the identification of bml regions to the velocity - width constraint we applied in ? we illustrate this point in ( right ) where we compare broadened features to unbroadened emission as determined by our analysis . the regions with the broadest @xmath0co line widths are qualitatively different from regions with narrow molecular lines . furthermore , we report the total number of pixels selected by the bml algorithm for newly identified bml regions in , which includes pixels that were ruled out as potential bml regions by the criteria described in . for newly identified bml regions , the algorithm selects these pixels with little contamination from other positions . that is , the total number of bml pixels identified using the criteria described in is generally @xmath212 the number of bml pixels we determine are associated with the snr as described in . we interpret this result to mean that a @xmath168 velocity - width threshold may be close to optimal for identifying bml regions toward snrs and therefore minimizes false positives from spurious detections of , for example , overlapping molecular lines . @xcite compare molecular line studies to other traditional detection methods for snr - mc interactions . these include detection of the oh @xmath5 maser , morphological correspondence between a snr and a mc , molecular line emission with a large high - to - low line ratio such as @xmath0co @xmath213 , detection of near - infrared emission from shock excitation such as [ fe ii ] or h@xmath11 emission , and infrared colors consistent with molecular shocks . it is valuable to utilize as many of these indicators as possible , and in our analysis we rely on morphological correspondence between the snr in radio emission and the mc as well as reports of oh @xmath5 maser emission . however , the detection of broadened @xmath0co line emission is a unique tool for tracing snr - mc interactions in our galaxy . spectroscopic information is vital for identifying the systemic velocity of molecular gas and therefore associating the kinematic distance to molecular emission with distances to snrs . this information reduces the probability of confusion by any chance alignment of a shock tracer with a snr . carbon monoxide is also the second most abundant molecule and possesses one of the highest bond - dissociation energies of any diatomic molecule . therefore , we would expect co to be more easily visible and less susceptible to dissociation behind a snr shock than oh , h@xmath11 or any other molecular shock tracer . finally , broad , low - j co emission is a direct tracer of disturbed gas in a mc , as opposed to line emission from atomic species or near - infrared colors which do not directly or unambiguously trace the molecular gas . probing the mc for molecular shocks and then drawing an association with the snr is well - suited for finding snr - mc interactions , even in a large survey . we applied the bml region algorithm to the @xmath0co @xmath2 spectra ( ) toward all 50 snrs in our sample to determine the location of broadened co features . the algorithm revealed broad - line detections toward nineteen snrs , including nine previously unidentified snr - mc systems . for the other ten systems with coincident bml regions , we discuss our detections toward known snr - mc systems in the context of previous studies in . one of the remaining 31 snrs , [email protected] , had previous indications of a bml region . we argue in that the associated mc may not be velocity - broadened . finally , in , we discuss the results of our algorithm for newly identified snr - mc systems and compare these detections to the observed properties of the snrs themselves . for reference , we include a discussion of previous observations for some remnants in , where we discuss details of the radio continuum , oh maser detections , and pulsar observations . bml regions were detected toward ten snrs with known snr - mc systems . here we compare the characteristics of the detected bml regions to results from previous studies . _ [email protected] _ was observed to have a snr - mc interaction by @xcite . the authors report detection of a small mc in @xmath0co @xmath208 at @xmath214 and coincident with oh maser detections to the south of the remnant . the co emission from this cloud is seen from @xmath215 to @xmath216 . the cloud is small ( @xmath217 m@xmath218 , n@xmath219 @xmath220 ) and has a co line width of only @xmath221 , consistent with the line width of an ambient mc approximately @xmath222 pc in size @xcite . @xcite report detection of two other mcs toward [email protected] , which they call the `` central '' and `` northwestern '' clouds , at systemic velocities between @xmath223 and @xmath224 . the central cloud is located directly to the east of the brightest @xmath225 enhancement in the remnant , implying a possible connection between nonthermal emission and the mc . the reported morphological correspondence , small size of the central cloud ( @xmath226 m@xmath218 ) , and moderately broadened co profile ( @xmath227 ) are all characteristic of a snr - mc interaction . the northwestern cloud appears along the sharp boundary of the snr shock front . while this evidence may indicate physical contact between the snr ejecta and molecular gas , there is little indication in @xmath0co @xmath208 emission to suggest an interaction scenario . the cloud extends to the northwest of the remnant as part of a larger complex , its observed line width ( @xmath228 ) is only slightly larger than the surrounding gas , and no other signs of snr - mc interaction appear in this region . this potential interaction region may be supported by the presence of enhanced tev gamma - ray emission , although the detection is just below the @xmath229 limit traditionally used to identify tev sources ( ) . in @xmath0co @xmath2 , we detect the same molecular emission as seen in @xmath0co @xmath208 emission toward [email protected] at a systemic velocity of @xmath214 . toward the emission that @xcite identify as the `` southern cloud , '' some pixels appear velocity - broadened in @xmath0co @xmath2 to a width of around @xmath230 . this value contrasts with the smaller velocity - width in @xmath208 of @xmath221 and suggests a recent or fast shock may be responsible for the large observed line widths . while this width is broad compared to emission observed in @xmath0co @xmath208 , it is also consistent with other shock indicators . the southern cloud is the site of oh maser emission as well as a significant enhancement in the radio continuum . we do not detect any broadened emission at the positions of the other clouds in this region . _ kes 75 _ was first observed to have a bml region in @xmath0co @xmath208 @xcite . a large mc is coincident with the northern half of the remnant and at a systemic velocity between @xmath231 and @xmath232 . the evidence for interaction is in the broadened blue wing of this feature ( @xmath231 to @xmath233 ) . this asymmetric and broadened feature is spectroscopically similar to the snr - mc interactions seen toward ic 443 and 3c 391 . kes 75 is also the site of a strong tev gamma - ray source @xcite , although this emission is thought to be associated with a pulsar wind nebula in the remnant itself . we confirm the bml detection in @xmath0co @xmath2 around @xmath234 . broadened molecular emission appears extended along the north and west of the remnant and is coincident with the brightest regions of 20 cm continuum emission . this observation is also consistent with detection of broadened emission in @xmath0co @xmath208 where line widths were observed to be on the order of @xmath235 . in our analysis , we find broader line widths in @xmath0co @xmath2 up to @xmath236 . _ 3c 391 _ is discussed in . _ kes 79 _ was observed to have a snr - mc interaction in bright @xmath0co @xmath208 emission to the east and south of the radio shell of the remnant . this morphological correspondence between the snr and mc strongly implies an association between these objects @xcite . the co line profile at these positions is centered around @xmath237 with an integrated intensity of @xmath238 and a peak brightness temperature of @xmath239 . thus , the line has an approximate equivalent width of @xmath240 , consistent with a shock - broadened cloud . however , the observation of @xmath0co @xmath208 was performed with the nrao @xmath241 m telescope and a beamwidth of @xmath242 . the integrated intensity measurements reported in @xcite are ambiguous and may include line emission from multiple , ambient mcs and not a single , shocked cloud . we confirm detection of velocity - broadened @xmath0co emission in @xmath2 from @xmath243 to @xmath244 toward kes 79 . broad line detections extend across the mc from southwest toward the northeast of the remnant with velocity - widths up to @xmath245 . agreement between our @xmath0co measurements and previous observations , as well as the oh measurements at similar velocities ( ) , argues for the presence of a snr - mc interaction toward this remnant . _ 3c 396 _ is observed to have a possible snr - mc interaction in @xmath0co @xmath2 and @xmath208 emission , where a bright cloud appears between @xmath246 and @xmath247 @xcite . at these velocities , the coincident mc extends around the remnant , with emission on the blue end of the line profile concentrated to the north and east and extending clockwise around the remnant toward redshifted emission along the western edge . this detection represents strong morphological agreement between the structure of the mc and the snr . the spectral lines themselves exhibit signs of broadening , with an asymmetric profile and broad wings toward redder velocities in the eastern emission and blueshifted broad wings in the southwestern emission . in some regions , these line profiles have widths @xmath169 . the presence of broadened molecular emission along the western edge of 3c 396 as well as the oh maser at the same velocity argues for the presence of a snr - mc interaction . we detect velocity - braodened @xmath0co @xmath2 emission centered around @xmath248 to the north of the remnant and extending to the west where the broadened emission is centered around @xmath249 . the detected broad line widths are generally between @xmath250 and @xmath251 . these broad - line detections are not bright or significantly enhanced compared to surrounding material . while our analysis supports the presence of a snr - mc interaction , detailed analysis of this region may reveal shocked material that is not significantly distinct from the surrounding gas . evidence for a snr - mc interaction toward _ 3c 397 _ comes from @xmath0co @xmath208 observations performed with the 13.7 m telescope at the purple mountain observatory at delingha @xcite . these data reveal a mc at @xmath252 and to the north and west of the remnant . at these positions , the snr shock appears to be embedded in the mc and morphological correspondence between the @xmath0co and snr radio continuum emission imply the objects are interacting . the co emission in these regions indicates that the mc emits a line profile with broad blue wings ( @xmath253 to @xmath254 ) , consistent with acceleration from a snr - mc interaction . the authors report a fwhm line width of @xmath255 toward the mc at this velocity . we confirm detection of velocity - broadened emission along the western edge of the remnant and centered at a systemic velocity of @xmath254 . the blue wing is not as bright in @xmath0co @xmath2 emission as in the lower - excitation line , although the overall line profile appears to be enhanced to @xmath256 . toward _ w49 _ , evidence for a snr - mc interaction can be seen in @xmath1co @xmath208 emission @xcite . broad lines have been detected at a systemic velocities from @xmath155 to @xmath257 in multiple components and with a narrower line profile centered at @xmath258 . averaged over the entirety of w49b , the line profile has a width of @xmath259 , although the spatially resolved emission is likely narrower . we detect broadened @xmath0co @xmath2 emission toward w49b in the same velocity range and centered at a systemic velocity of @xmath260 . regions of broadened emission are seen to the west of w49b , and the majority of our detections come from a cloud located along the southwestern boundary of the remnant . in this region , we report bml detections with widths of @xmath261 . while this finding agrees with previous detections of broadened molecular emission , some interpretations argue against the presence of a snr - mc interaction to the southwest of w49b . @xcite found that low - frequency radio emission in this region is almost completely attenuated , and the intervening molecular material responsible for absorption may only be seen toward the remnant in projection . however , at higher frequencies w49b becomes significantly brighter . indeed , at 90 , 20 , and 6 cm , the southwestern portions of the shell show significant local enhancement @xcite . added to this fact is the observation that spectral indices appear to flatten to the southwest of w49b . it may be that this region is not enhanced in nonthermal emission . overall , the evidence for the presence of a snr - mc interaction is ambiguous , though our observations support the presence of an interaction scenario . detailed characterization of the physical properties in the gas itself is needed to further verify the presence of a snr - mc interaction . _ cas a _ is discussed in . evidence for a potential snr - mc interaction toward _ hb 3 _ was reported by @xcite , in which the authors perform line mapping and @xmath0co @xmath208 observations . toward this remnant , a bright @xmath0co `` bar '' is coincident with a region of enhanced radio emission near @xmath262 and extending east in galactic coordinates . this morphological correspondence suggests the snr is interacting with the cloud and a region of enhanced particle acceleration is seen in the snr ejecta . evidence from @xcite supports this hypothesis where radio continuum from @xmath263 to @xmath264 is brighter toward the @xmath0co cloud . the spectral indices at these frequencies appear to steepen to @xmath265 . emission from the mc itself extends to the south of the remnant , roughly in the direction of the oh maser detection , and the line profile peaks at a velocity of @xmath266 . however , the w3 region is embedded in a mc at @xmath267 , and it is certain that the @xmath0co `` bar '' at @xmath266 is associated with this cloud . @xcite detect bright @xmath0co @xmath2 and @xmath268 emission centered on w3(oh ) in this same region . the correlation between the region , high line - ratio @xmath0co emission , and 1.1 mm dust continuum from the bolocam galactic plane survey @xcite argues that these bml regions are not associated with hb 3 . we detect @xmath0co @xmath2 emission in this region and centered at a systemic velocity of @xmath269 . our spectral analysis confirms that this emission is broadened , up to @xmath270 , although generally to a velocity width of @xmath271 to @xmath272 in @xmath0co @xmath2 . _ ic 443 _ is a well - observed snr with several indications of a snr - mc interaction . this interaction was first identified in broadened co emission by @xcite and has been confirmed by detection of molecular species toward the associated mcs @xcite , oh @xmath5 emission @xcite , and h@xmath11 emission @xcite . the physical conditions toward ic 443 are well - suited for every shock interaction tracer traditionally used to identify snr - mc interactions . as a result , this region is one of the most luminous molecular hydrogen sources in the galaxy . the interacting mc generally exhibits linewidths of @xmath273 , up to @xmath274 toward the most turbulent molecular gas @xcite . this result has been confirmed in several molecules , probing molecular gas of densities up to @xmath275 @xmath220 @xcite . the snr itself has a composite morphology and appears evolved , with an inferred age of @xmath276 @xcite . the overall radio morphology of the remnant was described by @xcite in @xmath277 continuum emission . two main subshells ( shells a and b ) make up the majority of the remnant in radio continuum emission . shell a appears brighter and is coincident with the molecular shock indicators along its southern boundary and across the center of the remnant as a whole . the western edge of this shell appears to be associated with both the oh maser emission and tev gamma - rays @xcite . given its approximate age and the assumed distance of @xmath278 , ic 443 has a large apparent size on the order of @xmath279 . we targeted a section of the remnant toward its center and along the southwestern edge of shell a. the shocked @xmath0co emission peaks around @xmath280 in this region and extends blueward in a broad , asymmetric line profile . our observations centered on a bright cloud referred to as shock clump b in @xcite . we measured velocity - widths in @xmath0co @xmath2 toward this region ranging from @xmath271 to @xmath281 . [email protected] was targeted with @xmath0co @xmath2 observations in @xcite . a large mc at @xmath282 extends along the north of the remnant , with a bright central core to the northeast of the remnant . the morphological correspondence between the outer blastwave of the snr and the mc is not strong , with at least @xmath283 separation between the 20 cm contours of the remnant and the mc . @xcite report a fwhm of @xmath284 at @xmath282 across the entire cloud . our observations indicate that the broad feature detected around @xmath285 with @xmath286 ( * ? ? ? * see figure 13 , therein ) is at least two clouds blended over the @xmath283 grid spacing applied in this study . while the broad and narrow components of emission at these velocities can be separated from @xmath287 to @xmath288 , the broad component is further separated into two features centered at @xmath289 and @xmath282 ( ) . these components are resolved in our spectra . however , these features may have been smoothed into a single , broad - line component in @xcite given the larger beam ( @xmath290 ) from the 6-m seoul radio astronomy observatory telescope . in contrast to the gaussian component analysis over a large region , our pixel - by - pixel analysis of the molecular emission reveals no indication of @xmath166 velocity - broadening . we detected broadened molecular features toward nine snrs with no previous indication of snr - mc interactions : [email protected] , [email protected] , [email protected] , [email protected] , [email protected] , [email protected] , [email protected] , [email protected] , [email protected] . in this section , we discuss the characteristics of the interaction as identified by our bml region algorithm along with previous work on the associated snrs . our results from the detection of velocity - broadened regions toward each remnant are summarized in . for comparison with each remnant , we use 90 cm radio continuum contours from @xcite ( hereafter , bggkl06 ) where @xmath291 and nvss @xmath292 contours @xcite for the remaining objects . [email protected] was initially identified in 20 cm emission by @xcite and subsequently confirmed by bggkl06 at 90 cm as a small ( @xmath293 ) snr with shell - like morphology . the remnant has flux density @xmath294 at 20 cm , but its morphology suggests an enhancement in surface brightness toward the east . a subsequent search for associated oh maser emission at 1612 , 1665 , 1667 , and @xmath5 with the green bank telescope ( gbt ) yielded a single - dish detection with no velocity information or localization @xcite . of particular note is a tev gamma - ray detection reported toward this remnant @xcite . the gamma - ray emission was observed with a @xmath295 offset from the center of [email protected] and detected with @xmath296 resolution . given the position of the remnant relative to the gamma - ray detection , there is reason to believe that the gamma - ray source originates from the northern half of this snr . in projection , the radio contours of [email protected] appear to flatten toward the northwest ( ) coincident with a mc between @xmath297 and @xmath298 . the spectrum of the molecular emission in this region appears velocity - broadened with @xmath0co gaussian widths fit between @xmath299 and @xmath300 . the coincidence of this broadened @xmath0co along with the morphology of the remnant , the possible oh maser toward this position , and the gamma - ray source associated with this region support the presence of a snr - mc interaction at this location at a systematic velocity near @xmath301 . we note that the velocity centroid of the bml region and the kinematic distance inferred from this velocity , @xmath302 , is also in good agreement with the value given by the @xmath157 relation ( @xmath303 ) . [email protected] has been observed to have typical radio shell - like morphology ( bggkl06 ) . its spectral index between @xmath304 and @xmath305 cm is crab - like ( @xmath306 ) implying nonthermal emission from the snr . there may also be confusion with thermal emission from other sources , which suggests that the snr has a steeper spectral profile in agreement with enhanced particle acceleration . h@xmath307 emission toward the remnant is correlated with nonthermal emission at @xmath292 along the northern edge of the remnant @xcite . this trend implies a connection between regions of strong snr - ism interaction and nonthermal emission from the snr along that edge of the shell . we observe an association between a bright region of 90 cm radio emission and a mc toward the southwest of this remnant ( ) . the molecular emission appears at a systemic velocity of @xmath308 to @xmath309 and extends in a small arc toward the center of the remnant . the region of velocity - broadened emission is concentrated toward a peak in the 90 cm contours , and the molecular emission to the east of this peak exhibits the most velocity - broadening . the @xmath0co gaussian widths of the velocity - broadened emission in this region are fit between @xmath250 and @xmath310 . [email protected] was identified as a snr by @xcite at 408 and @xmath311 . these observations revealed a small remnant ( @xmath312 ) with a steep spectral index of @xmath313 . follow - up with asca indicated there is a 65 ms pulsar toward the center of the remnant @xcite . subsequent radio detection of the remnant in bggkl06 revealed a probable shell - like snr with a flux density at 20 cm of approximately @xmath314 . emission from the remnant at 90 cm appears locally enhanced around the shell , and especially to the east and northwest . detection of oh maser emission has been reported from gbt observations @xcite , however , the lack of precise position and velocity information makes it difficult to associate the oh maser with the snr . there appears to be a strong tev gamma - ray source extended to the northwest of the remnant and hess detects emission toward [email protected] at @xmath315 @xcite . we detect velocity - broadened molecular emission toward this remnant from @xmath316 to @xmath317 . at these velocities , emission extends from the southeast to the north of the remnant , roughly coincident with the bright shell of [email protected] as seen in 90 cm contours ( ) . regions of velocity - broadened emission lie mostly toward the northern molecular emission , with some detections interior to the main radio shell . the gaussian line width of velocity - broadened emission in this region is between @xmath250 and @xmath310 . [email protected] was discovered in bggkl06 as a small , partial shell - like remnant . the radio spectral index from 90 to 20 cm is steep with @xmath318 . the weak flux density at 20 cm combined with the size suggest that the remnant has @xmath319 , and the @xmath157 relationship implies a distance of @xmath320 kpc . @xmath5 oh maser emission has been detected toward this remnant in single - dish observations with no reported velocity information or localization @xcite . we detect velocity - broadened @xmath0co emission coincident with this remnant at a systemic velocity of @xmath321 ( ) . the broadened molecular emission lies toward the eastern shock front of the remannt and to the north of the brightest region in 90 cm continuum emission . this emission is weakly detected in two pixels of our map , and the velocity - width between the two bml detections extends from @xmath250 to @xmath322 . however , the morphology of the coincident mc and snr are suggestive of an interaction at this velocity . as seen in projection , the shape of the snr appears flattened to the east and along the boundary of the snr . this shape contrasts with the northwest edge of the snr where the shell structure appears much fainter and more extended . we infer that the eastern shock front is deccelerated relative to the rest of the remnant , which would require a significant density enhancement in the ism . [email protected] was first confirmed as a snr by bggkl06 in 90 cm continuum emission , although @xcite listed the source g18.6375 - 0.2917 as a `` high - probability supernova remnant candidate '' before this confirmation . subsequent detection at 20 cm identified a partial shell with a spectral index of @xmath323 . the remnant is small ( @xmath324 ) and faint in 90 cm emission ( @xmath325 ) , in agreement with a far distance as determined from the @xmath157 relation ( @xmath326 kpc ) . as shown in , the region of brightest radio continuum emission toward this remnant extends in an arc to the east with fainter lobes of emission to the southwest and north . we find evidence for a snr - mc interaction to the north of the remnant at a systemic velocity of @xmath327 . the mc in this region covers most of the northern and western half of the snr , although the disturbed region appears localized to the north . the velocity - widths of emission lines detected toward this cloud are broad , with most lines exceeding @xmath328 . it has been suggested based on its high infrared to radio luminosity ratio that [email protected] is an region and not a snr @xcite . however , this object has a crab - like spectral index ( @xmath329 ) , which implies some nonthermal component to the emission from this source @xcite . while the radio emission is elongated , it correlates well with the spatial extent of the infrared emission and suggests that the two are associated . this observation is consistent with the possibility that [email protected] is an region and snr projected onto each other . we detect velocity - broadened lines toward the remnant at a systemic velocity of @xmath330 . the regions of broadened molecular emission lie along the brightest part of the remnant s northern shell ( ) . emission in this region appears moderately broadened , and the emission lines have velocity - widths around @xmath331 up to @xmath332 . as seen in projection , the morphology of the snr relative to the mc is highly suggestive . there appears to be a boundary along the northern edge of [email protected] where an ongoing interaction has broadened the coincident mc . [email protected] is often referred to as kes 73 and was identified as a potential snr in part of a larger radio source centered around galactic coordinates @xmath333 by @xcite . the remnant is compact with a central pulsar identified at x - ray wavelengths and an age of @xmath334 @xcite . @xcite were able to localize oh maser emission toward this source @xmath335 west of the snr around @xmath309 . subsequent observations revealed associated @xmath1co @xmath336 and emission toward this source and two nearby regions imply a connection with gas around @xmath196 @xcite . hess detected tev gamma - ray emission toward this remnant at @xmath337 significance @xcite . there appears to be a tev source extending to the south of the remnant , although several other counterparts have been discussed for tev detections in this region . we observe broadening from the same molecular emission toward [email protected] as seen in @xcite , although in @xmath0co @xmath2 the line profile of the broadened emission is centered around @xmath338 with average line widths of @xmath339 ( ) . however , the broadened regions lie at large separations from [email protected] ( @xmath340 ) . it is possible that these bml regions toward [email protected] are associated with other nearby radio sources ( * ? ? * see , e.g. , ) often assumed to be regions . these sources would need to extend to within @xmath341 in projection toward [email protected] to its west and southwest , which is not seen in the radio continuum . a second possibility , which has been raised in the context of cas a @xcite , is that faint , fast - moving ejecta from [email protected] could interact with the molecular gas without being visible in radio emission . if ejected at high - velocity during early phases of sn evolution , this material could carry enough momentum to turbulently accelerate the mc to the velocity width we observe . this hypothesis is supported by the observation that [email protected] is still relatively young and the turbulent clouds are nearby as seen in projection . presumably some fraction of the ejecta has not yet been significantly deccelerated by the surrounding ism and would not need to travel far to shock the observed mcs . [email protected] was discovered in vla observations at 6 and 3.7 cm by @xcite . the radio source is a faint , shell - like remnant , and several local enhancements were detected in the continuum . the spectral index from 6 to 3.7 cm is roughly @xmath342 across the remnant , although the brightest region of 6 cm continuum toward the eastern shell is steeper , with @xmath343 . the northwestern edge of the shell also appears relatively bright , although there is no evidence of steeper nonthermal emission . subsequent observations toward [email protected] with asca revealed a compact x - ray source thought to be a pulsar @xcite . the x - ray source is relatively faint , and no timing or age estimate have been performed , although the coincidence of this source strongly implies that [email protected] is a distant and possibly young snr from a core - collapse sn . we detected broadened @xmath0co emission toward [email protected] and roughly coincident with a region of bright 20 cm continuum emission , at a velocity of @xmath344 ( ) . this position is consistent with enhancements in the 6 cm continuum and spectral index , as well as a region of bright x - ray emission as mapped by @xcite . the majority of the molecular emission at these velocities coincides with the remnant , implying a strong morphological association between the snr and mcs . a weak red wing is also visible in the @xmath0co spectrum demonstrating the characteristic velocity profile of shocked - broadened molecular emission . the two pixels where we find evidence for velocity - broadened molecular emission are highly disturbed , with line widths @xmath345 . [email protected] was discovered in the nrao / vla sky survey ( nvss ) at @xmath292 @xcite . the remnant has clear shell - like structure visible in both radio and x - ray wavelengths @xcite . while the morphology of the remnant and its x - ray spectrum imply a strong synchrotron component , a radio spectral index has not yet been measured . of particular note are two lobes seen in nonthermal radio and x - ray emission along the eastern edge of the remnant @xcite . these features appear to be local enhancements in nonthermal emission ( ) . moreover , they are adjacent to regions where the radio shell appears to break and little or no emission can be seen . this morphology contrasts with the western edge where the radio shell appears more contiguous albeit no brighter than the eastern lobes . we detect broadened molecular emission toward a cloud at a systemic velocity of @xmath346 and roughly coincident with the eastern radio lobe of [email protected] . the associated mc extends along the southeastern and eastern part of the shell , although we detect velocity - broadening only along the northernmost portion of this cloud where the radio continuum of [email protected] is brightest . the gaussian velocity - width of molecular emission in this region ranges from @xmath347 to @xmath348 . the association with the nonthermal enhancement toward [email protected] is highly suggestive , and the spatial correlation between the mc and the shell of the remnant imply an interaction scenario . however , the kinematic distance inferred from this association is in tension with the values inferred from the @xmath157 relation ( @xmath349 ) and x - ray absorption ( @xmath350 ) @xcite . the @xmath157 estimate may be affected by the low flux density of [email protected] ( @xmath351 from @xcite ) , which renders surface brightness and size estimates uncertain . finally , it may be that the density of hydrogen toward this source is significantly larger than the @xmath352 assumed for the latter estimate . @xcite argued that , because half of all snrs are expected to result from type ii sne and their progenitors are young , massive stars born in large mc complexes , it is expected that roughly half of snrs should be in contact with mcs . new information on the incidence of core - collapse sne places additional constraints on the expected rate of this type of snr - mc interaction . here , we discuss the total number of snr - mc interactions in our galaxy as a fraction of all known galactic snrs , which we call the `` snr - mc interaction rate . '' we examine this value in light of our findings and the factors that may contribute to an enhanced or reduced rate . recent transient surveys have greatly improved our understanding of the intrinsic fraction of both thermonuclear and core - collapse sn types in volume - limited samples . for example , the lick observatory supernova search ( loss ) observed 175 sne between @xmath353 and @xmath354 mpc and determined that @xmath355 of sne are type ii , another @xmath356 are type ib / c and the remaining @xmath357 are type ia @xcite , implying that roughly @xmath358 of snrs come from core - collapse sne . we might expect the snr - mc interaction rate to be close to this value . however , discounting the four objects in our sample mentioned in , our fraction of candidate interactions ( 17/46 or @xmath359 ) is much lower than the observed fraction of core - collapse sne . if we rule out possible type ia sne by looking only at snrs with known pulsars or compact x - ray sources , which we infer resulted from core - collapse sne , the total rate appears to change somewhat ( 7/15 or @xmath360 ) . we show in below that virtually all snr - mc interactions identified by other means also show broad co and will be identified by our approach . however , in we illustrated the potential for false identifications . therefore , 7/46 and 7/15 should be interpreted as upper limits . the difference between these two rates also supports the hypothesis that snr - mc interactions tend to be associated with core - collapse sne , but both values are low compared to the expected rate . there appears to be an underlying physical mechanism that suppresses the number of snrs that interact with mcs compared to the known fraction of core - collapse sne or the number of snr - mc interactions detectable in co lines . in , we discuss the possibility of ongoing or recent interactions between snrs and mcs that do not excite broadened @xmath0co @xmath2 emission . we examine the possibility that some underlying physical mechanism suppresses the snr - mc interaction rate in and in we discuss the implications of a low snr - mc interaction rate for sequential star formation and simulations of sn feedback . finally in , we analyze the correlation between snr - mc interactions and tev gamma - ray emission from snrs . one of the most fundamental questions in our @xmath0co @xmath2 analysis of bml regions toward snrs is what fraction of snr - mc interactions are detectable in molecular emission ? @xcite argue that oh maser detections are a reliable ( if incomplete ) signpost of snr - mc interactions , so a test of the completeness of @xmath0co @xmath2 line broadening detections for snr - mc interactions might be `` what fraction of snrs with nearby oh maser emission have also been detected in broad - line molecular emission ? '' in , we list remnants with oh maser detections and whether molecular line studies , where they exist , support the presence of broadened molecular lines . lllc [email protected] & cs ( 1@xmath30 , 5@xmath34 , 7@xmath36 ) & y & 1,2 + [email protected] & & & + [email protected] & & & + [email protected] & co ( 1@xmath30 ) & n & 3 + [email protected] & co ( 1@xmath30 ) & n & 3 + [email protected] & co ( 1@xmath30 , 3@xmath32 ) & y & 4 + [email protected] & & & + [email protected] & & & + [email protected] & co ( 1@xmath30 , 2@xmath31 ) & y & 5,6 + [email protected] & co ( 1@xmath30 ) , @xmath1co ( 1@xmath30 ) , c@xmath361o ( 1@xmath30 ) , & y & 7 + & hco@xmath362 ( 1@xmath30 ) & & + [email protected] & co ( 2@xmath31 ) & y & 5 + [email protected] & co ( 2@xmath31 ) , cs ( 2@xmath31 , 3@xmath32 , 5@xmath34 ) , & y & 5,8 + & hco@xmath362 ( 1@xmath30 ) & & + [email protected] & co ( 1@xmath30 , 2@xmath31 ) @xmath1co ( 1@xmath30 ) & y & 9 + [email protected] & co ( 2@xmath31 ) & y & 10 + [email protected] & co ( 1@xmath30 , 2@xmath31 ) , @xmath1co ( 1@xmath30 ) , & y & 11 + & hco@xmath362 ( 1@xmath30 ) & & + [email protected] & co ( 1@xmath30 , 2@xmath31 ) & y & 5,12 + [email protected] & & & + [email protected] & & & + [email protected] & & & + [email protected] & & & + [email protected] & co ( 1@xmath30 ) & y & 6 + [email protected] & co ( 1@xmath30 ) & y & 13 + [email protected] & co ( 1@xmath30 , 2@xmath31 , 4@xmath33 ) , & y & 14 + & @xmath1co ( 1@xmath30 , 2@xmath31 ) , cs ( 2@xmath31 , 3@xmath32 ) , & & + & hco@xmath362 ( 1@xmath30 ) , hcn ( 1@xmath30 ) , & & + & h@xmath11co ( 3@xmath363@xmath32@xmath364 , 3@xmath365@xmath32@xmath366 ) & & + [email protected] & & & + [email protected] & co ( 1@xmath30 , 2@xmath31 ) , @xmath1co ( 1@xmath30 , 2@xmath31 ) , & ? & 15 + & c@xmath361o ( 1@xmath30 , 2@xmath31 ) , cs ( 2@xmath31 , 3@xmath32 ) , & & + & hco@xmath362 ( 1@xmath30 ) , hcn ( 1@xmath30 ) , & & + & h@xmath11co ( 3@xmath363@xmath32@xmath364 , 3@xmath365@xmath32@xmath366 ) , & & + & sio ( 2@xmath31 , 5@xmath34 ) & & while all of the studies cited in detect molecular features at roughly the same systemic velocity as the coincident oh maser emission , the incidence of broad - line detections is 13/15 or @xmath367 . another study is inconclusive : @xcite confirm the presence of shocked molecular hydrogen emission and broad - line absorption toward [email protected] but their results are ambiguous for the co lines , so we have not included this object . molecular line broadening toward shocked mcs may not be detected for several reasons . molecular gas from a single cloud can be optically thick and obscure the shocked regions . the faint line wings characteristic of broadened molecular lines may also be obscured by bright , narrow emission from other clouds . these effects are usually more severe in lower energy transitions where line emission from quiescent clouds is intrinsically broad . indeed , both of the bml non - detections in were performed in @xmath0co @xmath208 and toward snrs near the galactic center ( g5.4 - 1.2 , g5.7 - 0.0 ) @xcite . observations in higher energy transitions might show less emission from quiescent clouds along the line - of - sight while revealing broad - line emission . every observation we report that was performed in transitions at higher energies than @xmath0co @xmath208 has found signs of broadened molecular emission . we infer that broad molecular line emission is a highly complete ( e.g. , @xmath368 ) tracer of snr - mc interactions , and @xmath0co @xmath2 may exhibit broadened molecular lines in almost all instances of a snr - shocked mc . detection of broadened @xmath0co @xmath2 lines may be a necessary , but not sufficient , condition for the presence of snr - mc interactions . this hypothesis would imply the limiting factor in assessing the `` true '' number of snr - mc interactions in @xmath0co @xmath2 is not false negatives from optically thick or obscured emission , but false positives from multiple mcs in velocity - space that give the appearance of a broad line or from bml regions toward other sources such as regions . we report evidence of both effects , for example toward hb 3 where the coincident bml region appears to be associated with the nearby region w3(oh ) , [email protected] where the line emission from multiple velocity components appears to have been mistaken for broad - line emission in previous work , and [email protected] where bml regions are detected at multiple velocities implying contribution from other sources . given the high completeness we infer for detections of snr - mc interactions observed in @xmath0co @xmath2 , the ratio of bml regions we detected to observed snrs ( i.e. , @xmath359 ) may only be an upper limit to the `` true '' snr - mc interaction rate in the galaxy . a lower snr - mc interaction rate only exacerbates the discrepancy with the fraction of core - collapse sne . we infer that some additional mechanism is required in order to suppress the number of snr - mc interactions occurring in the galaxy , and we explore possibilities and their consequences below . suppression of the snr - mc interaction rate relies on a sn event that is too distant from a mc for interaction to occur on a short timescale . for core - collapse events , we propose that there are two main mechanisms by which suppression can occur . one possibility is that massive stars are kicked or migrate away from their parent mc such that they are too distant to cause a snr - mc interaction . the second is a delay between the sn event and snr - mc interaction , which we can express as some characteristic timescale or , equivalently , a characteristic separation between massive stars and mcs . these mechanisms are not mutually exclusive , and indeed , the former might provide an explanation for the delay between sn explosions and snr - mc interactions . the evolution of the snr - mc interaction rate with age provides a way of disentangling each possibility and its underlying cause . for example , one explanation of the low snr - mc interaction rate may be a large , distinct population of runaway massive stars . @xcite detected 2547 runaway star candidates in hipparcos , including over 200 stars with peculiar velocities from @xmath369 toward ob associations . for a star migrating directly away from its parent cloud at @xmath370 over @xmath371 yr , a subsequent core - collapse sn would explode roughly @xmath372 pc from its starting position . thus , the time between explosion and snr - mc interaction would be delayed by approximately @xmath373 ( i.e. , the age at which a snr has a radius of @xmath374 ) . age is a critical factor in the snr - mc interaction rate . we hypothesize that the snr - mc interaction rate for snrs of varying ages may be solely a function of the `` delay '' between the explosion and interaction . this supposition is supported by the observation that the incidence of runaway stars is appreciable compared to the overall population of young , nearby stars . @xcite found that the runaway frequency of ob stars is @xmath375 for their sample . this population of stars may explode as sne and interact with mcs only as evolved snrs or never exhibit signatures of snr - mc interactions . assuming massive stars are `` kicked '' or otherwise migrate a large distance from their parent mc , an analysis of the snr - mc interaction rate with snr age would provide a constraint on the delay caused by this migration . a constant snr - mc interaction rate with age would suggest that there are two distinct populations of massive stars ; those that remain close to their parent mc and those that migrate a large distance such that all known galactic snrs are too young for the latter to have encountered a mc . the more likely scenario is that the snr - mc interaction rate increases with snr age , which suggests that massive stars migrate a short distance ( i.e. , @xmath376 ) from their parent mc over their lifetimes . we would expect the snr - mc interaction rate to level off at a certain snr age . this analysis may provide a constraint on the distribution of massive star runaway velocities and the fraction of massive stars that are runaways . we note several snrs with age determinations from free - expansion velocity or pulsar timing ( ) . however , the number of such snrs is too low to draw any statistically meaningful conclusions . as more snr - mc interactions are identified , accurate age determinatons may be obtained from kinematic distances and modeling of snr size , morphology , and surface brightness . @xcite provides an analytic overview of age determinations that accounts for sn parameters as well as momentum and radiative loss . once ages have been determined for a sufficient number of snrs , an estimate of the dependence of the snr - mc interaction rate with age should be possible . another rich field of study is the direct impact snrs may have on star formation . models of sequential or triggered star formation rely on an understanding of the feedback between massive stars and star forming regions . while the relationship between ob associations and star - forming regions has long been recognized @xcite , the contribution of shocks from snrs is more ambiguous . some studies have proposed that these shocks will compress molecular regions and accelerate triggered star formation @xcite , although it is possible the shock velocities involved are dissociative and will destroy the molecular gas instead . there is also evidence that snrs play little or no role in regulating star formation in individual star clusters due to the time delay before sne occur , but turbulent outflows powered by sne may suppress star formation in giant mcs or even galaxy - wide star formation @xcite . recent numerical simulations of sn feedback have examined the impact sn feedback has on galactic environments in order to evaluate the importance of snrs in star formation ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? many of these simulations fail to produce the observed stellar masses due to excessive cooling or the lack of momentum transfer from sne to giant mcs . this problem may be exacerbated by a large fraction of sne that explode far from any molecular gas . over short timescales , a large fraction of sne may play no role in galaxy - wide feedback . in the absence of momentum transfer from snrs , it is possible that stellar feedback mechanisms other than sne are dominant . for example , regions are known to photoionize dense mcs and may suppress star formation where a sufficient level of background radiation is present . this hypothesis is supported by observations of giant regions ( such as 30 doradus in the lmc ) indicating that radiation pressure from stars is the dominant mechanism in adding energy and momentum to the surrounding region @xcite . the relative role of each mechanism must be assessed in the context of specific timescales and physical conditions . our observations support arguments that snrs may contribute less to stellar feedback than previously thought . particle acceleration is one of the most important galactic processes in which snrs play a central role . cosmic rays are produced in snrs where the sn explosion can deliver enough energy to rapidly accelerate nuclear particles @xcite , and the energy density of cosmic rays produced in the shock region increases with the density of the surrounding medium @xcite . enhanced turbulence and magnetic fields can excite significant hadronic particle acceleration which produces @xmath377 mesons that decay immediately into gamma - rays . consistent with this hypothesis , there are a number of instances where tev gamma - rays are reported from regions of interaction between snrs and mcs ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? a challenge to such studies is the complex and dense population of mcs and the high density of tev gamma - ray sources in the inner milky way where most known tev sources lie . there may also be a tendency to target regions of possible interaction in searching for gamma - ray sources , potentially introducing a bias . these issues can be mitigated in our survey . we use the positions of the specific snrs targeted in the co survey as priors to identify possible tev counterparts , which we accept if their positions coincide with the snrs within @xmath378 . this positional tolerance was adopted from a conservative estimate of the positional accuracy of @xmath379 for the tev measurements ( e.g. , * ? ? ? * felix aharonian , private communication ) , plus an allowance for the extent of the snr . we can then compare the number of detections for the snrs where we detect broad co lines indicative of snr - mc interactions with the number where we do not find broad lines . we adopt the conventional threshold of @xmath380 for a tev detection . @xcite has evaluated the hess milky way survey ( the source of nearly all the tev detections we use ) and shown that all of the sources at this detection level have been identified , that is , the sample is complete to this level of significance . we correct the detection rates for chance coincidences by identifying tev sources with the same positional tolerance relative to 120 positions selected arbitrarily in the same region . where there were possible identifications , we report the separation between the approximate location of the snr - mc interaction region or the center of the snr for remnants without snr - mc interactions and the nearest tev gamma - ray source in . we have taken positions for the tev sources from tevcat @xcite . there are nine cases within our positional criteria , four with indications of mc interactions and five without . our test on arbitrary positions indicates that there are about two false identifications , one for the cases with interactions and one for those without . that is , there are three true cases coinciding with interactions and four not coinciding . we have estimated the number of snrs with tev measurements from @xcite , as shown in . the two sample sizes are virtually the same ( 19 without interactions , 20 with ) , so the rates of detections are identical within statistical errors . lcccccl [email protected] & @xmath381 & @xmath382 & d & 5.1 & hess j1804 - 216 & @xcite + [email protected] & @xmath383 & @xmath384 & d & 22 & hess j1809 - 193 & @xcite + [email protected] & @xmath35 & @xmath385 & & 1.4 & hess j1813 - 178 & @xcite + [email protected] & @xmath46 & @xmath386 & & 57 & hess j1825 - 137 & @xcite + [email protected] & @xmath48 & @xmath387 & & 42 & hess j1825 - 137 & @xcite + [email protected] & @xmath54 & @xmath388 & & 17 & hess j1831 - 098 & @xcite + [email protected] & @xmath389 & @xmath390 & d & 39 & hess j1841 - 055 & @xcite + [email protected] & @xmath391 & @xmath392 & d & 1.5 & hess j1846 - 029 & @xcite + [email protected] & @xmath104 & @xmath393 & & 0.1 & ver j2016 + 371 & @xcite + [email protected] & @xmath394 & @xmath395 & d & 7.9 & tev j2323 + 588 & @xcite + [email protected] & @xmath118 & @xmath396 & & 2.1 & tev j0025 + 641 & @xcite + [email protected] & @xmath120 & @xmath397 & & 2.0 & tev j0209 + 648 & @xcite + [email protected] & @xmath124 & @xmath398 & & 0.2 & tev j0534 + 220 & @xcite + [email protected] & @xmath399 & @xmath400 & d & 8.4 & tev j0616 + 255 & @xcite the similar number of associations between snrs and gamma - ray emission in these two samples does not support a strong correlation between snr - mc interactions and tev gamma - ray emission . the observed gamma - ray detections may also be contaminated by particle acceleration in pulsars , which are known gamma - ray sources . for example , the crab pulsar has long been associated with hadronic particle acceleration and tev gamma - ray emission @xcite . tev gamma - ray surveys typically distinguish between `` compact '' and `` extended'' gamma - ray sources , which may aid in resolving the ambiguity between gamma - rays from pulsars and more extended regions along the snr shell . however , this distinction may fail to separate pulsar point sources from snr - mc interactions where the conditions for particle acceleration are met only in a small region . moreover , clouds are observed near snrs with a high enough density gradient to excite particle acceleration at shock fronts @xcite . in the absence of a pulsar , it may not be necessary to have a snr - mc interaction to create a gamma - ray source toward a snr , although the contact discontinuity between snr ejecta and the densest parts of the ism would still be expected to produce the brightest and hardest gamma - ray sources . we can account for these uncertainties to an extent . ten out of the fourteen snrs we report with tev gamma - ray emission at @xmath229 significance have pulsars or coincident , compact x - ray sources , including five of the six objects with bml regions and five of the eight without . removing these sources leaves only [email protected] and tycho as the two objects with tev gamma - ray sources detected at close separations , and while we report a coincident bml region toward the former , we do not detect any strong @xmath0co emission toward the latter . [email protected] is a recently discovered object ( ) and it is possible that an undiscovered pulsar exists in the snr and provides the particle acceleration necessary to generate tev gamma - rays . @xcite discuss several possible sources for the gamma - ray detection in this region , including the pulsar psr b1800 - 21 , although this source appears to be separated from the gamma - ray emission by at least @xmath378 . barring these possibilities , [email protected] represents the best candidate in our sample for tev gamma - ray emission from a snr - mc interaction . in addition to [email protected] , compelling evidence for tev gamma - ray emission can be seen toward a number of snr - mc interactions , notably w28 @xcite and ic 443 @xcite . although these cases show that snr - mc interactions can produce tev sources in some cases , the lack of a strong correlation in our study shows that other processes must play an important , perhaps dominant , role . larger statistical samples are needed to evaluate the association of tev sources with snr - mc interactions more definitively . @xcite lists all measurements of snrs , detections and upper limits from hess . out of the nineteen snrs we observed with evidence for snr - mc interaction , twelve also have evidence for gamma - ray emission at the @xmath401 significance level including six of the nine snrs with newly reported interactions . this fraction is appreciably higher than the overall incidence of gamma - ray emission toward snrs , which is 24/50 in our sample . previous systematic studies for gamma - ray emission toward snrs have found an even lower ratio of snrs with coincident tev gamma - ray emission ( 1/3 in * ? ? ? * ) . at this level , our findings seem to support the hypothesis that tev gamma - ray emission is correlated with snr - mc interactions . however , confirmation of gamma - ray emission at this level is problematic . detections at @xmath401 may be close to the confusion limit with background sources along the galactic plane . in addition , the diffuse emission of the galaxy can contribute to such detections without being associated with a specific source . combined with the positional uncertainty in gamma - ray experiments , the correlation between snr - mc interactions and tev sources at @xmath401 is tantalizing but not necessarily reliable . we conclude that there is no convincing evidence with the existing data that snr - mc interactions are a dominant source of tev gamma - ray emission from snrs . other sources of particle acceleration in snrs , especially pulsars and interaction with the gas in the cold neutral medium , may contribute significantly to the number of gamma - ray detections . our survey of @xmath0co @xmath2 molecular spectroscopy toward 50 snrs is the first large , systematic search for this shock tracer toward snrs . we demonstrate that our technique for analyzing @xmath0co @xmath2 data cubes satisfactorily reproduces broad - line detections in systems with known bml features . in applying this technique to our data set , we have found : 1 . bml features were detected toward nine remnants ( [email protected] , kes 75 , 3c 391 , kes 79 , 3c 396 , 3c 397 , w49b , cas a , hb 3 , ic 443 ) with previous evidence of snr - mc interactions and nine remnants ( [email protected] , [email protected] , [email protected] , [email protected] , [email protected] , [email protected] , [email protected] , [email protected] , [email protected] ) with no previous evidence of interaction . the bml region detected toward hb 3 is likely associated with the region w3(oh ) . we are uncertain whether the bml regions associated with 3c 396 and w49b represent snr - mc interactions , and detailed follow - up toward these clouds is required to draw any firm conclusions . for one of the remaining remnants , [email protected] , we find that previous indications of a snr - mc interaction may have been due to the blending of multiple narrow line components . 2 . from our analysis of snrs with signs of molecular shocks from oh masers , we infer broadened @xmath0co @xmath2 features are a necessary but not sufficient condition to determine the presence of snr - mc interactions . confirmation must come from other sources , such as @xmath0co line ratios , to demonstrate that the gas in bml regions is physically distinct from the surrounding material . the total number of snr - mc interactions we observed as a fraction of observed snrs may only be an upper limit to the `` true '' snr - mc interaction rate in the galaxy . the fraction of detected bml regions toward snrs with known pulsars or compact x - ray sources ( 7/15 ) is slightly larger than the overall number ( 17/46 ) , supporting the hypothesis that snr - mc interactions are associated with core - collapse sne from massive stars . however , the number of bml regions we detect toward snrs is significantly lower than the incidence of core - collapse sne . we infer that age may be a critical factor in the overall rate of snr - mc interactions . a low snr - mc interaction rate may also have far - reaching consequences for models of stellar feedback and sequential star formation . 4 . we found no convincing evidence for a correlation between tev gamma - ray sources and snr - mc interactions . the number of tev detections within @xmath378 of snrs with bmls is the same as the number for snrs with no evidence for interactions . we conclude that snr - mc interactions are not a dominant source of tev gamma - rays in the galaxy . as an alternative , pulsars or interaction with gas in the cold neutral medium may contribute significantly to the detection rates for tev gamma - rays toward snrs . the heinrich hertz submillimeter telescope is operated by the arizona radio observatory with partial support from national science foundation grant ast 1140030 . funding for this research was provided by nasa through contract number 1255094 issued by jpl / caltech . we would like to express our gratitude to crystal brogan for providing her data from bggkl06 for remnants used in this study as well as david moffett for providing 20 cm radio continuum data of 3c 391 . we also thank joachim hahn for his helpful comments . kepler ( [email protected] ) is a young snr whose proper motion is fast enough to be measured . from _ acis - s measurements of kepler s expansion and a shock velocity inferred from x - ray synchrotron emission , @xcite determined that kepler is located at a distance of @xmath402 kpc . [email protected] was observed in neutral hydrogen by @xcite who found that weak absorption can be seen up to @xmath403 and roughly consistent with a distance of @xmath404 kpc . [email protected] has an association with stellar cluster cl @xmath405 whose distance is known to be @xmath406 kpc @xcite . [email protected] does not yet have a trustworthy distance estimate , although multiple @xmath138-d relationships have been calculated for this remnant . @xcite calculated the distance to be @xmath407 kpc ( updated to @xmath408 kpc for the correct galactocentric distance as in * ? ? ? @xcite use the @xmath138-d relationship and place the remnant slightly further at @xmath409 kpc . we use @xmath408 kpc in our analysis for consistency with previous studies . [email protected] has an associated pulsar with high galactic column density along its line - of - sight @xcite , roughly consistent with a distance of @xmath372 kpc . it has been argued that the complex of regions near [email protected] implies an association with the snr . several studies have used absorption estimates and inferred a distance of @xmath410 kpc to [email protected] ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? however , new absorption estimates to the remnant itself imply this assumption may be in error and the remnant is located beyond the regions at @xmath411 kpc @xcite , which we use in our analysis . [email protected] has been identified from absorption between systemic velocities @xmath412 and @xmath248 for a distance of @xmath406 kpc @xcite . radio recombination lines at systemic velocities from @xmath413 to @xmath414 provide kinematic constraints on the distance to [email protected] @xcite . these values , along with the distance to the nearby pulsar ax j1845 - 0258 @xcite , led @xcite to estimate that [email protected] is at a distance of @xmath372 kpc . @xcite measured the distance to [email protected] ( 4c-04.71 ) in absorption to be @xmath408 kpc . based on x - ray absorption measurements , @xcite determined that the distance to [email protected] is @xmath415 kpc ( i.e. , @xmath416 kpc ) . [email protected] ( kes 75 ) has several distance estimates derived from absorption , separate @xmath157 relationships , and co associations ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? we adopt the distance @xmath417 kpc , derived by @xcite from absorption studies . @xcite derived a distance of @xmath418 kpc to [email protected] based on @xmath157 estimates , and this value has been used in previous studies of this remnant ( e.g. , * ? ? ? * ) . we adopt this distance in our study . the distance to [email protected] was measured in x - ray absorption by @xcite , who determined the remnant lies at @xmath419 kpc . @xcite both measured the distance to [email protected] ( kes 79 ) kinematically from and oh absorption , respectively . their measurements both agree with a distance @xmath347 kpc . the distance to 3c 396 has been derived in several ways , sometimes with discrepant results . @xcite use absorption to determine 3c 396 is @xmath420 kpc , while x - ray column density and co associations indicate @xmath421 kpc @xcite . we adopt the latest kinematic estimate of @xmath422 kpc as in @xcite . toward 3c 397 , absorption indicates a distance of at least @xmath423 kpc while kinematic estimates from co give @xmath424 kpc @xcite . we use this estimate in our analysis . enhanced absorption toward w49b suggests the distance is @xmath425 kpc @xcite , from which we adopt the lower estimate of @xmath426 kpc . @xcite determine the distance to [email protected] from morphological association with co around @xmath427 to be @xmath422 kpc . optical observations of [email protected] reveal line emission in h@xmath307 , h@xmath142 , [ n ii ] , [ s ii ] , and [ o iii ] @xcite . these data suggest that the snr has a large angular size ( @xmath428 ) and slow shocks ( @xmath429 ) . [email protected] is also obscured by a column density of @xmath430 @xmath431 . for a galactic longitude of @xmath432 and a galactocentric distance of the sun of @xmath433 kpc , the minimum galactocentric distance along a line - of - sight near the mid - plane is approximately @xmath434 kpc . we can apply an estimate of the distribution of neutral hydrogen along the line - of - sight to [email protected] determined in @xcite to infer a distance to this snr . for @xmath435 @xmath220 , @xmath436 kpc and assuming the galactocentric distance is @xmath434 kpc @xmath437 kpc , the average mid - plane volume density of neutral hydrogen is @xmath438 the distance to which the neutral hydrogen density profile integrates to the measured value of @xmath439 toward [email protected] is @xmath440 kpc , which we use as the distance to this snr . associated and co features toward [email protected] yield a distance of @xmath441 kpc @xcite . the distance to [email protected] ( ctb 80 ) has been determined from emission to be approximately @xmath442 kpc @xcite . absorption and a possible co association toward [email protected] ( ctb 87 ) yield a distance estimate of @xmath443 kpc @xcite . [email protected] has an associated pulsar whose coordinates and spectra best place the snr in the outer arm at @xmath444 kpc @xcite . @xcite determined the distance to [email protected] from x - ray absorption to be @xmath445 kpc . proper motion measurements of cas a along with spectroscopic measurements of the expansion velocity suggest the remnant is located at a distance of @xmath446 kpc @xcite . @xcite determined the distance to ctb 1 kinematically from optical spectroscopy to be @xmath447 kpc . the distance to tycho has historically been controversial ( e.g. , * ? ? ? measurements of absorption at the near and far side of tycho suggest the remnant is located at @xmath448 2.53.0 kpc @xcite while x - ray proper motion studies suggest @xmath449 kpc @xcite . we adopt the near distance estimate @xmath450 kpc in agreement with models of the shock wave @xcite and measurements of a reported companion star @xcite . several observations of [email protected] ( 3c 58 ) in absorption indicate a distance of @xmath451 kpc @xcite , from which we adopt the latest estimate of @xmath199 kpc . as we indicate in , hb 3 is located near w3(oh ) in the w3 star - forming region . @xcite measure the distance to w3(oh ) kinematically from water masers and find it is located at @xmath452 kpc . proper motions of ejecta toward [email protected] ( crab / crab nebula ) in optical and radio wavelengths yield a distance of @xmath199 kpc @xcite . @xcite constrained the distance to ic 443 by examining stars whose spectra indicated na and ca absorption from the remnant . these stars constrained bracketed the distance at @xmath442 kpc . the remaining three snrs [email protected] , [email protected] , and [email protected] have no distance estimates and either have no known type or are classified as having composite morphology ( ) . however , the targets have observed sizes and @xmath204 surface brightnesses , to which we can apply the @xmath137 relation and obtain a distance approximation . we acknowledge that this relationship should not be applied to remnants that are not shell - like unless the shell structure is well - resolved or bright enough that all other emission can be ignored . therefore , these distance values should not be interpreted as a measurement to each snr but a starting point we calculate for the purposes of our analysis . using the relationship calculated in @xcite and the angular size and @xmath453 flux in @xcite , we find distances of @xmath454 kpc to [email protected] , @xmath455 kpc to [email protected] , and @xmath456 kpc to [email protected] . _ [email protected] _ is a composite remnant initially identified at @xmath225 in @xcite . the remnant has a steep spectral index between @xmath292 and @xmath225 of @xmath457 . multiple oh @xmath5 maser searches have revealed a source along the southern half of the remnant , consistent with one of the brightest regions of the snr at @xmath225 @xcite . this maser emission occurs at @xmath257 , and a tentative detection of oh ( 1667 mhz ) absorption has also been reported at @xmath214 . _ kes 75 _ was initially identified as a snr ( referred to as g@xmath458 ) between 408 and @xmath311 by its spectral index @xcite . the spectral index measured in that study demonstrated nonthermal emission with @xmath459 . low - frequency observations of this snr are dominated by the outer shell where the emission demonstrates has an even steeper spectral index , up to @xmath460 @xcite . while the overall morphology appears to be composite , the steep spectral indices , even at high frequencies , suggests that the shell - like structure may be the dominant source of nonthermal emission . an x - ray pulsar was identified toward kes 75 by @xcite . timing measurements suggest that the pulsar is relatively young at @xmath461 . _ kes 79 _ was first confirmed as a snr at 408 and @xmath311 by @xcite . subsequent pulsar detection and x - ray timing observations were performed with xmm - newton @xcite . these observations confirmed the existence of a 105 ms x - ray pulsar toward kes 79 ; however , the characteristic age was determined to be @xmath462 . this value is significantly larger than the age estimated in @xcite , where the authors identified a bright inner ring of material at @xmath463 as a wind - driven shell and , assuming this material is expanding outwards at @xmath464 , inferred an age of @xmath465 yr . thus , while kes 79 is likely a shell - like snr with an embedded pulsar , the timing measurements imply the pulsar was spun up at birth and the younger age is more accurate . oh lines at 1666 , 1667 , and @xmath5 have been seen in both emission and absorption toward kes 79 @xcite . the oh @xmath5 maser was observed at @xmath466 in a single - dish detection , although the source has yet to be localized . in absorption , oh has been observed at @xmath467 and @xmath468 and in a broad feature from @xmath469 to @xmath470 . it is inferred that this last feature corresponds to a coincident mc , and the broadening in oh is due to a shock interaction with the snr . _ 3c 396 _ was classified as a snr by @xcite . 20 cm emission maps reveal composite morphology with significant brightening along the western half of the remnant , and especially toward the southwest @xcite . the spectrum at this frequency appears nonthermal ( @xmath471 ) , and measurements of the spectral index at 90 cm agree with this value @xcite . @xcite reported detection of oh @xmath5 maser emission at more than one pointing in single - dish observations of the remnant , although no velocity information or localization was obtained from these observations . this detection was subsequently confirmed by @xcite in vla observations , and the authors report the presence of a @xmath472 oh line . _ 3c 397 _ was resolved and distinguished from a nearby region to its west by @xcite . the remnant is bright and compact , with several enhancements in the nonthermal radio emission to its west . the spectral index is steep from 6 to 20 cm in this region ( @xmath473 ) relative to the rest of the shell ( @xmath474 ) @xcite . observations at @xmath5 revealed no oh maser emission and imply that if there is a snr - mc interaction the physical conditions may be too extreme to excite a maser @xcite . _ w49b _ was initially identified as a strong radio source and probable snr at @xmath475 by @xcite . the remnant radio spectrum has a moderately nonthermal spectral index between @xmath475 and 1@xmath225 of @xmath476 @xcite . thus far , no oh maser emission has been detected toward this remnant in targeted searches ( e.g. , * ? ? ? [email protected] _ is a diffuse , low surface brightness radio source identified as a snr in @xcite . the source is significantly polarized ( @xmath477 ) at 4.7@xmath225 , but has a flat spectral index at this frequency of @xmath478 . subsequent observations in the x - ray with asca and at @xmath479 with arecibo revealed a 136 ms pulsar , which combined with the snr morphology , suggests that [email protected] is a crab - like remnant with an age of @xmath480 @xcite . _ hb 3 _ is a large snr with well - defined shell structure at @xmath481 @xcite . the remnant is embedded in a star - forming region that contains the nearby region w3 . spectral indices from @xmath482 to @xmath481 ( @xmath342 ) suggest hb 3 has a strong nonthermal component . there is also evidence for a snr - mc interaction from several oh @xmath5 masers that have been detected to the southeast of this snr and near w3 @xcite . one of these detections , referred to as `` w3 northwest '' in @xcite , occurs at @xmath483 and is directly to the southeast of the remnant . toward hb 3 , this systemic velocity corresponds to a distance of @xmath484 , slightly farther than inferred from this @xmath157 relation , although hb 3 is extended ( @xmath485 ) and presumably an old remnant at this distance ( @xmath486 in @xcite ) . thus , surface brightness and size estimates are less reliable than for younger remnants and may yield discrepant distances compared to other methods . , m. , hoppe , s. , komin , n. , et al . 2008 , in american institute of physics conference series , vol . 1085 , american institute of physics conference series , ed . f. a. aharonian , w. hofmann , & f. rieger , 285288	we present molecular spectroscopy toward 50 galactic supernova remnants ( snrs ) taken at millimeter wavelengths in @xmath0co and @xmath1co @xmath2 with the heinrich hertz submillimeter telescope as part of a systematic survey for broad molecular line ( bml ) regions indicative of interactions with molecular clouds ( mcs ) . these observations reveal bml regions toward nineteen snrs , including nine newly identified bml regions associated with snrs ( [email protected] , [email protected] , [email protected] , [email protected] , [email protected] , [email protected] , [email protected] , [email protected] , [email protected] ) . the remaining ten snrs with bml regions confirm previous evidence for mc interaction in most cases ( [email protected] , kes 75 , 3c 391 , kes 79 , 3c 396 , 3c 397 , w49b , cas a , ic 443 ) , although we confirm that the bml region toward hb 3 is associated with the w3(oh ) region , not the snr . based on the systemic velocity of each mc , molecular line diagnostics , and cloud morphology , we test whether these detections represent snr - mc interactions . one of the targets ( [email protected] ) had previous indications of a bml region , but we did not detect broadened emission toward it . although broadened @xmath0co @xmath2 line emission should be detectable toward virtually all snr - mc interactions we find relatively few examples ; therefore , the number of interactions is low . this result favors mechanisms other than sn feedback as the basic trigger for star formation . in addition , we find no significant association between tev gamma - ray sources and mc interactions , contrary to predictions that snr - mc interfaces are the primary venues for cosmic ray acceleration .
0.5 in the e2 type elliptical galaxy ngc 5903 ( @xmath4 ) forms a pair with the e0.5 elliptical galaxy ngc 5898 ( @xmath5 ) ; these two are the dominant members of a compact group that also contains an so type elongated dwarf galaxy eso 514@xmath0g003 ( @xmath6 ; maia , da costa & latham 1989 ; see , also , appleton , pedlar & wikinson 1990 , hereinafter apw90 ) . the group lies at a distance of @xmath7 mpc ( apw90 ; macchetto et al . 1996 ; rickes , pastoriza & bonatto 2008 ; serra & oosterloo 2010 ) , taking a hubble constant of 70 kms@xmath8 mpc@xmath8 . ( komatsu et al . 2011 ) . at this distance 1 arcmin corresponds to a projected distance of 10.2 kpc . the two dominant galaxies , ngc 5898 and ngc 5903 , are separated by 5.47 arcmin on the sky , while the third member , eso514@xmath0g003 is located 2.95 arcmin south of ngc 5903 ( rickes et al . basic optical and radio parameters of these galaxies are summarized in table 1 . in the soft x - ray band , only upper limits are available for their luminosities ; these limits are 10@xmath9 , 10@xmath10 and 10@xmath11 ergs s@xmath8 for ngc 5903 , ngc 5898 ( beuing et al . 1999 ) and eso514@xmath0g003 ( burstein et al . 1997 ) , respectively . the first radio observations of this group were reported at 2640 mhz using the twin - element interferometer at the owens valley radio observatory , with an east - west spacing of 100 ft , yielding a flux density of 90@xmath1230 mjy for ngc 5903 ( rogstad & ekers 1969 ) . this was followed up with 5 ghz observations using the parkes radio telescope with a pencil - beam of 4 arcmin ( disney & wall 1977 ) . the radio source was found to be well resolved , having a size of @xmath13 arcmin and an integrated flux density of 104@xmath1230 mjy , with its peak closer to the dwarf galaxy eso514@xmath0g003 . a somewhat clearer picture emerged when this group was observed in the course of the ooty lunar occultation survey at 327 mhz ( gopal - krishna 1978 ) . the higher resolution achieved in those observations revealed two resolved sources , the brighter one coinciding with ngc 5903 . the fainter radio source identified with eso514@xmath0g003 was found to have a size of @xmath14 arcmin , with its major axis inclined eastward from the optical axis of the elongated host so galaxy . also , a comparison of these occultation observations with the available parkes pencil - beam measurements ( disney & wall 1977 ) revealed an unusually steep spectrum for the extended radio counterpart of ngc 5903 ( gopal - krishna 1978 ; table 1 ) . the massive elliptical companion ngc 5898 has , however , remained undetected in all the radio observations mentioned above , as well as in the later radio observations ( see below ) . a major advance in the investigation of this system came from the work of apw90 who used the vla at 1.4 ghz to image both the radio continuum ( 30 arcsec beam ) and hi line emission ( 45 arcsec beam ) . their hi map showed an hi mass of @xmath15m@xmath16 to be associated with ngc 5903 , nearly a third of which is in a kinematically disturbed state and distributed along the galaxy minor axis , while the remainder is in the form of two huge hi trails extending towards the north and south from the inner hi distribution , with projected lengths of approximately 73 kpc and 27 kpc , respectively . the hi trails are probably the tidal debris left from an encounter with the disk galaxy eso 514@xmath00050 located 16 arcmin to the northeast from ngc 5903 ( apw90 ) . interestingly , while no hi was found associated with the massive elliptical companion ngc 5898 , the southern hi filament was found to approach the dwarf companion eso 514@xmath00050 . the conspicuous lack of both radio continuum and hi emission from ngc 5898 seemed particularly striking , when contrasted with the situation for the remaining two members of the group , namely ngc 5903 and eso 514@xmath0g003 . both these galaxies , which were already known to host extended radio sources ( see above ) , were now also found to be potential sites of hi gas accretion . thus , the apparently correlated presence ( or absence ) of synchrotron radio emission and ongoing hi accretion , exhibited by all 3 early - type galaxies of this group could be a manifestation of the importance of cold gas accretion for triggering the nuclear jet activity in elliiptical galaxies , as hypothesized in many papers ( sect . however , this seemingly self - consistent picture encountered a setback due to the assertion of apw90 that the double radio source ( of size @xmath17 arcmin ) which had previously been identified with the lenticular eso514@xmath0g003 in the ooty lunar occultation survey mentioned above , was in fact significantly offset from the galaxy and hence probably a background radio source unrelated to eso514@xmath0g003 . in this paper we shall revisit these discrepant findings , using the recent observations made with the giant metrewave radio telescope ( gmrt , swarup et al . 1991 ; ananthakrishnan 2005 ) , which have revealed in unprecedented detail the structure of the extended radio synchrotron emission from this compact group of galaxies at 150 mhz . in sect . 2 we briefly recapitulate the main optical properties of this system and in sect . 3 we present the new radio data a discussion and conclusions are in sect . the salient features of the two dominant members of the group are presented below . * ngc 5903 : * this galaxy of the morphological type e2 has an extended red nucleus ( sparks et al . 1985 ) . macchetto et al . ( 1996 ) have detected in its inner parts , a @xmath18 kpc@xmath19 wide region of ionised gas , extended along the galaxy major axis . this region encompasses a dusty patch of size @xmath203.3 kpc inferred from the @xmath21 colour index map , which is slightly elongated at p.a . @xmath22 ( ferrari et al . 1999 ) . according to rickes et al . ( 2008 ) , the observed mgii abundance gradient is consistent with the origin of this massive galaxy , of dynamical mass @xmath23 m@xmath24 , in merger events . their stellar population synthesis has revealed at least two stellar populations with ages of 13 gyr and 5 gyr and , possibly , a younger population of age @xmath14 gyr , which is mainly discernible in the central parts of this elliptical . deep v - band imaging has revealed a modest degree of tidal disturbance in the stellar body of this galaxy , which is typical of nearby massive ellipticals in groups , which are thought to grow mainly via dry " mergers ( tal et al . 2009 ) . however , serra & oosterloo ( 2010 ) have argued that the presence of cold gas may often play a significant role in the assembly of massive galaxies outside cluster environments . finally , slit spectra taken along the optical major axis of this elliptical show little rotation of the stellar body , while the ionized gas has an irregular velocity profile along the radius ( caon , macchetto & pastiroza 2000 ; also longo et al . 1994 ) . * ngc 5898 : * this nearly circular e0.5 elliptical shows an outer arc of dust and an inner red patch of elliptical shape . the gap between them indicates either intermittent accretion or a rapid dissipational process ( sparks et al . 1985 ) . macchetto et al . ( 1996 ) have detected a roughly circular central patch of ionised gas of diameter @xmath13 kpc . a deep v - band image shows tidal disturbance that is much stronger than is typical of nearby massive ellipticals in groups ( tal et al . 2009 ) , and thus suggestive of a major merger event . this is consistent with the complex kinematics revealed by the detection of a counter - rotating stellar core on the galaxy major axis ( pa @xmath25 ) , with a transition radius of 1 kpc and a gas rotation which , remarkably , is opposite to that of the stars on the galaxy minor axis ( caon et al . 2000 ; bertola & bettoni 1988 ) . figure 1 reproduces a new radio map of this compact group , taken from the 2nd data release ( dr2 ) of the ongoing tifr.gmrt.sky.survey ( tgss ) which is in the process of imaging the entire sky north of declination @xmath26 at 150 mhz with a resolution of @xmath27 arcsec and a typical rms noise of 57 mjy / beam ( the value for the present map is on the lower side , being only 3 4 mjy / beam ) . the radio contours are overlaid on the r - band optical photograph reproduced from the digital sky survey . in figure 2 we show the same radio image in grey scale , overlaid on which are the outermost ( lowest ) 4 contours of hi emission , reproduced from the vla map made by apw90 . lastly , we show in figure 3 the 150 mhz ( tgss ) and 1.4 ghz ( nvss ) emission contours in the region of the dwarf so galaxy eso514@xmath0g003 . the tgss map at 150 mhz has unravelled , for the first time , the complexity of the radio emission associated with this group of 3 early - type galaxies , although the massive elliptical ngc 5898 is still undetected . the 150 mhz image of the other massive elliptical , ngc 5903 , exhibits a number of interesting features , including a jet - like structure of size @xmath20 1 arcmin and flux density 350 @xmath12 65 mjy , protruding from the elliptical towards the southwest . the galaxy is surrounded by a diffuse radio halo of diameter @xmath28 7.5 arcmin ( 77 kpc ) and flux density @xmath28 6.7 @xmath12 1.1 jy encompassing a web of radio filaments and extending over the region bounded by the galaxy triplet . since the filaments have not been discerned in the previously published radio maps at higher frequencies , probably due to their lower resolution ( sect . 1 ) , it is presently not possible to determine their radio spectrum . nonetheless , the available data do provide reliable estimate of the integrated radio spectrum of this elliptical ngc 5903 between 150 mhz and 1.4 ghz where the maps have adequate resolution ( table 1 ) . although the highest frequency at which this source has been mapped is 5 ghz , the published value of 104 @xmath12 30 mjy measured with the 4 arcmin parkes pencil - beam can only be regarded as a generous upper limit , since much of this emission is contributed by the dwarf neighbour eso 514@xmath0g003 , as inferred from the marked shift in the radio centroid at 5 ghz towards it ( see table 2 of disney & wall 1977 ) . the spectral index for ngc 5903 is ultra - steep ( @xmath29 , with @xmath30 ) , in agreement with the previous estimate ( gopal - krishna 1978 ) . the 150 mhz map is also useful for clarifying the confusion over the radio emission from eso 514@xmath0g003 , the least massive , lenticular member of this triplet . as mentioned in sect . 1 , apw90 found the bright symmetric double radio source in their 1.4 ghz vla map ( resolution @xmath31 ) to be significantly displaced from the optical galaxy ; they concluded that it is an unrelated background radio source ( with an undetected optical counterpart ) . on the other hand , the lunar occultation scans at 327 mhz were consistent with a physical association of the radio source with the lenticular galaxy ( gopal - krishna 1978 ) . to pursue the matter , we have overlaid on the galaxy image the contours of radio emission at 150 mhz ( tgss ) and 1.4 ghz ( nvss ) , also marking the position of the mid - point of the symmetric double radio source seen in the 1.4 ghz vla map by apw90 ( their fig . as seen from figure 3 , the position offset of the galaxy eso514@xmath0g003 from the tgss position of the centre of the double radio source at 150 mhz is 8 arcsec ( which is at least a factor of two smaller than the offset of the galaxy from the apw90 position mentioned above ) , which is consistent with the expected tgss position uncertainty , given the confusion arising from the radio halo of ngc 5903 . likewise , a reasonable agreement of the galaxy position is also found with the nvss position at 1.4 ghz , though the radio source is poorly resolved due to the larger beam . thus , based on both tgss ( 150 mhz ) and nvss ( 1.4 ghz ) positions we conclude that the double radio source is associated with the lenticular eso514@xmath0g003 , in agreement with the lunar occultation results at 327 mhz ( sect . 1 ; gopal - krishna 1978 ) . note the westward extension of the double radio source visible in the tgss map raises the possibility that the radio source may actually be a triple source ( with a flux density of @xmath20 1150 @xmath12 190 mjy at 150 mhz ) , in which case the agreement with the galaxy position would be even better . based on their 1.4 ghz vla observations , apw90 characterised the radio counterpart of ngc 5903 as a head - tail radio source , with the tail extending to the southwest . such a radio morphology seems puzzling , since the host galaxy belongs to a very small group sparse in hot intra - group gas ( sect . the present tgss map at 150 mhz has revealed faint radio emission on the opposite , northeastern side of the galaxy as well . the radio source may thus be an asymmetric radio halo around ngc 5903 , with most of the radio emission located on the south - western side of the galaxy , within the region demarcated by the galaxy triplet ( figure 1 ) . the combination of high sensitivity and resolution of the 150 mhz tgss map has revealed an extensive network of radio filaments , the brightest of which fills the space delineated by the prominent hi trails which are seen to extend towards the south and southwest from ngc 5903 , in a ` bay ' like configuration ( figure 2 ) . as the filamentary structure has not been resolved at any other frequency , its radio spectrum is presently unknown . we shall make a plausible assumption that the radio emission is synchrotron radiation . figure 2 shows the morphological relationship of the radio continuum filaments to the hi trails reproduced from apw90 . based on the observed hi velocity gradients and other arguments , apw have estimated that the huge hi filaments are transient features that would have dissipated had they been older than @xmath32 yrs . the ultra - steep spectrum of the radio source ( @xmath33 , table 1 ) suggests that the source is very old and its age is likely to be of order of the maximum lifetime of the radio mode ( roughly @xmath34 yr , e.g. , leahy , muxlow & stephens 1989 ; liu et al . 1992 ; barai & wiita 2006 ; croton et al . 2006 ; shabala et al . thus , the triggering of the currently observed radio source in ngc 5903 may well have coincided with the formation of the hi trails and cold gas accretion through them , onto the galactic nucleus . this situation is rare and contrasts with the usual pattern found for nearby radio - loud elipticals , wherein the detected radio source is much younger than the associated extended hi structure ( typically a regular rotating disk , or ring - like formation , stretching beyond the optical extent of the host elliptical ) ( e.g. , tadhunter et al . 2005 ; emonts et al . 2006 ) . this has led to the view that if any radio activity was triggered at the time of a ( major ) merger between two gas - rich galaxies , resulting in an elliptical ( e.g. , naab , burkert & hernquist 1999 ) , the relic of that radio source is generally not visible any more ( e.g. , emonts et al . note that the latest major starburst episode in this elliptical is estimated to have occurred approximately 1 gyr ago ( rickes et al . 2008 ) , which presumably predates the birth of the radio source we see today . however , based on the recent evidence , the link between starburst and radio - loud nuclear activity does not seem to be strong ( dicken et al . 2011 ) . further , this nearby galaxy provides a striking manifestation of the influence of the hi tidal trails on the radio morphology . from figure 2 the brightest radio filament is aligned with the inner ( eastern ) edge of the southwestern bright hi trail . this alignment , together with the lateral offset from the hi ridge - line , appears to signify a dynamical interaction of the hi trail ( presumably infalling onto ngc 5903 ) with the surrounding synchrotron plasma that led to compression of the magnetic field along its eastern edge . this process could boost the radio emission locally , if relativistic particles are entangled in the regions of compressed magnetic field associated with the hi filaments . such a ` trapping ' of the synchrotron plasma by the hi filaments amounts to a kind of prolonged ` inertial containment ' of the synchrotron plasma , which could explain the ultra - steep spectrum of the diffuse radio source at metre wavelengths , even though such spectrum is generally not expected for sources outside clusters of galaxies , where the confining hot gaseous environment is lacking ( sect . 1 ) . lastly , a highly suggestive correlation emerging from the present work derives from our establishing the radio counterpart to the so galaxy eso514@xmath0g003 . with this , two members of this triplet of early - type galaxies are shown to have radio counterparts and these two galaxies are also the only ones that appear connected to hi trails and therefore likely to be accreting the cool hi gas . the third member , the dominant elliptical ngc 5898 is radio quiet and also unconnected to any hi trail ( apw90 ) . with the obvious caveat of small number statistics ( such a relation has a 1/8 chance of occurring randomly in the present case , if both hi gas and radio emission are independent and each have a probability of @xmath35 ) , this correlation between radio loudness of an early - type galaxy and its physical connection to an hi filament ( or , vice versa ) , seems entirely consistent with the importance of cold gas accretion in triggering and sustaining powerful jet activity in the galactic nuclei , as highlighted in the recent literature ( e.g. , best et al . 2005 ; croton et al . it is tempting to speculate that the correlation with radio emission arises because , in comparison to ionized gas ( viz . , the case of bondi accretion : allen et al . 2006 ; hardcastle et al . 2007 ; balmaverde et al . 2008 ; buttiglione et al . 2010 ) , cold gas is more amenable to fast accretion merely because any infalling cold gas would encounter much less resistance from the substantial magnetic fields likely to exist within the cores of massive galaxies . we are conducting a search of other small groups mapped by tgss that also have hi data available , to investigate whether the correlation between hi filaments and radio activity in early - type galaxies seen in this particular triplet remains valid . the giant metrewave radio telescope ( gmrt ) is a national facility operated by the national centre for radio astrophysics ( ncra ) of the tata institute of fundamental research ( tifr ) . we thank the staff at ncra and gmrt for their support . this research has used the tifr . survey ( http://tgss.ncra.tifr.res.in ) data products , nasa s astrophysics data system and nasa / ipac extragalactic database ( ned ) , which is operated by the jet propulsion laboratory , california institute of technology , under contract with national aeronautics and space administration . allen s. w. , dunn r. j. h. , fabian a. c. , taylor g. b. , reynolds c. s. , 2006 , mnras , 372 , 21 ananthakrishnan s. , 2005 , icrc , 10 , 125 appleton p.n . , pedlar a. , wilkinson a. , 1990 , apj , 357 , 426 balmaverde b. , baldi r. d. , capetti a. , 2008 , a&a , 486 , 119 barai p. , wiita p. j. , 2006 , mrnas , 372 , 381 bertola f. , bettoni d. , 1988 , apj , 329 , 102 best p.n . , et al . , 2005 , mnras , 362 , 25 beuing j. , dobereiner s. , bhringer h , bender r. , 1999 , mnras , 302 , 209 burstein d. , jones c. , forman w. , marston a.p . , marzke r. o. , 1997 , apjs , 111 , 163 buttiglione s. , et al . , 2010 , a&a , 509 , a6 caon n. , macchetto d. , pastiroza m. , 2000 , apjs , 127 , 39 croton d.j . , et al . , 2006 , mnras , 365 , 11 dicken d. , et al . , arxiv:1111.4476 disney m. j. , wall j.v . , 1977 , mnras , 172 , 235 emonts b. h. c. , morganti r. , tadhunter c.n . , 2006 , a&a , 454 , 125 emonts b.h.c . , morganti r. , oosterloo t.a . , 2007 , a&a , 464 , l1 gopal - krishna , 1978 , mnras , 185 , 579 ferrari f. , pastoriza m.g . , macchetto f. , caon n. , 1999 , a&as , 136 , 269 hardcastle m.j . , evans d.a . , croston j.h . , 2007 , mnras , 376 , 1849 komatsu e. , et al . , 2011 , apjs , 192 , 18 leahy j. p. , muxlow t. w. b. , stephens p. w. , 1989 , mnras , 239 , 401 liu r. , pooley g. , riley j.m . , 1992 , mnras , 257 , 545 longo g. , zaggia s.r . , busarello g. , richter g. , 1994 , a&as , 105 , 433 macchetto f. , et al . , 1996 , a&as , 120 , 463 maia m.a . , da costa l.n . , latham d. w. , 1989 , apjs , 69 , 809 naab t. , burkert a. , hernquist l. , 1999 , apj , 523l , 133 rickes m.g . , pastoriza m.g . , bonatto c , 2008 , mnras , 384 , 1427 rogstad d. , ekers r. , 1969 , apj , 157 , 481 serra p. , oosterloo t.a . , 2010 , mnras , 401l , 29 shabala s. s. , ash s. , alexander p. , riley j.m . , 2008 , mnras , 388 , 625 sparks w.b . , et al . , 1985 , mnras , 217 , 87 swarup g. , ananthakrishnan s. , kapahi v. k. , et al . , 1991 , current sci , 60 , 95 tadhunter c. , robinson t. g. , gonzlez delgado r. m. , wills k. , morganti r. , mnras , 356 , 480 tal t. , van dokkum p.g . , nelan j. , bezanson r. , 2009 , aj , 138 , 1417	we discuss the nature of the multi - component radio continuum and hi emission associated with the nearby galaxy group comprised of two dominant ellipticals , ngc 5898 and ngc 5903 , and a dwarf lenticular eso514@xmath0g003 . striking new details of radio emission are unveiled from the 2nd data release of the ongoing tifr.gmrt.sky.survey ( tgss ) which provides images with a resolution of @xmath1x@xmath2 and a typical rms noise of 5 mjy at 150 mhz . previous radio observations of this compact triplet of galaxies include images at higher frequencies of the radio continuum as well as hi emission , the latter showing huge hi trails originating from the vicinity of ngc 5903 where hi is in a kinematically disturbed state . the tgss 150 mhz image has revealed a large asymmetric radio halo around ngc 5903 and also established that the dwarf so galaxy eso514@xmath0g003 is the host to a previously known bright double radio source . the radio emission from ngc 5903 is found to have a very steep radio spectrum @xmath3 and to envelope a network of radio continuum filaments bearing a spatial relationship to the hi trails . another noteworthy aspect of this triplet of early - type galaxies highlighted by the present study is that both its radio loud members , namely ngc 5903 and eso514@xmath0g003 , are also the only galaxies that are seen to be connected to an hi filament . this correlation is consistent with the premise that cold gas accretion is of prime importance for triggering powerful jet activity in the nuclei of early - type galaxies . [ firstpage ] galaxies : active galaxies : ism galaxies : jets radio continuum : general galaxies : interactions galaxies : elliptical and lenticular , cd
essentially all missions sensitive to cosmic x - rays have detected intense x - ray outbursts with timescales from seconds to hours . peak fluxes range to 1 crab and above ( 1 crab=2@xmath310@xmath4 ergs @xmath5 s@xmath6 in 2 - 10 kev band ) . x - ray outbursts are unlike classical x - ray transients , which persist for weeks or months . x - ray outbursts fade within a day and are normally only seen once . historically , x - ray outbursts seen for less than a day that lack persistent x - ray counterparts were called `` fast x - ray transients''(fxts ) . recently a new term , `` x - ray flashes''(xrfs ) , was coined for short ( less than 1000 sec ) intense bursts of x - ray emission@xcite . xrfs lack detectable gamma - ray emission , but are reminiscent of gamma - ray bursts in their time history . it can be difficult in some cases to distinguish xrfs and other fxts using archival data e.g. where the time resolution is inadequate . the sources of fxts might be heterogenous ; the experiments that measure them are certainly so . fxts have been surveyed in a variety of energy bands , sometimes with low spectral and time resolution , and typically with low angular accuracy . there is no straightforward way to obtain a uniform large set of fxt parameters . it is no surprise that different experiments have suggested different sources for fxt , including flare stars , compact objects , extragalactic sources , and x - ray emission from grbs [ e.g. pye & mchardy 1983 ( hereafter * ? ? ? * ) ; ambruster & wood 1986 ( hereafter * ? ? ? * ) ; castro - tirado et al . 1999 ( hereafter * ? ? ? however , the quantitative contribution of these several sources is still unknown , and room remains for unknown phenomena . the discovery by bepposax of long - lasting x - ray emission from grbs , the so called x - ray afterglow @xcite , and the strong prompt x - ray emission from grb seen by ginga @xcite and watch @xcite , shows that some fast x - ray transients are related to grbs . theoretical work predicts the existence of grb types that are x - ray rich or gamma - ray quiet . this might be caused by physical properties of the grb central engine and surrounding media , by the geometry of grb source with respect to the observer s line of sight , or by the redshift of classical ( standard ) grbs ( see recent review by @xcite ) . in this paper we attempt to construct a uniform data set of fxts from the heterogeneous data available to date . in section 2 we describe the experiments , and in section 3 our data analysis . in section 4 we discuss the connection of fxts and grbs , and estimate the fraction of grb - related fxts in fxt data . in section 5 we discuss other possible sources of fxt progenitors . the importance of sensitive uniform sampling has been well recognized in the case of grbs ( see * ? ? ? * ; * ? ? ? * and elsewhere ) . the batse / cgro experiment , with its large homogenous and uniform database , has allowed the study of statistical properties of classical grb . a similar database , if it existed , would allow an unbiased determination of the important parameters of fxts , and could test different hypotheses for their origin . although fxts are by no means rare , we still do not have a large uniform set of these events . in its absence , we have combined the data provided by several very different experiments to quantify , as best we can , the statistics of fxt occurrence . while we recognize the imperfection of this approach , it is the only way to estimate the fxt distribution over several orders of magnitude in fluence . a brief description of the instruments and data sets is given below . @xcite published a catalog of fxts detected in the 2 - 18 kev band by ariel-5 . flux measurements from the rapid - spinning satellite were integrated in 100 minute bins corresponding to the satellite orbit . these fluxes provide a good estimate of the integral fluence in the time bin . transient events were selected by their detection in a single , or at most a few , satellite orbits . the lightcurves therefore have 100-min time resolution . the limiting sensitivity was @xmath720 mcrab ( 5.5@xmath8 ) in 100 minutes . @xcite found a cumulative log n - log f peak flux relation for 27 detected fxts consistent with a power law with index @xmath9 = -0.8 @xmath10 0.5 . @xcite estimated the total number of fxts above their threshold as n=150 - 180 per year for the whole sky . the heao-1 satellite carried two x - ray surveys , a1 and a2 . these instruments scanned a great circle on the sky every 35 minutes . sources were sampled for a short interval once or twice per scan ( 10 sec for a1 and 60 sec for the a2 experiment ) . @xcite published a survey of transients detected by heao-1 a1 in the 0.5 - 20 kev band . this @xcite survey was based on excess emission over an 12-hour interval . 10 fxts were detected , with a limiting sensitivity about @xmath74 mcrab . the log n - log f distribution was consistent with a power law with index @xmath11 . @xcite estimated the total number of fxts as n=1500 - 3000 per year above their threshold , assuming that a typical event lasted more than 1.5 hours . connors et al . ( 1986 , hereafter * ? ? ? * ) published the heao-1 a2 survey of fxts in the 2 - 20 kev band . 8 fxts were detected with a limiting sensitivity about @xmath74 - 6 mcrab . the log n - log f distribution fit a power law with index @[email protected] . the duration of the detected fxts was poorly determined , and might have anywhere between 60 s and @xmath72000 s. the all - sky rate of such fxts would have been between 10@xmath12 per year for durations 2000 s and 3@xmath310@xmath13 per year for durations of 60 s. @xcite published the detection of 7 fxts by the watch all - sky monitor onboard the granat satellite . fxts were detected in a 8 - 15 kev energy channel . watch , with a time resolution of several seconds , was able to resolve the fast transients it detected ( @xmath7100 min for 3 fxts and @xmath71 day for the rest ) . the typical peak flux for these sources was several hundred mcrab . because the lightcurves are well - sampled , we have a good estimate of the total fluences . sazonov et al . ( 1998 ) used watch to observe an additional 95 events in the energy band from 8 to 60 kev . 86 of them were observed by other grb experiments , and confirmed to be grbs , and 47 events were localized to 1 degree or better . we used the sazonov et al . ( 1998 ) data to constrain the @xmath2 ratio for gamma - ray bursts ( see section 4 ) . vikhlinin ( 1998 , hereafter * ? ? ? * ) searched for faint x - ray bursts ( duration 10 - 300 s ) in pointed rosat pspc observations with a total exposure of 1.6@xmath310@xmath14 s. only the hard rosat energy channel ( 0.5 - 2 kev ) was used for their study . a total of 141 x - ray flares with duration 100 - 300 s were detected . the quiescent flux was consistent with zero in approximately half of the burst sources , while positive quiescent emission was detected for the rest . 112 bursts were identified as galactic stars . one burst was from the x - ray binary lmc x-4 . the 28 remaining fxts were not identified , but for most of them starlike optical counterparts were found in the digital sky survey . the limiting fluence of these events ( 10 - 100 s duration ) corresponds to 2.6@xmath310@xmath15 ergs @xmath5 in the 0.5 - 2 kev band . we have translated the fluence from rosat energy band to 2 - 10 kev under assumption of crab - like spectrum and neglecting foreground absorption . this implies a limiting fluence of 3.5 mcrab@xmath3sec . the estimated rate of such events is @xmath710@xmath16 per year for the whole sky . greiner et al . ( 2000 ) searched for grb afterglows in the rosat all - sky survey . during the survey , the telescope s field of view scanned a great circle on the sky , sampling sources for 10 to 30 s. the sensitivity limit per scan is 10@xmath17 erg s@xmath6 @xmath5 . greiner et al . ( 2000 ) applied tight restrictions on the time profile to extract grb afterglow candidates , and selected 23 afterglow candidates . closer examination showed that about 14 of afterglow candidates were in fact flares from late - type stars of class m - dme . with less restrictive requirements on the time profile greiner et al . ( 2000 ) found more afterglow candidates , mostly associated with flare stars . gotthelf , hamilton & helfand ( 1996 ; hereafter * ? ? ? * ) reported the detection of 42 faint x - ray flares in the einstein ipc data . those flares have typical fluxes from 10@xmath15 ergs @xmath5 to 10@xmath18 ergs @xmath5 in the 0.2 - 3.5 kev energy range , and their durations were less than 10 s. their detection rate corresponds to about 10@xmath16 flares per year for whole sky . however , they concluded that only the hardest of their events , consisting of 18 flashes , was undoubtedly of cosmic origin . the remaining soft events might have been detector artifacts . strohmayer et al . ( 1998 ) reported on spectral fits and the @xmath2 ratio for 22 grbs observed with ginga over 2 to 400 kev band . we used their fits , together with data from other instruments , in our @xmath2 distribution . batse grbs are cataloged in the 4b and ` current ' batse databases . a recent catalog of bright batse grbs was published by preece et al . ( 2000 , hereafter * ? ? ? two searches by kommers et al . ( 1999 ) and stern et al . ( 2001 ) showed that the batse database contains almost as many untriggered grbs as triggered grbs . the untriggered grbs tend to be fainter than the triggered grb ( see fig.8 of * ? ? ? * ) , for natural reasons . we used both triggered and untriggered grbs to obtain a log n - log s for grbs in the gamma - ray band . spectral fits for 53 grbs detected with batse were published by band et al . ( 1993 ) . using the @xcite catalog , we calculated spectral parameters and estimated @xmath2 for 81 more batse grbs . we used bepposax afterglow observations ( see review in * ? ? ? * ) as inputs to for our estimate of the fraction of grb - related fxts . frontera et al . ( 2000a , 2000b , 2001 ) , int zand et al . ( 1999 , 2000a , 2001 ) , pian et al . ( 2001 ) and nicastro et al . ( 2001 ) presented the x - ray and gamma - ray fluences for 15 grbs in the 2 - 10 kev and 40 - 700 kev bands measured by bepposax . we use these data to help estimate @xmath2 . although we discuss the statistics of grbs and xrfs reported by heise , int zand , & kuulkers ( 2000 ) , and @xcite we did not include them in our quantitative analysis because no numerical data for those xrfs have been published in useful form . smith et al . ( 2001 ) recently presented rxte asm and konus / wind fluence measurements in the 1.5 - 12 kev and 50 - 200 kev energy bands for 14 grbs detected by both these instruments . we have used these data for our @xmath2 estimate . our study started from the analyses published for the individual instruments . we used both published results , and , in some cases , data from public archives . however , because results of previous studies were strongly instrument dependent , they can not be directly compared . we need to convert data from the individual instruments into a quasi - uniform data set . our analysis will proceed in two directions : in the forward direction ( approach # 1 ) , we will fit a power law luminosity distribution ( log n - log s ) to the data . in the backward direction ( approach # 2 ) we extract a model - independent luminosity distribution by plotting each data set on the log n - log s plane . approach # 2 requires the several assumptions discussed below . most of the aforementioned experiments measured the fluence ( time - integrated flux ) of detected events , but the heao-1 instruments provided only intermittent sampling of fluxes . these heao-1 results are important , however , because they were able to detect fainter fxts than ariel-5 and watch . rosat and einstein detected even fainter fxts , but those experiments were sensitive in a much softer energy band , and are subject to additional corrections when converted to the 2 - 10 kev band . the watch data also need correction , because that instrument was not sensitive to softer x - rays , but in this case the overlap with other x - ray instruments is larger and hence the correction is not so critical . additionally , interstellar absorption , which can be safely neglected in most cases above 2 kev , is an important factor in the rosat and einstein bands , adding more uncertainty to the results . finally , we did not have information about individual fxts detected with rosat and einstein , and used instead the integral estimates presented by @xcite and @xcite . for our primary analysis ( approach # 1 above ) we have used data from ariel-5 ( energy range 2 - 18 kev , minimum detectable fluence about 100 crab@xmath3sec ) , heao-1 a1 & a2 experiments energy range 0.5 - 20 kev & 2 - 20 kev , minimum detectable fluence 0.5 crab@xmath3sec ) , and watch ( energy range 8 - 15 kev , minimum detectable fluence about several hundred crab@xmath3sec ) . the fluence sensitivity limits above are given for long - duration events , up to 1 day . the fluence sensitivity is much higher for short intense bursts , where background is not important . to include heao-1 data into our analysis , we need to make assumptions about the duration and time profile of the transients . fortunately , because of the nearly @xmath71 slope of the cumulative distribution , almost the same distribution follows from a wide range of assumptions on duration . we have only lower ( the scan duration ) and upper ( the scan interval ) limits to the duration of the heao events . we estimated fluences for the heao-1 events by multiplying the measured flux by an assumed event duration . our initial estimate of event duration was taken as the geometric mean of the minimum possible fxt duration ( usually the scan duration of 10 - 60 sec ) and the maximum possible fxt duration , which is several hours for the a1 experiment and @xmath74000 s for the a2 experiment . to quantify the transient statistics , we assume that they are distributed as @xmath19 , where @xmath20 is the normalization at 1 crab@xmath3sec , and determine the parameters of this distribution with a maximum likelihood analysis . in our first analysis , we assumed that the spectrum of the events was similar to the crab nebula , i.e. , a power law with photon index @xmath21 . we then calculated the minimum detection fluence for each experiment . the overall event rates were taken from the published results of the original analyses @xcite . more detailed discussion of sky coverage , thresholds and detection bands can be found in the references above . the parameters of our maximum - likelihood determination of the parameters @xmath9 and @xmath20 are shown in fig.1 . it is known that many flare stars produce short , intense x - ray flares @xcite with thermal spectra . because many fxt have been identified with flare stars @xcite , we have calculated the fxt distribution that results if the fxt consist of mixture of crab - like and thermal spectra spectrum . a mixture of 30% thermal - spectrum fxt , with @xmath22 , and 70% crab - spectrum fxts maximizes the maximum likelihood function , considering the range kt = 0.5 - 4.5 kev , and ` thermal ' burst fraction 0 to 100% . given this mix of fxt spectra , our best estimate of @xmath23 yields @[email protected] and @xmath24 at @xmath25 . a conservative , 3@xmath8 limit bounds the parameters to @xmath9=[0.7 - 1.2 ] , @xmath20=[@xmath26-@xmath27 in any case , fxts are common , occurring somewhere on the sky dozens to hundreds of times each day . they are one or two orders of magnitude more common than classical gamma - ray bursts ( kommers et al . 1999 ) . note that it is impossible to determine whether all the events that make up this distribution fit our strict definition of occurring only once . if some of the fxts come from magnetic stars , rs cvns , or other nearby objects , they may indeed repeat . these repetitions would have been seen by heao-1 , ariel-5 , etc . if those instruments had better sky coverage or longer lifetimes . our distribution ( @xmath28 and @xmath29 , 70% crab - like vs. 30% thermal spectra ) is shown on fig.2 . it is not significantly different from the distribution that follows if we assume that all bursts have a crab - like spectra . fig.2 also shows the segments of the distribution seen by the individual instruments ( our analysis approach # 2 ) . the shaded areas represent 1@xmath8 errors for the individual experiment log n - log s curves , including the normalization and the statistics of the number of events seen by each experiment . although the duration of the heao-1 events is quite uncertain , this uncertainty does not affect the overall distribution . we plot the backward - derived log n - log s assuming the geometric - mean time duration discussed above . but because the overall distribution has a slope close to 1 , changes in the assumed duration simply slide the log n - log s contribution from the heao-1 instruments up or down along the same curve . the dashed / dotted boundaries show the effect of changing the duration to the extreme possibilities on the heao-1 a2 log n - log s distribution . the rosat results from @xcite and einstein results from @xcite were not used to determine our net log n - log s distribution . however , they are plotted on fig.2 for comparison . we have translated the fluence measured in the rosat energy band to 2 - 10 kev by assuming a crab - like spectrum with negligible foreground absorption . the events reported by @xcite had fluences from 3.5 to 200 mcrab@xmath3sec . we estimated the total number of events per year for rosat events which had no detectable quiescent emission as @xmath72@xmath310@xmath16 year@xmath6 , and for events without optical counterparts as @xmath74@xmath310@xmath13 events year@xmath6 . we have taken these values as upper and lower limits to the fxt rate at this fluence ( shown as the cross on fig.2 ) the _ einstein _ data were also converted to 2 - 10 kev assuming a crab - like spectrum ( shown as the rhomb on fig.2 ) , which is consistent with the spectral shape found by @xcite for hard events . our integrated distribution is consistent with a power - law with a slope @xmath30 . a spatially homogenous distribution of identical sources ( standard candles ) would have a slope -1.5 . our distribution is consistent with the slope of @xmath310.8 derived by pm83 from ariel-5 events . by adding data from several other experiments , we have extended the pm83 distribution by several orders of magnitude towards the faint end . the rosat and einstein points , derived from @xcite and @xcite , are consistent with an extrapolation of our log n - log s curve . fig.3 shows the sky distribution of the events used in fig.2 ( ariel , heao , and watch marked with crosses , and rosat and einstein events marked with triangles ) . the sky distribution is consistent with isotropy . the dipole moment of the distribution towards galactic center is [email protected] , and the quadrupole moment is [email protected] ( for an isotropic distribution the dipole moment is zero , and the quadrupole moment is 1/3 ) . we see no significant anisotropy for any subset of these data . however we need to note that there are no reliable exposure maps for mentioned above experiments , making it difficult to be definite about the lack of a disk population in these data . the connection between fxts and grbs is especially intriguing . as a result of the ginga and watch studies of prompt x - ray emission from grbs @xcite , and the bepposax discovery of grb afterglows @xcite , it is evident that grbs can emit a large fraction of their energy in the x - ray region ( see also * ? ? ? this emission is both coincident in time with the gamma - ray burst ( prompt emission ) and delayed ( afterglow ) . recently bepposax has found a few grb - like events with very little or no gamma - ray emission but strong x - ray emission @xcite . these events were designated `` x - ray flashes '' @xcite . however , kippen et al . ( 2001 , 2002 ) argue that they merge smoothly with regular gamma - ray bursts in a single distribution of their parameters . such x - ray rich grbs were estimated to total about 30% of all grbs @xcite . whether xrfs and grbs originate from a single class of progenitors with a smooth distribution of parameters , or are at least in some cases intrinsically different , is not yet clear . comparison of grbs and xrfs is fraught with selection effects , including the fact that the observed xrfs were selected in a different energy band ( x - rays ) than gamma - ray bursts . in the absence of gamma - ray data , one might classify a conventional grb that was detected in the x - ray band as an xrf . grindlay ( 1999 ) examined ariel-5 and heao-1 data for an excess of gamma - ray quiet fxts , above those expected from conventional gamma - ray bursts , that might indicate beaming of grbs . he estimated that the grb share of ariel-5 and heao-1 fxt events was about 50% . we wish to refine this estimate by bringing several parameters into our analysis . these include the fraction of grbs that emit in the x - rays , the ratio of the fluence emitted in the x - ray band to the fluence emitted in the gamma - ray band , the fraction of grb that produce x - ray afterglows , and the ratio of the x - ray fluence emitted in prompt phase to the fluence emitted in the early afterglow ( the first @xmath32 sec ) bepposax and watch @xcite indicate that the first parameter is close to unity for long - duration grbs . long - duration grbs are defined as events with @xmath33 sec in the batse data set @xcite . the second parameter of interest is the ratio of the fluence emitted in the x - ray band ( 2 - 10 kev ) to the fluence emitted in the gamma - ray band ( 50 - 300 kev ) , @xmath2 . the ratio @xmath2 can be estimated from the few events observed in both the x - ray and gamma - ray band , or , using the larger batse sample , by extrapolating the measured spectrum into the x - ray band . if the latter is consistent with the former estimate , one could use the full batse grb database to find the @xmath2 distribution band et al . ( 1993 ) has shown that grb spectra can be fit by the expression : @xmath34 @xmath35^{\alpha-\beta}\big(\frac{e}{100kev}\big)^{\beta}exp(\beta-\alpha ) , \ ; \ ; ( \alpha-\beta)e_{0}\le e , \ ; \ ; ( 1)\ ] ] in the 20 kev-2 mev energy band . using typical values for the parameters ( see for example * ? ? ? * ; * ? ? ? * ) , the average value of @xmath2 is about several percent . however , preece et al . ( 1996 ) has shown that such an approach leads to an underestimate of x - ray fluxes for about 15% of batse grbs . the magnitude of this underestimate can be relatively high ( up to order of magnitude above band s formula ) for an individual grb . ginga , which was able to simultaneously measure x- and gamma - ray emission from grbs , found an average ratio of prompt emission @xmath2 of 0.24 for 22 grb events , although the logarithmic mean @xmath2 was 0.07 @xcite . @xmath2 was close to unity for several events . bepposax confirmed that such a high ratio is not rare @xcite . data from different instruments are subject to significant differences in their energy bands and event selection criteria and often do not agree between themselves . to refine our estimate of @xmath2 we have used all relevant available data including batse , ginga , rxte , bepposax and watch . fits to the spectra of 53 batse and 22 ginga grbs , using the band formula , are found in band et . al ( 1993 ) and strohmayer et al . ( 1998 ) . using the catalog of batse grbs by @xcite , we have calculated the band parameters for 81 additional grbs from batse data . for all these bursts we calculated @xmath2 by integrating the band formula . smith et al . ( 2001 ) presented rxte asm and konus measurements of 14 grb in 1.5 - 12 kev and 50 - 200 kev energy bands . to convert asm counts to x - ray fluence we fit the three - color asm data to a power law for each grb ( assuming a single power law in the 1.5 - 12 kev region ) . to obtain the 50 - 300 kev fluence from the single - channel konus data we assumed an average value for the band parameters from @xcite . bepposax data for 15 grbs are presented at frontera et al . ( 2000a,2000b,2001 ) , int zand et al . ( 1999,2000a,2000b,2001 ) , pian et al . ( 2001 ) and nicastro et al . ( 2001 ) , who give x - ray and gamma - ray fluences for the 2 - 10 kev and 40 - 700 kev bands . when available , we used the individual band s parameters for each grb , and otherwise used average values from @xcite for 50 - 300 kev channel . sazonov et al . ( 1998 ) published a catalog of 95 grbs detected with watch . we used 82 of these events , those for which fluences are available for both energy channels ( 8 - 20 kev and 20 - 60 kev ) . full data were unavailable for the other 13 events . for the watch data we extrapolated @xmath36 and @xmath37 to our 2 - 10 kev and 50 - 300 kev channels by assuming that the @xmath9 band s parameter was fixed at average value found from ginga , rxte and bepposax measurements , @xmath7 -0.85 , while the other parameters were fixed at average values from @xcite . fig.4 shows the distribution of @xmath2 for all these events . the solid line shows the total @xmath2 distribution derived from the ginga , rxte / konus , bepposax and watch measurements ( 134 grb total ) , while the dashed line represents the batse events ( 134 grb total ) . note that the batse ratio comes from an extrapolation of the measured spectra to the x - ray band . the non - batse events , which are triggered at lower energies than batse , tend to have larger median @xmath2 , along with a tail of even higher @xmath2 , x - ray - rich events . we will call them x - ray detected grb ( xdgrb ) . the later distribution is in a good agreement with bepposax measurements @xcite . we need to note that the @xmath2 distribution certainly depends on how one weighs events from the different experiments . we consider the xdgrb estimate of @xmath2 to be more reliable than the ` batse ' estimate , because the ` batse ' @xmath2 ratio was calculated by extrapolation of band s parameters from the 20 kev-2 mev region , while for the xdgrb @xmath2 was actually measured , or calculated using band s parameters fit to 2 kev-700 kev measurements . the shift between the peaks of the two distributions is in the direction expected from selection effects ( see appendix ) . to determine the integrated x - ray fluence from grbs , we need to consider not only the prompt emission , but also the afterglow emission ( the third and forth parameters mentioned at the beginning of this section ) . data from bepposax show that 80 - 100% of grbs that emit prompt x - ray emission also show an x - ray afterglow @xcite . we use this fraction ( 100 % ) as our estimate of the third parameter . to estimate the fourth parameter ( the ratio of the x - ray fluence emitted in the prompt phase to the fluence from the early afterglow ) we use predictions of the standard fireball model . the standard fireball model ( wijers , rees , & meszaros 1997 ; sari , piran , & narayan 1998 ; @xcite ) predicts that in the afterglow phase , the x - ray flux decreases as a power law , as the grb - induced blast waves sweep up surrounding matter and decelerate . the index of this power law may change its value on time scales from minutes to hours . in this `` external shock '' model , both the tail of the prompt x - ray emission and the afterglow are produced by the blast wave ( wijers , rees , & meszaros 1997 ) . the external shock model predicts that the amplitude of x - ray prompt emission during the final part of the grb should be equal to the back - extrapolated flux from the afterglow . this back - extrapolation holds for most grbs that were observed by bepposax @xcite . however , because the afterglow measurements by bepposax start 6 - 8 hours after the grb , there is no direct measurement of the x - ray flux immediately after the grb . if the x - ray afterglow decays as a power law , its integral fluence can reach 50%-200% of the prompt emission @xcite . we have assembled from the literature decay indices for 15 grb , and have calculated indices for 10 more , based on the published estimates of prompt and afterglow fluxes . the decay indices @xmath9 fall in the range -1.0 to -1.9 , most commonly about @xmath71.3 , and the distribution of indices can be satisfactorily fit with a gaussian . in fig.5 , we combine all four parameters discussed above to obtain a log n - log s distribution for x - ray emission from grbs . this estimate is based on the gamma - ray ( 50 - 300 kev ) log n - log s distribution for all batse grbs with duration @xmath33 s. the solid curve is based on our xdgrb distribution of @xmath2 , sampled by monte carlo methods . to the prompt x - ray fluences we add a contribution from the afterglow obtained by monte carlo sampling of the afterglow decay parameter @xmath9 and integrating the afterglow out to @xmath26 seconds . these curves plot our estimate of the x - ray fluence from cataloged , triggered batse events . we also show for comparison the expected fxt distributions for less - preferred distributions of @xmath2 : the mean and logarithmic mean of ginga ( 0.07 and 0.24 , * ? ? ? * ) and our batse " distribution . the less - preferred @xmath2 yield numbers of fxts from grbs that fall below the preferred estimate . we would like to understand how these curves would change if we included all the grbs that are too faint to trigger batse . some of them may be , in fact , x - ray rich grb or xrfs ( see @xcite ) . we can get some notion of this correction by adding to our sample the sample of untriggered batse events obtained by kommers et al . ( 1998 ) and stern et al . ( 2000 ) , which enhance the batse sample of the faint end . we find little change for all but the faintest part of the xrf log n - log s curve . fig.6 shows the xrf log n - log s distribution for triggered , untriggered , and all known batse grbs , based on the xdgrb estimate of @xmath2 . fig.7 compares our prediction of the fxt s from grbs to our best - fit distribution of all fxt . the plotted distribution ( solid curve ) is based on the xdgrb estimate of @xmath2 , and the population of all ( triggered plus untriggered ) batse grbs . the distribution derived from the ` batse'@xmath2 ( shown with dotted curve ) . bepposax data ( heise , int zand , & kuulkers 2000 , @xcite ) suggest that about 70% of fxt detected by bepposax are grb - related . the ` batse ' @xmath2 would yeld a value of 0.1 - 1% for this ratio , far less than observed . fig.7 also compares our estimate of the log n - log s curve for fxt from grbs with the rate of candidate grb afterglows reported by greiner et al . ( 2000 ) . for any reasonable assumption about the @xmath2 distribution , the greiner et al . ( 2000 ) rate of events without optical counterparts is in good agreement with our prediction , suggesting that a substantial fraction of their residual , non - stellar , events are indeed grbs . it is clear that the fraction of fast x - ray transients that can be attributed to classical gamma - ray bursts drops with decreasing fluence . the predicted x - ray event rate for `` normal '' grbs falls a factor of @xmath7 3 - 5 below the total fxt rate at high fluences , but a factor of 100 below at the faint end . even if we are off by a factor of two or three due to unknown effects , a large majority of fxts are not counterparts of classical grbs ( e.g , not x - ray flashes ) , especially at the faint end . however , we can not rule out a possibility that exists a population of x - ray rich grb - like events that comprise a large fraction of the fxts at the faint end . the timescales and spectra of fxts are diverse . timescales range from seconds to hours , and spectra range from very hard ( kt@xmath3820 kev , * ? ? ? * ) to soft ( kt@xmath71 kev , * ? ? ? @xcite suggested the time profiles could be divided into classes , perhaps indicating multiple progenitor classes . the few fxt identifications made to date confirm the heterogeneous nature of fxts . the lower right corner of our x - ray log n - log s curve is formed by several events with duration up to one day , which were seen mostly by watch @xcite . a t tauri star was suggested by @xcite as the progenitor for one of the events , while the rest were probably generated by x - ray binaries . several x - ray binaries have been recently found to generate short , powerful outbursts that might be classified as fxts . for example , outbursts from v4641 sgr up to 12.2 crab ( 2 - 12 kev ) with durations less than a day were observed by the rxte and bepposax ( smith , levine , & morgan 1999 , @xcite ) .. during the rxte pca observation , v4641 sgr showed x - ray fluctuations by a factor of 4 on timescales of seconds and @xmath39 on timescales of minutes @xcite . the mass function identifies the system as a high - mass x - ray binary harboring a black hole @xcite . a slightly longer type of x - ray transient , ci cam , with a single short outburst about 2 days long , was also discovered by rxte ( @xcite , revnivtsev , emelyanov & borozdin 1999 ) . `` super type i x - ray bursts '' that last from 30 min to 3 hours were reported recently from several low - mass x - ray binaries ( e.g. , * ? ? ? * ; * ? ? ? similar events might been detected as high - fluence fxts by low - resolution experiments . these or similar sources might contribute to the lower right corner of the fxt log n - log s curve . int zand et al . ( 1998 ) , and kaptein et al . ( 2000 ) announced the discovery of type i x - ray bursts from ` empty ' places . these bursts , observed from locations without steady emission , would be naturally classified as fxts . the origin of such bursts might be low accretion rate neutron star binaries . we can get an idea of the rate of ` empty ' place bursts from bepposax@xcite . it observed the galactic center region for @xmath40 s and detected about @xmath71500 x - ray type i bursts from @xmath41=31 sources . at the same time it observed @xmath42=4 x - ray type i bursts from 4 ` empty place ' x - ray bursters . if all these events were generated on low - accretion x - ray neutron stars ( with , say , x - ray luminosity less than several percent of the eddington limit ) , and their space distribution follows the space distribution of normal x - ray bursters , we can estimate the total number of bursts from ` empty place ' per year per sky as : @xmath43 events , where @xmath44 50 is the total number of known normal bursters . if their distribution is isotropic , the number of bursts from ` empty ' places would be at least an order of magnitude greater.this simplified estimate agrees well with the bepposax estimate @xcite . they found that the total number of such sources in our galaxy is between 30 and 4000 , while the recurrence time between the bursts is a half year ( for 30 sources ) and several tens of years ( for 4000 sources ) the fluence from such events would be comparable with outbursts from rs cvn binaries , yielding about 50 - 100 events per sky in the @xmath45 region . @xcite identified a large fraction ( 6 out of 27 ) of ariel-5 s fast x - ray transients with rs cvn binary systems . another rs cvn object was proposed as a possible fxt source by @xcite . rs cvn systems are binaries formed by a cool giant or subgiant with an active corona and a less massive companion in a close synchronous orbit . rs cvn binaries typically exhibit a peak flux @xmath710@xmath46 erg / s for a duration of 1 to 10 h. outbursts from nearby rs cvn systems would have a fluence that puts them in the middle part of our diagram . this is certainly true for the identifications proposed by @xcite and @xcite . our analysis of grb x - rays indicates that this contribution to the fxt log n - log s is most significant around 100 - 1000 crab@xmath47 . heise et al . ( 2000 ) reported that about 70% of all fxts detected with bepposax are grb - related , i.e. , either conventional grbs or xrfs . 1 our best estimate ( xdgrb ) indicates that the grb contribution to fxts ( see @xmath2 distribution discussion ) peaks at 20 - 30% . however , taking into account the uncertainties of both studies , we consider these results as moderately consistent between themselves , and consistent with @xcite . we note that bepposax / wfc detected @xmath71.5@xmath310@xmath48 thermonuclear x - ray bursts from @xmath735 sources , but did not classify them as fxts because they had clear identifications . some of these bursts might have been identified as fxts by earlier experiments like ariel-5 . this would increase the total number of fxts and decrease the grb fraction . our estimate of grb - related fxts is also consistent with the rosat result @xcite at low fluences . we have shown ( see fig.7 ) that the contribution of classical batse grbs drops dramatically for lower fluences . other sources must dominate fxt statistics that region . @xcite found 6 of their 10 fxts to be flares from late type dme - dke stars , i.e. , m or k dwarfs with balmer lines in emission . @xcite also identified 3 of 10 heao-1 a1 fxts with flare stars . flares of dke - dme have peak fluxes of 10@xmath49 - 10@xmath46 erg s@xmath6 and duration from @xmath7minutes to @xmath7hours . their spectra have temperature @xmath71 kev , and they should be well detected by rosat . indeed , @xcite found star - like counterparts for 132 out of 141 reported fxts . we therefore suggest that the least luminous and most numerous part of log n - log s distribution is formed by nearby flare stars . there is however a fraction of lower - fluence fxts that are not identified with star - like objects . while the contribution of classical grbs should be negligible in this region , other types of progenitors may play a more important role . one of them , gamma - ray quiet grbs might arise in several ways ( e.g. * ? ? ? * ; * ? ? ? * ) . recent bepposax results have shown that some grbs are faint in the standard 50 - 300 kev batse band but bright in x - rays @xcite . out of the 53 grb - like transients detected with the bepposax wfc , 17 were not registered by bepposax grb monitor , which is sensitive in 40 - 400 kev energy region . @xcite found that 9 of such xrfs were also seen by batse and recorded as untriggered events at batse database . according to sazonov et al . ( 1998 ) 10% of watch events had detectable emission in 8 - 20 kev energy channel but no significant emission in the hard channel . our analysis ( see fig.6 ) confirms that untriggered batse events contribute only modestly to the lower fluence part of log n - log s if they follow the same @xmath2 distribution as the triggered events . but grb - like outbursts with softer spectra might be more significant at lower fluences . other exotic progenitors may contribute also . schafer , king , & deliyannis 2000 , suggested that ordinary main sequence stars like our sun may exhibit rare powerful superflares with @xmath50 - @xmath51 ergs , some of them in x - rays . however , one can not quantify their contribution until one better understand the frequency of such superflares . extragalactic sources may also generate fxt phenomena . x - ray variability is a fundamental property of active galactic nuclei . in the last decade it was found that bl lac objects ( blazars ) exhibit strong correlated variability in both x - ray and tev gamma - ray bands on short timescales of days to hours @xcite . the brightest of such events may be detected as fxts . previous searches for fxt progenitors demonstrated that we probably deal with a mix of close ( flare stars ) and extragalactic ( grb - related ) events . the sky distribution for both these populations is expected to be isotropic . this is consistent with the sky distribution of detected fxt ( fig.3 ) . however , the absense of reliable exposure maps for many observations and possible selection effects leave room for the existence among fxts of a non - isotropic population with the galactic disk or bulge distribution . the slope of the log n - log s distribution of such additional component is expected to be flatter than -1.5 for a variety of possible progenitors . the most natural candidates for an additional component of fxts are low mass x - ray binaries ( lmxbs ) . @xcite have analyzed bepposax data and found several x - ray type - i bursts from empty places , where no persistent emission was detected down to a several mcrab limit . @xcite have proposed a new class of low - luminosity bursters . @xcite have shown that all detected bursts from empty places are indeed the members of such class and follow lmxb space distribution . since lmxbs are concentrated in the galactic bulge ( see e.g. recent review by grimm , gilfanov , & sunyaev 2002 ) , any associated extra fxts should be clustered around the galactic center . however , most early surveys had difficulty in detecting fxts from this part of the sky @xcite . their localization accuracy was poor ( typically several degrees ) , and any x - ray flash from a populated region of the sky would tend to be attributed to known x - ray sources . hence , the population of lmxb with low persistent flux and emitting rare bursts would need to be be quite numerous to have been clearly detected by those instruments . emelyanov et al . ( 2001 ) have shown that ttm observations of the galactic bulge put strong constraints on the number - frequency function of low - accreting x - ray bursters indicating that the number of low - luminosity x - ray bursters @xmath3 the burst frequency from a burster is small . if the fxts that are distributed as a bulge population are fewer than several percent of all fxts , they can not be found with the usual statistical tests ( e.g. , a k - s test would not identify a bulge fraction less than 3% of the total as significant ) . the deviation from uniformity in the fxt distribution would be further masked by the poor ability of previous studies to find fxts at low latitudes ( see e.g. * ? ? ? * for the discussion of source confusion in ariel-5 survey ) . our two most reliably identified source populations , gamma - ray bursts and stellar flares , although isotropically distributed , might yield a log n - log s distribution with a slope significantly flatter than -3/2 ( as would be expected for a uniform distribution of standard candles ) . in the case of gamma - ray bursts , the distribution is known to flatten due to cosmological effects . even more important than the cosmological flattening , the luminosity distribution of fxts from grbs is flattened by the broad distribution of @xmath52 . we demonstrate below that nearby flare stars would also produce a distribution flatter than -3/2 . exosat observed 22 late - type flare stars ( dme - dke ) in the vicinity of the sun ( 2 - 20 pc ) in the 1 - 10 kev energy region . they found that these stars generated x - ray flares with total energies from 10@xmath53 to 10@xmath54 ergs and @xmath55 ( pallavicini , tagliaferri , & stella 1990 ) , e.g. with a differential frequency - flare distribution fits a power law n(e)@xmath56 . a similar distribution function has been observed for the sun , with differential frequency - flare energy distribution index @xmath57 @xcite or 2.0 @xcite , and for optical flares on the m dwarf flare star ad leo ( pettersen , coleman & evans 1984 ) . the integral distribution of x - ray flares from flare stars is : @xmath58 where @xmath59 is the differential distribution function of x - ray flares from flare stars , e is the flare energy , @xmath60 is the nearest flare star , and s = e/(4 @xmath61 ) is the measured fluence . this distribution holds for sources uniformly distributed through the space ( as is certainly within a radius of 100 pc ) , with a differential frequency- flare energy function described with power law @xmath62 . we derive : @xmath63 substituting @xmath64 , where @xmath9 is the slope of the cumulative frequency - flare energy function , we get : @xmath65 if the total flare energy dynamical range @xmath66 $ ] , is greater than the distance dynamical range @xmath67 $ ] , then for a range of detected fluences @xmath68 $ ] the slope of the detected cumulative number - fluence function @xmath69 will be the same as the slope of the integral frequency- flare energy function @xmath70 exosat data confirm that @xmath71 for at least @xmath72 ( pallavicini , tagliaferri , & stella 1990 ) . so , for example , if @xmath73 , the dynamical range where @xmath74 is @xmath75 . integrating over the full range of distances ( 2 - 100 pc ) and flare energies ( @xmath76-@xmath77 ) , we estimate that the integral fluence distribution @xmath69 from stellar flares has a slope that steepens slowly from @xmath78 to -3/2 with increasing fluence ( fig.8 ) . note that over 3 orders of magnitude in fluence , it is close to a power - law with slope @xmath79 . the intrinsic distribution of flare energy drives the observed log n - log s fluence distribution for the stellar component of fxts . the fluence ( crab@xmath3s ) in the 2 - 10 kev region is related to the flare energy as @xmath80 ( erg ) @xmath3 m ( at crab@xmath3sec ) @xmath3 @xmath81 ( at @xmath82 ) . the heao-1 experiments had a fluence detection threshold about 1 crab@xmath3s , while ariel-5 had a threshold of about 100 crab@xmath3s . as a result , heao-1 detected flares with energy @xmath83 ergs from sources located within 2 pc , and the most energetic flares from distances up to 100 pc . ariel-5 was able to detect flares with @xmath84 within 2 pc , and assuming the most energetic flares within 10 pc . taking the star density near the sun @xmath70.1 star @xmath85 , and assuming , based on the standard 4-component model of the galaxy @xcite , that stars are distributed uniformly within 100 pc of the sun , heao-1 should have seen flares from up to @xmath86 flare stars , while ariel-5 was able to detect @xmath87 flare stars . the catalog of chromospherically active stars @xcite includes 3 stars closer than 10 pc , while pallavicini , tagliaferri & stella ( 1990 ) observed 12 flare stars within 10 pc . basing on these data we accept @xmath88 as a lower limit for fraction of dme - dke stars around the sun . in this case the heao-1 experiments were able to detect x - ray flares from @xmath26 flare stars . because the heao-1 experiments were sensitive to fainter events than ariel-5 , and because flare star fxts are generally fainter than rs cvn fxts , we expect that the ratio between late - type flare stars and rs cvns should increase as we move from the ariel-5 to the heao-1 fluence band . this is consistent with their proposed identifications : heao-1 identified 10 events with dme - dke stars , while ariel-5 reported only one such identification . in summary , over a wide range of fluences , the intrinsic distribution of flare energies dominates over the geometric term , so that the -3/2 power law expected for standard candles does not apply . the shallow slope of the fxt distribution is , therefore , obtained naturally even for uniformly distributed progenitors . we have produced a composite log n - log s relationship for fast x - ray transients , by combining data from several experiments . this relation spans 4 orders of magnitude , and is fit to a power law with a slope of @xmath0 . extrapolation of this slope to lower fluences is consistent with the sensitive rosat and einstein experiments . the sources of fast x - ray transients are undoubtedly heterogeneous . the @xmath1 -1 power law must come from the summation of contributions from several different progenitor classes . both grbs and flares stars will contribute components to this distribution that are flatter than the ` isotropic ' slope of @xmath89 . the exact shape of the global fluence - frequency relationship might deviate from this simple power law , even inside the range we measured . however , due to the small number of events that we analyzed , and resulting large uncertainty in the deduced rate , a more complicated luminosity function is not justified ( see for example detailed discussion at murdoch , crawford , & jauncey ( 1973 ) ) . we use archival measurements of the x - ray and gamma - ray fluences for 134 grb to estimate the distribution of the x - ray to gamma - ray ratio . this @xmath2 distribution shows the presence of highly significant tail of x - ray rich events . extrapolating from the batse grb catalog , we find that the fraction of fast x - ray transients attributable to classical gamma - ray bursts is a function of fluence , being greatest for high fluences and dropping dramatically at the faint end . the exact fraction of grb - associated fast x - ray transients is sensitive to our assumptions about the @xmath2 distribution , but our analysis certainly shows that most of fainter fxts are not counterparts of classical gamma - ray bursts . the fraction of fxts from non - grb sources , such as magnetic flares on nearby stars , is the highest for the faintest flashes . our understanding of the fast x - ray transient phenomenon is still poor . the situation is similar to that for gamma - ray bursts before batse . there have been many fxts detected , but by a variety of instruments each with specific restrictions . the data show convincingly that fxts are ubiquitous . but the number of events detected by each particular experiment is small , and their statistical analysis is complicated . we can identify some classes of progenitor sources , but we can determine the fractional contribution of these classes only roughly . there remains room for the discovery of new types of sources that produce fast x - ray transients , but we certainly can not yet confirm the existence of new classes like `` gamma - ray quiet grbs '' or `` superflares '' at normal stars . a large uniform dataset from a dedicated , sensitive wide - field x - ray experiment ( see e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) could revolutionize our understanding of the fxt phenomenon . the greatest opportunity for discovery may lie in the faintest fxts , posing a challenge to instrument builders . the grb data were taken from the batse archive . the work of va was partially supported by russian foundation for basic research grant 01 - 02 - 17295 . kb is glad to acknowledge helpful discussions of the subject with jean int zand and john heise . va would like to aknowledge the advices from eugene churazov and marat gilfanov . differences in the sensitivity band and observational strategy significantly affect the @xmath2 ratio measured with different instruments . in our analysis we made use of all data we aware of to obtain the most instrument - independent estimate possible . we calculated the mean , median and variance of the measured distributions of @xmath2 for the ` batse ' and xdgrb samples ( table 1 ) we also fitted the distribution @xmath2 for the xdgrb and ` batse ' samples with a gaussian , and with a gaussian plus quadratic polynomial . these functions were choosen simply to match the data with a smooth curve . a gaussian alone can not fit the tail of distribution . the difference between the gaussian cores of the distribution is in the direction expected from the triggering effects , but the tail in the xdgrb distribution can not be caused by triggering effects . the inset on fig.4 shows the best - fit curves for the gaussian plus quadratic polynomial fit to the both samples , to show qualitatively how they differ . to understand the effect of triggering on the @xmath2 distribution , we consider the simplified case in which the x - ray / gamma - ray fluence ratio follows a log - normal distribution . in other words , if @xmath9 is defined as @xmath90 , @xmath9 is distributed as a gaussian . the gaussian distribution of @xmath9 is dn / d@xmath91 $ ] , where @xmath92 is the mean of the distribution and @xmath8 is its standard deviation , for events triggered in the band of @xmath93 . if the log n - log s distribution goes as @xmath94 , then the x - ray triggered distribution of @xmath9 , for the same population of events , will follow a gaussian distribution of the same width , but centered on @xmath95 . we assume that the x - ray trigger band is the same band as for @xmath96 . in the case of an isotropic , @xmath97 power law , the distribution will be displaced 1.5 standard deviations to the right when we trigger on x - rays instead of gamma rays . for trigger bands other than those used for @xmath36 or @xmath37 , the displacement will be different . this selection effect may explain the fact that watch generally selects softer grbs than batse . while batse triggers in the 50 - 300 kev band , watch triggers in the 8 - 100 kev band . however , the nominal trigger range for ginga , at 50 - 400 kev is harder even than batse . the softer @xmath2 ratio obtained by ginga may be the result of a secondary selection done by strohmayer et al . ( 1998 ) . while ginga observed about 120 grbs @xcite , only the grbs with good quality spectral data in both @xmath98- and x - ray detectors were selected for spectral analysis , possibly driving the selection of softer , x - ray bright , grbs . we can make a rough check of the selection effect by taking the power law @xmath99 from batse measurements , calculating the displacement @xmath100 , and comparing it with the displacement between the batse and the other , mostly x - ray triggered , @xmath2 distributions . we predict a displacement for @xmath101 1.3 ( derived from middle part of batse grbs log n - log s ) of 0.6 - 0.7 , compared to a displacement in range of 0.5 - 0.7 ( table 1 ) , between the @xmath2 distributions for ` batse ' and xdgrb . the direction is as expected : triggering at softer energies moves the @xmath2 distribution to larger values . we have tested whether the difference between the ` batse ' and xdgrb distributions was caused by the effect of trigger - band selection on a single distribution of grbs , or if there existed an additional x - ray rich grb population not detected by batse . to do this we have fit the @xmath2 distribution of the batse grb sample to a log - normal distribution . a maximum - likelihood estimate of the fit parameters gives @xmath102 and @xmath103 . this result shows that except of the triggering band effect there is also an additional contribution of x - ray rich events for xdgrb case . borozdin , k.n . , priedhorsky , w.c . , arefiev , v.a . , et al . 1999 , in : _ small missions for energetic astrophysics : ultraviolet to gamma - ray : los alamos , new mexico , february 1999 _ , ed . s.p . brumby _ , aip conference proc . , 499 , 20 heise , j. , & int zand , j. , 2001 , in : gamma - ray bursts in the afterglow era , proceedings of the international workshop , rome , 17 - 20 october , 2000 . ed : e. costa , f. frontera , and j. hjorth . berlin heidelberg : springer , 2001 , p. 16 . kippen , m. , woods , p. , heise , j. , int zand j. , preece , r. , & briggs , m. 2001 , in : gamma - ray bursts in the afterglow era , proceedings of the international workshop , rome , 17 - 20 october , 2000 . ed : e. costa , f. frontera , and j. hjorth . berlin heidelberg : springer , 2001 , p. 22 . zand int , j. 2001 , in : exploring the gamma - ray universe . proceedings of the 4-integral workshop , 4 - 8 september 2000 , alicante , spain . ed : b. battrick , sc.ed : a. gimenez , v. reglero & c. winkler . esa sp-459 , noordwijk : esa publications division , 2001 , p. 463 - 470 ccccccccccc case & & + & & & + & mean & median & @xmath8 & @xmath92 & @xmath8 & @xmath92 & @xmath8 & a@xmath104 & b@xmath104 & c@xmath104 + batse@xmath105 & -1.67 & -1.68 & 0.48 & -1.53 & 0.43 & -1.53 & 0.43 & 0.0 & 0.0 & 0.0 + xdgrb@xmath106 & -0.95 & -1.12 & 0.68 & -1.06 & 0.30 & -1.08 & 0.22 & -0.01 & -0.01 & 0.04 +	fast x - ray transients ( fxts ) with timescales from seconds to hours have been seen by numerous space instruments . because they occur at unpredictable locations , they are difficult to observe with narrow - field instruments . only a few hundred have been detected , although their all - sky rate is in the tens of thousands per year . we have assembled archival data from ariel-5 , heao-1 ( a-1 and a-2 ) , watch , rosat , and _ einstein _ to produce a global fluence - frequency relationship for these events . fitting the log n - log s distribution over several orders of magnitude to simple power law we find a slope of @xmath0 . the sources of fxts are undoubtedly heterogeneous , representing several physical phenomena ; the @xmath1 -1 power law is an approximate result of the summation of these multiple sources . two major contributions come from gamma - ray bursts and stellar flares . these two types of progenitors are distributed isotropically in the sky , however their individual luminosity distributions are both flatter than the -3/2 power law that applies to uniformly distributed standard candles . extrapolating from the batse catalog of grbs , we find that the fraction of x - ray flashes that can be the x - ray counterparts of gamma - ray bursts is a function of fluence . the exact fraction of grb - induced x - ray counterparts is sensitive to the @xmath2 distribution , which we estimate from available experimental measurements . certainly most fxts are not counterparts of standard gamma - ray bursts . the fraction of fxts from non - grb sources , such as magnetic stars , is greatest for the faintest fxts . our understanding of the fxt phenomenon remains limited and would greatly benefit from a large , homogeneous data set , which requires a wide - field , sensitive instrument .
impact factor was first described in 1972 and the isi journal impact factor is today the leading indicator of journal influence @xcite , despite its weaknesses @xcite . the impact factor of a journal in a given year measures the average number of citations to recent articles from articles published in the given year . the conventional two - year impact factor is calculated based on citation data from a three - year period . for example , the impact factor of journal @xmath0 in 2014 is the ratio between the number of citations from all considered journals in source year 2014 to articles published in @xmath0 in target years 20122013 and the number of articles published in @xmath0 in the target years . a five - year impact factor is calculated in a similar way with a five - year target window . the advantage with the widely used impact factor is that it is easy to calculate and explain , once the selection of journals is made . even though impact factor is not seen as a flow - based metric , it is in fact an example of a zero - order markov model of flow between journals . the simple count of citations to journals corresponds to measuring the visit frequency of a random walker that visits journals proportional to their citation counts . while the measure is widely used , a major drawback is the underlying assumption that all citations carry equal weight , irrespective of origin . several rankings have been suggested to overcome the problem with uniform citation weights @xcite . the eigenfactor score @xcite and its per - article normalized article influence score , for example , builds on the pagerank algorithm @xcite and takes advantage of the entire network of citations . generally speaking , the eigenfactor score measures the relative journal visit rate of a random walker that navigates between journals by following random citations . therefore , the eigenfactor score of a journal can be interpreted as a proxy for how often a researcher who randomly navigates the citation landscape accesses content from the journal . in this way , the eigenfactor score corresponds to a first - order markov model for evaluating journal influence . in a recursive fashion , important journals are those that are highly cited by important journals . in practice , a citation from an influential journal will be worth more than a citation from a less significant journal , because its importance is inherited from the citing journal . however , the inherited importance is aggregated across a journal and pushed further no matter where it came from . as a result , the actual inheritance structure of the article - level citation network is lost with strongest effect on multidisciplinary journals . while the main difference between impact factor and eigenfactor is that they correspond to a zero- and a first - order markov model of flow between journals , respectively , they differ in two other ways as well . first , while the conventional impact factor uses a two - year citation target window , eigenfactor uses a five - year target window by default . the extended time window was introduced because , in many fields , articles are not frequently cited until several years after publication . moreover , the eigenfactor score considers inheritance of importance _ between _ journals and therefore ignores self - citations . as a result , the incentive to boost the ranking of a journal with self - citations vanishes . in this paper , we focus on the general effects of markov order rather than specific implementations . therefore , we exclusively study rankings with five - year target windows and exclude all self - citations . to model citation flows between journals , we first aggregate article - level citation data in journals and then model the network flow with a random walk process . we construct citation flows with different amounts of memory by aggregating the citation data in networks that correspond to zero- , first- , and second - order markov models . below we in turn describe how we aggregate the data and model the flow . we use article - level citation data from thomson reuters web of science 1980 - 2013 . the data include almost one billion citations between more than 30 million articles published in about 20,000 journals . in this study , we focus on articles published in the years 20072012 and their citations to articles published in 2002 - 2007 . specifically , we are interested in articles published in 2007 , their citations to articles published in 20022006 , and citations to the articles published in 2007 from articles published in 20082012 . we need the two overlapping time windows to construct the second - order markov model . figure [ fig : schematic ] illustrates how we construct journal citation networks with different amount of memory from article - level citation data . in fig . [ fig : schematic]a , we show a schematic citation network with articles published in 11 different journals . the articles were published in three different time periods , the early target years 20022007 , the early source year 2007 , which also is the target year of the late source years 20082012 . for the zero- and first - order markov models , we used the early target and source years 20022007 , and for the second - order markov model we also included the late source years 20082012 . we excluded proceedings , but included all journals @xmath1 that received citations during the target period . for the zero - order markov model , we counted the number of citations to articles published in the early target years 2002 - 2006 from articles published in the early source year 2007 . to construct the journal network , we aggregated these citations in the journals of the cited articles . that is , each citation @xmath2 between an article published in journal @xmath3 in the early source year to an article published in journal @xmath4 in the early target years , adds a weight of one to the cited journal @xmath4 , @xmath5 . this procedure is exemplified in figs . [ fig : schematic]a and b , with articles published in the early target years in green and articles published in the early source year in blue . figure [ fig : schematic]a shows how one article published in journal 1 receives three citations , how four articles published in journal 3 receive eight citations , and how one article published in journal 4 receives two citations . for this zero - order markov network shown in fig . [ fig : schematic]b , journals are connected to other journals with weights proportional to the number of incoming citations , independent of citation source . that is , a random walk process on a zero - order markov network is memoryless such that the next step does not depend on the currently visited journal . for the first - order markov model , we aggregated the citations described above in pairs of citing and cited journals . that is , each citation between an article published in journal @xmath3 in the early source year to an article published in journal @xmath4 in the early target years adds a link weight of one between the citing and the cited journals , @xmath6 . figure [ fig : schematic]c illustrates how the 13 incoming links in the zero - order markov model have specific sources of the citing journals in the first - order markov model . accordingly , a random walk process on a first - order markov network has a one - step memory such that the next step depends on the currently visited journal . for the second - order markov model , we also included citations from articles published in the late source years . we used citation chains @xmath7 , trigrams of articles published in journal @xmath8 in the late source years that cite articles in journal @xmath3 in the early source year that in turn cite articles in journal @xmath4 in the early target years , as illustrated in fig . [ fig : schematic]a . to construct the second - order markov network , we aggregated the trigrams in memory nodes @xmath9 , such that each citation chain @xmath7 adds a link weight of one between memory nodes @xmath9 and @xmath10 , @xmath11 . that is , each journal has @xmath12 memory nodes , one for each other journal that cites it . constructed in this way , a random walk process on a second - order markov network has a two - step memory such that the next step depends not only on the currently visited journal , but also on the previously visited journal . the procedure to construct a second - order markov network above assumes that each article in the early source years only is cited by one journal in the later source years . for each citation from an article in journal @xmath3 in the early source year to an article in journal @xmath4 in the early target years @xmath13 , we identify all @xmath14 articles published in any journal @xmath8 in the late source years that cite the article in the early source year , and add a fractional link weight of @xmath15 between memory nodes @xmath9 and @xmath10 . moreover , if we can not identify a trigram @xmath7 , because the article in the early source year was never cited by an article in the late source years , we add a fractional link weight @xmath16 between memory nodes @xmath9 and @xmath10 for all @xmath12 memory nodes @xmath9 of journal @xmath3 . in this way , we obtain the first - order markov network if we aggregate the memory nodes in their respective journals . we use a random walk process on the networks with different amounts of memory to obtain the journal rankings . the random walk processes can be seen as proxies for how researchers navigate scholarly literature , as they read articles and follow citations in their search for information . in the zero - order markov model , a researcher would pick any citation and follow it to the cited journal irrespective of where the currently read article is published ( fig . [ fig : schematicwalk]a ) . in the first - order markov model , a researcher would pick a citation from any article published in the same journal as the currently read article and follow it to the cited journal ( fig . [ fig : schematicwalk]b ) . in the second - order markov model , a researcher would pick a citation from an article published in the same journal as the currently read article that is also cited by an article published in the previously visited journal and follow it to the cited journal ( fig . [ fig : schematicwalk]c ) . in this way , the random walk processes correspond to researchers with zero- , one- , and two - step memory . the first- , second- , and third order markov models are obtained with the same random walk process on the three networks with zero- , one- , and two - step memory . formally , we represent the journal visited at time @xmath17 by the random variable @xmath18 . the random walk process generates a sequence of visited journals @xmath19 . in general , the journal visited at time @xmath20 depends on the full history of the dynamic process , @xmath21 but for the processes we consider here the memory is limited . for the zero - order markov model illustrated in fig . [ fig : schematicwalk]a , the probability to step to journal @xmath4 next is simply given by the relative number of citations to that journal irrespective of the currently visited journal , @xmath22 which therefore also is the stationary solution of the zero - order markov model , @xmath23 for the first - order markov model illustrated in fig . [ fig : schematicwalk]b , the probability to step to journal @xmath4 next from journal @xmath3 is given by the relative number of citations to @xmath4 from @xmath3 , @xmath24 accordingly , the probability that the random walker visits node @xmath4 in step @xmath20 is in principle @xmath25 however , to ensure a unique solution independent of where the random walker is initiated , at a low rate @xmath26 the random walker instead moves according to the zero - order markov model , @xmath27 with stationary solution given by @xmath28 the zero - order markov step corresponds to random teleportation to journals proportional to their number of incoming citations . this link - weighted teleportation gives results that are more robust to changes in the teleportation rate @xmath26 @xcite . we use teleportation rate @xmath29 in all analyses . note that this teleportation scheme is slightly different from the one used in eigenfactor @xcite . however , unrecorded teleportation to a journal proportional to the number of articles it publishes followed by a recorded first - order markov step , as used in eigenfactor , is approximately the same as a single recorded zero - order markov step . for example , they would be identical if all articles cited the same number of articles . for the second - order markov model illustrated in fig . [ fig : schematicwalk]c , the random walker moves from memory node to memory node proportional to the link weights between the memory nodes . for example , the probability to visit memory node @xmath10 after visiting memory node @xmath9 is @xmath30 accordingly , the probability that the random walker visits memory node @xmath10 in step @xmath20 is in principle @xmath31 but to ensure a unique solution we include teleportation steps also in this process , @xmath32 here @xmath33 is given by the relative number of links to memory node @xmath10 , which is equivalent to the relative number of links between node @xmath3 and @xmath4 , @xmath34 consequently , the stationary solution is given by @xmath35 this teleportation scheme gives unbiased comparisons because journals receive the same amount of teleported flow as in the first - order markov model , @xmath36 and proportional to the stationary solution of the zero - order markov model in eq . ( [ eq : zerosolution ] ) . we obtain the nontrivial stationary solutions of eq . ( [ eq : firststationary ] ) and ( [ eq : secondstationary ] ) with the power - iteration method @xcite . for per - article rankings , analogous to the impact factor and the article influence score , we simply divide the stationary solution of a journal by the number of articles published by that journal in the early target years . for easy comparison between the rankings of the different markov models , we normalize with respect to the average journal . in this way , a ranking score of a journal larger than one tells how many times higher the stationary distribution per article is compared with the average journal . the common framework for the three ranking models makes it easy to study effects of the markov order alone . however , the common framework also means that the models studied here are not identical to the established impact factor and eigenfactor , and conclusions should be treated with care even if the differences are small . in summary , unlike impact factor , we disregard all self - links , and unlike eigenfactor , we use recorded teleportation to journals proportional to their citation counts . in this section , we show the results of comparisons between ranking scores obtained with zero- , first- , and second - order markov models . we first show comparisons of the top journals in explicit ranking lists , and then show quantitative results for ranking scores and robustness . figure [ fig : ranking ] shows the rankings of the top 20 journals obtained with the three different markov models . the ranking scores are given by the per article stationary distribution of random walkers normalized such that the average journal has score 1 , as described above . as with impact factor , review journals with few highly cited reviews have the highest rankings in all three models . they are followed by high impact multidisciplinary journals . journals that lose from the zero- to the first - order flow model also tend to lose from the first- to the second - order model , and , vice versa , journals that gain from the zero- to the first - order flow model also tend to gain from the first- to the second - order model . however , the multidisciplinary journals only gain marginally from the first- to the second - order model . for the similar ranking analysis with the less complete and more biased citation data from jstor reported in ref . @xcite , the effect on multidisciplinary journals was even stronger because leaking flow between fields did not cancel to the same degree . in any case , and as schematically illustrated in fig . [ fig : schematicwalk]c , the relative rankings show the largest change from the zero- to the first - order model . 1 . * zero - order markov * 2 . ( start_same1 ) ; 3 . ( start1 ) ; 4 . ( start2 ) ; 5 . ( start3 ) ; 6 . ( start4 ) ; 7 . ( start5 ) ; 8 . ( start6 ) ; 9 . ( start_same2 ) ; 10 . ( start7 ) ; 11 . ( start8 ) ; 12 . ( start_removed1 ) ; 13 . ( start9 ) ; 14 . ( start10 ) ; 15 . ( start11 ) ; 16 . ( start12 ) ; 17 . ( start13 ) ; 18 . ( start14 ) ; 19 . ( start15 ) ; 20 . ( start_removed2 ) ; 21 . ( start16 ) ; 22 . ( zero_to_first_new ) ; 1 . * first - order markov * 2 . ( start_same1end ) ; ( start_same3 ) ; 3 . ( start6end ) ; ( start_same4 ) ; 4 . ( start8end ) ; ( start17 ) ; 5 . ( start9end ) ; ( start18 ) ; 6 . ( start7end ) ; ( start19 ) ; 7 . ( start10end ) ; ( start20 ) ; 8 . ( start4end ) ; ( start21 ) ; 9 . ( start_same2end ) ; ( start22 ) ; 10 . ( start1end ) ; ( start23 ) ; 11 . ( start3end ) ; ( start24 ) ; 12 . ( start12end ) ; ( start25 ) ; 13 . ( start2end ) ; ( start26 ) ; 14 . ( start5end ) ; ( start27 ) ; 15 . ( start13end ) ; ( start28 ) ; 16 . ( start_new1 ) ; ( start29 ) ; 17 . ( start14end ) ; ( start_same5 ) ; 18 . ( start15end ) ; ( start30 ) ; 19 . ( start11end ) ; ( start_removed3 ) ; 20 . ( start16end ) ; ( start_removed4 ) ; 21 . ( start_new2 ) ; ( start31 ) ; 22 . ( first_removed_and_new1 ) ; ( first_removed_and_new2 ) ; 1 . * second - order markov * 2 . ( start_same3end ) ; 3 . ( start_same4end ) ; 4 . ( start18end ) ; 5 . ( start17end ) ; 6 . ( start20end ) ; 7 . ( start23end ) ; 8 . ( start19end ) ; 9 . ( start21end ) ; 10 . ( start22end ) ; 11 . ( start25end ) ; 12 . ( start24end ) ; 13 . ( start28end ) ; 14 . ( start29end ) ; 15 . ( start26end ) ; 16 . ( start27end ) ; 17 . ( start_same5end ) ; 18 . ( start_new3 ) ; 19 . ( start31end ) ; 20 . ( start30end ) ; 21 . ( start_new4 ) ; 22 . ( first_to_second_removed ) ; tikz[remember picture , overlay ] ( @xmath37 ) ( @xmath38 ) ; tikz[remember picture , overlay ] ( @xmath39 ) ( @xmath40 ) ; ( @xmath41 ) ( @xmath42 ) ; ( @xmath43 ) ( @xmath44 ) ; ( @xmath45 ) ( @xmath46 ) ; tikz[remember picture , overlay ] ( @xmath47 ) ( @xmath48 ) ; tikz[remember picture , overlay ] ( @xmath49 ) ( @xmath50 ) ; ( @xmath51 ) ( @xmath52 ) ; ( @xmath53 ) ( @xmath54 ) ; tikz[remember picture , overlay ] ( @xmath55 ) ( @xmath56 ) ; tikz[remember picture , overlay ] ( @xmath57 ) ( @xmath58 ) ; tikz[remember picture , overlay ] ( @xmath59 ) ( @xmath60 ) ; ( @xmath61 ) ( @xmath62 ) ; ( @xmath63 ) ( @xmath64 ) ; ( @xmath65 ) ( @xmath66 ) ; ( @xmath67 ) ( @xmath68 ) ; tikz[remember picture , overlay ] ( @xmath69 ) ( @xmath70 ) ; tikz[remember picture , overlay ] ( @xmath71 ) ( @xmath72 ) ; ( @xmath73 ) ( @xmath74 ) ; ( @xmath75 ) ( @xmath76 ) ; ( @xmath77 ) ( @xmath78 ) ; ( @xmath79 ) ( @xmath80 ) ; ( @xmath81 ) ( @xmath82 ) ; ( @xmath83 ) ( @xmath84 ) ; tikz[remember picture , overlay ] ( @xmath85 ) ( @xmath86 ) ; ( @xmath87 ) ( @xmath88 ) ; tikz[remember picture , overlay ] ( @xmath89 ) ( @xmath90 ) ; tikz[remember picture , overlay ] ( @xmath91 ) ( @xmath92 ) ; tikz[remember picture , overlay ] ( @xmath93 ) ( @xmath94 ) ; tikz[remember picture , overlay ] ( @xmath95 ) ( @xmath96 ) ; tikz[remember picture , overlay ] ( @xmath97 ) ( @xmath98 ) ; ( @xmath99 ) ( @xmath100 ) ; tikz[remember picture , overlay ] ( @xmath101 ) ( @xmath102 ) ; ( @xmath103 ) ( @xmath104 ) ; ( @xmath105 ) ( @xmath106 ) ; tikz[remember picture , overlay ] ( @xmath107 ) ( @xmath108 ) ; ( @xmath109 ) ( @xmath110 ) ; ( @xmath111 ) ( @xmath110 ) ; ( @xmath112 ) ( @xmath113 ) ; ( @xmath114 ) ( @xmath113 ) ; ( @xmath115 ) ( @xmath116 ) ; ( @xmath115 ) ( @xmath117 ) ; ( @xmath118 ) ( @xmath119 ) ; ( @xmath118 ) ( @xmath120 ) ; the absolute ranking scores of the top journals in fig . [ fig : ranking ] show a similar increase from zero- to first- and from first- to second - order markov dynamics . in the zero - order markov model , the five top ranking scores are about 30 times higher than the average article , in the first - order markov model they are about 40 times higher , and in the second - order markov model they are about 45 times higher . moreover , it is a trend that the higher - order markov models give wider range of scores . this effect can be explained by their non - uniform citations values ; citations from top ranked journals are worth more than citations from average ranked journals @xcite . in the second - order markov model with more detailed structural information and more specific re - distribution of flow value , the range of scores is even wider . ] figure [ fig : distribution ] shows the cumulative journal frequency and ranking scores . the cumulative ranking scores show that the top 100 journals in the zero - order markov model share 15.9% of all flow , whereas the top 100 journals in the second - order markov model share 22.7% of all flow . the first - order model is in between the other models with 21.2% of all flow . overall , the higher - order markov models show a wide range of scores from the lowest to the highest values . a method that is good in theory is of little use if the results are not robust in practice . for journal rankings , the most crucial factor is how robust the results are to the particular selection of journals included in the analysis . for journals indexed by thomson reuters web of science , the citation data are more or less complete for the indexed journals . however , only a fraction of journals are indexed and the rankings inevitably depend on the selection . therefore , we examined the robustness of the different models by performing analysis on random sub - samples of the set of all journals . we generated sub - samples that contained 90% , 80% , , 10% of all journals by randomly including the journals . since highly ranked journals are more likely to be included in practice , we complemented this uniform sampling with a proportional sampling in which we included journals proportional their citation counts . for each sub - sample size , we generated 10 samples and measured the ranking similarity between all pairs of rankings for each model . we used the normalized mutual information for rankings to measure the similarity @xcite . this measure quantifies between 0 and 1 how much information one ranking provides about the other for journals common to both rankings . results close to 1 mean that few journal pairs swap ranking order between rankings and indicate that the results are robust to the selection of journals . figure [ fig : samplerobustness ] shows that the ranking robustness to journal selection tends to increase with markov order . all models become less robust with decreasing sample sizes , but the median ranking similarities among all journals are generally higher for the first- and second - order markov models ( fig . [ fig : samplerobustness]a ) , and among top 100 journals highest for the second - order markov model ( fig . [ fig : samplerobustness]b ) . for example , for the ranking similarity among top 100 ranked journals with 90% proportionally sampled journals reported in fig . [ fig : samplerobustness]b , the second - order model is more robust than the zero- and first - order models in 89% and 84% of the comparisons , respectively . at least three factors influence the robustness : the weight of citations , how local the model is , and the range of ranking scores . in the zero - order model , all citations carry equal weight and perturbations of low - ranked journals will have as high impact as perturbations of high - impact journals . in the higher - order models , the weight of citations from low - ranked journals is reduced and has less impact on the ranking . the zero - order and also the second - order model are more local in the sense that perturbations do not propagate across the network . for the zero - order model , it is simply because only directly cited journals are affected . for the second - order model , it is because the model can capture more constrained dynamics as illustrated in fig . [ fig : schematicwalk ] and demonstrated in ref . @xcite . finally , the range of scores can influence the robustness in two ways . first , simply because any two journals tend to be farther apart in units of ranking score . second , because in the normalized mutual information between rankings we make pairwise comparisons with weights proportional to the ranking scores of the journals . since the top journals have more extreme ranking values in the higher - order models , effectively fewer journal pairs will dominate the similarity measure . all these effects together make the higher - order markov models more robust to the selection of journals included in the analysis . finally , to validate that the data are sufficient for analysis with higher - order markov models , we conducted a cross - validation test @xcite . if the data were not sufficient , the models would overfit the data and this could lead to false conclusions . for example , the ranking scores of the higher - order markov models could be different from the lower - order markov models simply because of noise in sparse data . we performed the cross - validation test by predicting movements of the random walker in the second - order markov model in eq . ( [ eq : secondordernextstep ] ) with the zero- , first- , and second - order models . first , we divided all articles in 2007 into 10 random sets and generated 10 corresponding sets of trigrams , such that aggregating them all would give the complete set of trigrams . then , in each of 10 folds , we aggregated nine sets into a training set and used the last set for validation . for each fold and each order model @xmath121 , we measured the cross - entropy @xmath122 of the probability distributions @xmath123 of the validation set and @xmath124 of the training set for the zero- , first- and second - order models @xmath125 , @xmath126 that is , we measured the cost in bits of predicting the next journal in a random walk on the validation set with transition rates obtained from the training set . we found that navigation on the validation set costs 10.1(1 ) bits with the zero - order , 9.1(1 ) bits with the first - order , and 9.2(1 ) bits with the second - order markov model fitted on the training set . thus , the two higher - order models have a clear advantage over the zero - order model . while the two higher - order models perform similarly averaged over all journals , a journal - by - journal comparison highlights their differences . the second - order model can better predict pathways through high - impact multidisciplinary journals ( see fig . [ fig : schematicwalk]c ) , and therefore gives a higher robustness for top 100 journals ( fig . [ fig : samplerobustness]b ) , at an increased risk of overfitting pathways through field - specific journals with fewer citations . to quantify this effect , we derived the ratio of the posterior probabilities of the second- to the first - order model from the cross - entropy with bayes theorem . with uniform prior on the models @xmath127 and @xmath128 , the ratio between the posterior probabilities of the two models is @xmath129 table [ tab : diff ] shows that this model probability ratio is particularly high for multidisciplinary journals such as science and nature . overall , the zero - order model underfits the data , the first - order model underfits multidisciplinary journals , and the second - order model has a tendency to overfit movements in field - specific journals with fewer citations , but succeeds in capturing movements in multidisciplinary journals . this result suggests that the best model is a combination of the first- and second - order markov model . we have shown that the robustness of flow - based rankings to the selection of included journals tends to increase with increasing markov order . lower - order rankings , of which impact factor is an example , depend more on the particular selection of journals because all citations carry equal weight , and because the range between the lowest and highest ranked journals is smaller than for higher - order models . since the decision about which journals to include is difficult to make objectively and rarely made transparently , the robustness of a ranking scheme is important . whereas our analysis indicates that higher - order models perform better , the performance gain for the second - order markov model comes at the cost of requiring more citation data over a longer time period . while rankings can have many different objectives and be subject to various constraints that would favour other ranking schemes , if the sole objective of the ranking is to accurately capture likely pathways of researchers navigating between journals , model selection shows that using the more complex models pay off . however , the first - order markov model underfits multidisciplinary journals and the second - order markov model shows a tendency to overfit journals with limited data . the results suggest that an adaptive method that combines first- , second- , and even higher - order dynamics for multidisciplinary journals could further improve the ranking . we thank s. karlsson and c. wiklander for providing the journal citation data . m.r . was supported by the swedish research council grant 2012 - 3729 . stevan harnad , tim brody , franois vallires , les carr , steve hitchcock , yves gingras , charles oppenheim , heinrich stamerjohanns , and eberhard r hilf . the access / impact problem and the green and gold roads to open access . , 30(4):310314 , 2004 .	as the number of scientific journals has multiplied , journal rankings have become increasingly important for scientific decisions . from submissions and subscriptions to grants and hirings , researchers , policy makers , and funding agencies make important decisions with influence from journal rankings such as the isi journal impact factor . typically , the rankings are derived from the citation network between a selection of journals and unavoidably depend on this selection . however , little is known about how robust rankings are to the selection of included journals . here we compare the robustness of three journal rankings based on network flows induced on citation networks . they model pathways of researchers navigating scholarly literature , stepping between journals and remembering their previous steps to different degree : zero - step memory as impact factor , one - step memory as eigenfactor , and two - step memory , corresponding to zero- , first- , and second - order markov models of citation flow between journals . we conclude that higher - order markov models perform better and are more robust to the selection of journals . whereas our analysis indicates that higher - order models perform better , the performance gain for the second - order markov model comes at the cost of requiring more citation data over a longer time period . biblabel[1]#1 . science builds on previous science in a recursive quest for new knowledge @xcite . researchers put great effort into finding the best work by other researchers and into achieving maximum visibility of their own work . therefore , they both search for good work and seek to publish in prominent journals . inevitably , where researchers publish becomes a proxy for how good their work is , which in turn influences decisions regarding hiring , promotion , and tenure , as well as university rankings and academic funding @xcite . as a consequence , researchers depend on the perceived importance of the journals they publish in . while actually reading the work published in a journal is the only way to qualitatively evaluate the scientific content , different metrics are nevertheless used to quantitatively assess the importance of scientific journals @xcite . in different ways , the metrics extract information from the network of citations between articles published in the journals . in this paper , we analyze three flow - based journal rankings @xcite that at different order of approximations seek to capture the pathways of researchers navigating scholarly literature . specifically , the metrics measure the journal visit frequency of random walk processes that correspond to zero- , first- , and second - order markov models . that is , given a citation network between journals and a random walker following the citations , movements in a zero - order model are independent of the currently visited journal , movements in a first - order model depend only on the currently visited journal , and movements in a second - order model depend both on the currently visited journal and the previously visited journal . evaluating ranking methods inevitably becomes subjective , as their objectives often are different . which method is best , the most transparent @xcite , the most difficult to game @xcite , or the one with highest predictive power @xcite ? irrespective of the specific objective , perhaps the most important criterion is nevertheless the robustness of the method @xcite . because journal rankings depend on the selection of journals included in the analysis , we compare the robustness of rankings obtained with zero- , first- , and second - order markov models with random resampling techniques . we first describe the commonly used metrics impact factor and eigenfactor , which correspond to specific implementations of zero- and first - order markov models , respectively . then we put them in the same mathematical framework and show how a second - order markov model can be devised in a similar way . we use data from thomson reuters web of science and compare the methods both qualitatively and quantitatively in terms of ranking order , ranking score distributions , and robustness .
spin - orbit coupling ( soc ) has the potential to make novel electronics applications possible,@xcite as it allows one to control the electron s spin degree of freedom through its motion . most systems of interest for such applications are nano- or mesoscopic in size , including semiconductor quantum dots,@xcite metallic nanoparticles,@xcite and quantum corrals defined on surfaces.@xcite the energy spectrum , and more generally , properties of these systems are ( typically ) described by random matrix theory ( rmt).@xcite in rmt , the system s properties are a consequence of its symmetries in the classic wigner - dyson ensembles , the key symmetries are time - reversal ( @xmath0 ) and spin - rotation ( @xmath1 ) invariance.@xcite with both @xmath0- and @xmath1-invariance , the system is described by the gaussian orthogonal ensemble ( goe ) ; soc breaks the @xmath1-invariance ( while preserving @xmath0-invariance ) , driving the system to the gaussian symplectic ensemble ( gse ) . [ systems with broken @xmath0-invariance are described by the gaussian unitary ensemble ( gue ) . ] more specifically , systems with soc are described by random @xmath2 matrices ( with @xmath3@xmath4@xmath5 ) having quaternion components @xmath6 where @xmath7 is an @xmath3@xmath8@xmath3 symmetric matrix , and the @xmath9 are @xmath3@xmath8@xmath3 antisymmetric matrices ; @xmath10 are the pauli matrices , and @xmath11 is the 2@xmath82 identity matrix . @xmath12 in eq . [ goegsematrix ] is related to the soc of the microscopic hamiltonian @xmath12=0 in the goe , while @xmath12=@xmath13 in the gse . as most nanoscale systems of interest are described by rmt,@xcite it is important to understand the regimes / behaviors which arise with soc and the properties in these regimes . in this work , we consider the spatial properties of wave functions in ( two - dimensional ) chaotic nanoscale systems with soc . the spatial properties of wave functions often determine the system s response to experimental probes , and are important for devices / applications . @xcite while other works have discussed properties / consequences of eigenvector statistics with soc,@xcite here we consider the spatial properties of eigenvectors and , in particular , how these properties evolve with the soc . in what follows , we consider the properties of the hamiltonian @xmath14 where @xmath15 is a confining and/or disorder potential . results obtained via rmt are compared with those obtained by direct simulation of eq . [ startinghammy ] for a stadium billiard.@xcite to characterize the system and understand its properties , one- and two - point distribution functions were computed in the crossover from the goe to the gse ( with increasing soc ) . in particular , it is found that excellent agreement between rmt and microscopic simulations are obtained in a mean - field " description of the ( goe - gse ) crossover ( see below ) . a key observation from our results is that correlations of wave function amplitudes are suppressed with soc . interestingly , however , these correlations play a more important role in the two - point distribution function(s ) , compared to the goe ( with vanishing soc ) . the rest of the paper is organized as follows . the description of wave function statistics in rmt and , in particular , the description of the goe - gse crossover is discussed in sec . [ rmt ] . details of our microscopic calculations namely the stadium billiard considered as well as the numerical approach employed are presented in sec . our results are presented in sec . [ res ] one- and two - point distribution functions obtained via rmt are compared with numerical results from the stadium billiard . finally , sec . [ concl ] contains a summary of our results as well as remarks on experimental consequences . in rmt , wave function correlations are governed by the functional probability distribution@xcite @xmath16 \ , . \label{fieldtheory}\ ] ] @xmath17 is the functional inverse of the two - point correlation function @xmath18 , where the angular brackets @xmath19 denote an average with respect to @xmath20 ; the parameter @xmath21 depends on the system s symmetries @xmath21=1 ( @xmath21=2 ) in the goe ( gue ) , while @xmath21=4 in the gse . [ @xmath22 is a normalization constant . ] @xmath20 is the probability that a particular energy eigenfunction with spin-@xmath1 is equal to the specified function @xmath23 . a key property of eqs . [ goegsematrix ] and [ startinghammy ] is their invariance under time - reversal ; as a result , the energy levels are two - fold degenerate the eigenstates @xmath24 are degenerate , where @xmath25 is the time - reversal operator . explicitly , @xmath26 as a consequence of this two - fold degeneracy , the wave function amplitude probed numerically and experimentally is @xmath27=@xmath28+@xmath29 . as noted above , we are interested in the regimes / behaviors which arise with soc we will not only be interested in the gse , but also in the crossover from the goe to the gse . as such , we decompose the complex wave functions @xmath30 and @xmath31 in eq . [ wavefunction ] into their real and imaginary parts . then , the wave function amplitude is parameterized as @xmath32 where the parameters @xmath33 , which satisfy the constraint @xmath34 , characterize the crossover @xmath35=1 with @xmath36=0 for @xmath37 in the goe , while @xmath36=@xmath38 ( @xmath39 ) in the gse ; in the crossover , the @xmath40 fluctuate and , hence , physical quantities must be averaged over their distribution . , @xmath41 ( from eq . [ wavefunctioncrossover ] ) , where @xmath42=@xmath43 . ( a ) @xmath44 ( b ) @xmath45 ( c ) @xmath46 ( d ) @xmath47 . notice all @xmath48 change rapidly for @xmath49 and become sharply peaked at @xmath50 for large @xmath42.,width=327 ] we obtained @xmath51 , the distribution of the @xmath40 , numerically from eq . [ goegsematrix ] by considering the various orthogonal invariants@xcite the results are shown in fig . [ fig : distribution ] . we see that the @xmath51 change rapidly for @xmath52 in particular , the @xmath51 are broad for small @xmath12 , but become sharply peaked gaussian - like for larger values of @xmath12 , moving towards @xmath36=1/2 ( @xmath53 ) with increasing @xmath12 . figure [ fig : avgvariance ] shows the variance of the @xmath40 , @xmath54@xmath55@xmath56 , as a function of @xmath12 ; the inset shows how the average values of the @xmath40 , @xmath57 , evolve with @xmath12 . we see that the variance is extremely small for larger values of @xmath12 ; even for small values of @xmath12 ( where the @xmath51 are broad and asymmetric ) , the variance does not exceed 0.03 . as noted above , physical quantities must be averaged over the @xmath51 ; however , as will be seen below , rather good results are obtained in a mean - field " description ( due to the small variances ) , similar to what has been observed in the goe - gue crossover@xcite rather good results are obtained by approximating the @xmath40 by their average values ( rather than averaging over the @xmath58 . vs. @xmath42=@xmath43 . inset : average values of @xmath40 vs. @xmath42.,width=330 ] from eq . [ fieldtheory ] , all spatial correlations can be obtained once the two - point correlation function @xmath18 is known . to determine this , we expand the wave function as @xmath59 where the two - component spinors @xmath60 and @xmath60 are eigenstates of eq . [ startinghammy ] with @xmath61 . explicitly , the eigenvalues @xmath62 and corresponding eigenstates @xmath63 are ( @xmath64 ) @xmath65 where @xmath66=@xmath67+@xmath68@xmath69 . the spectrum above describes two spin - split chiral surfaces with energy @xmath70 , shown schematically in fig . [ fig : setup]a , where @xmath71 . to compute @xmath18 , the fourier coefficients ( in eq . [ fourier ] ) are taken to be gaussian random variables having zero mean and variance given by@xcite ( @xmath72,@xmath73=+,@xmath55 ) @xmath74 writing the wave function as per eq . [ wavefunction ] and using the parameterization in eq . [ wavefunctioncrossover ] , we obtain@xcite ( @xmath68,@xmath75=1,2 ) @xmath76 where @xmath77 , \label{correlatorf } \\ g & = & \frac{1}{2 } \left [ j_1(k_+ r ) - j_1(k_- r ) \right ] . \label{correlatorg}\end{aligned}\ ] ] in eqs . [ correlatorf ] and [ correlatorg ] , @xmath78 ( @xmath79 ) is the bessel function of order-0 ( order-1),@xcite @xmath80=@xmath81 , and @xmath82 are the wave vectors associated with the chiral branches at energy @xmath70 . the physics of eq . [ conjecture ] ( and eqs . [ correlator1 ] and [ correlator2 ] ) is that the system ergodically samples the energy surfaces@xcite ( shown schematically in fig . [ fig : setup]a ) . and @xmath83 . ( b ) stadium billiard considered in this work . ( c ) spatial scan of the ldos of a typical chaotic eigenfunction.,title="fig:",width=264,height=230 ] and @xmath83 . ( b ) stadium billiard considered in this work . ( c ) spatial scan of the ldos of a typical chaotic eigenfunction.,title="fig:",height=157 ] as described above , we are interested in comparing results obtained via rmt with those obtained by direct simulation of eq . [ startinghammy ] . to this end , we have computed the local density of states ( ldos ) for a stadium billiard,@xcite where the billiard s wall was constructed with a unitary delta - function potential@xcite @xmath84 with @xmath85 parameterizing the wall ( and @xmath86 ) . the retarded green s function ( gf ) for the system , @xmath87 is computed from the dyson equation @xmath88 in this equation , @xmath89 is the free - particle gf , i.e. the gf in the absence of the corral s wall , but in the presence of soc,@xcite @xmath90 where @xmath91 and @xmath92/r$ ] , with @xmath93 and @xmath94 being hankel functions,@xcite with @xmath80=@xmath81 and @xmath95 defined as before . the ldos is then obtained from the gf via @xmath96 $ ] . the stadium billiard we consider is shown schematically in fig . [ fig : setup]b . with energy in units of @xmath97=@xmath98 and soc in units of @xmath99=1/@xmath100 , where @xmath101 is the radius of the stadium s circular cap , we have considered eigenstates with energy @xmath70@xmath102@xmath103 , and have investigated socs in the range @xmath104@xmath105@xmath106@xmath107 . [ choosing @xmath101=70 , and @xmath108=@xmath109 ( with @xmath110 being the electron s rest mass ) , one obtains @xmath99=@xmath111evm , a value consistent with e.g. electrons on an au(111 ) surface.@xcite ] a spatial scan of the ldos for a typical eigenstate considered is shown in fig . [ fig : setup]c ; from the ldos , one- and two - point distribution functions were computed , going from the goe to the gse ( with increasing @xmath106 ) . we now analyze the properties of the system , comparing results from rmt with those obtained by direct simulation of eq . [ startinghammy ] for a stadium billiard . we begin by determining the regimes which arise as function of the soc strength . to this end , we consider the one - point function @xmath112 , which is obtained from eq . [ fieldtheory ] by integrating out the degrees of freedom except at @xmath113 . using eq . [ wavefunctioncrossover ] , we obtain @xmath114 \right\ } \nonumber \label{1ptfunction } \\ & & \hspace{0.087 in } \times ~ i_0\left[\frac{\nu}{4 } \left(\frac{1}{\gamma_1 ^ 2 } - \frac{1}{\gamma_2 ^ 2}\right ) ( 1-z)\right ] i_0\left[\frac{\nu}{4 } \left(\frac{1}{\gamma_3 ^ 2 } - \frac{1}{\gamma_4 ^ 2}\right ) z \right ] , \nonumber\end{aligned}\ ] ] where @xmath115 is the modified bessel function of order - zero.@xcite this expression reduces to @xmath116 in the goe ( @xmath35=1 and @xmath117=0 for @xmath68@xmath1181 ) and @xmath119 in the gse ( @xmath36=1/2 @xmath120 ) . . from top to bottom : @xmath106=0 ( goe ) , @xmath106=0.2 , @xmath106=0.5 , @xmath106=1.5 ( gse ) . each curve has been vertically offset by one unit for clarity.,width=327 ] figure [ fig : onept ] shows numerical results for @xmath121 for different values of the soc ; the results are compared with eq . [ goegsecrossover ] in a mean - field " description , i.e. with the @xmath40 evaluated at their average values for @xmath106=0.2@xmath99 ( @xmath106=0.5@xmath99 ) , we find @xmath122 ( @xmath123).@xcite the physics of eq . [ startinghammy ] is determined by its two length scales the spin - flip length @xmath124=@xmath125 and the linear dimension of the system @xmath126 ( @xmath127 ) . figure [ fig : onept ] shows how the system evolves toward the gse as the soc is increased . in particular , we find the system to be in the gse for @xmath128 i.e. @xmath129 ; once the system is in this gse regime , the statistics do not change further as the soc is increased . we now turn to spatial correlations of eigenfunctions . we first consider the amplitude correlator @xmath130 . using the parameterization in eq . [ wavefunctioncrossover ] , we obtain @xmath131 \ . \label{doscorrelator}\end{aligned}\ ] ] notice that this reduces to @xmath132 in the goe , and to @xmath133 in the gse where @xmath134=@xmath135+@xmath136 . numerical results for @xmath137 are shown in fig . [ fig : doscorrelate ] and are compared with eq . [ doscorrelator ] , with the @xmath40 evaluated at their average values ( as before ) . we see that the maximum is larger in the goe ; more generally , the correlations decay more rapidly with soc amplitude correlations are suppressed as @xmath1-invariance is broken . . from top to bottom : @xmath106=0 ( goe ) , @xmath106=0.2@xmath99 , @xmath106=0.5@xmath99 , @xmath106=1.5@xmath99 ( gse ) . each curve has been vertically offset by 1/4 unit for clarity.,width=327 ] having determined the parameter regimes and , in particular , how large the soc must be to be in the gse , we now consider in greater detail the properties of the system in the gse . to this end , we consider the joint distribution function @xmath138 , which is obtained from eq . [ fieldtheory ] by integrating out the degrees of freedom except those at @xmath113 and @xmath139 . for the gse we obtain @xmath140 where @xmath141 is the modified bessel function of order-@xmath142;@xcite for comparison , we also consider @xmath143 in the goe@xcite @xmath144 in the above equations , @xmath145=@xmath146 and @xmath147=@xmath148 where @xmath149=@xmath150 ( @xmath149=@xmath151 ) for the gse ( goe ) . we now consider the properties and consequences of @xmath152 . we begin by considering the conditional probability @xmath153 which describes the wave function distribution at @xmath154 , provided @xmath155 . it follows from eq . [ conditional ] that correlations between fluctuations at different points depend on their amplitudes@xcite regions of high amplitude ( i.e. large @xmath156 ) are correlated over larger distances , while regions of small amplitude are correlated over shorter distances . @xmath157 for the gse is shown in fig . [ fig : conditional]a for several values of @xmath158 ; @xmath159 for the goe is shown in fig . [ fig : conditional]b for comparison , for several values of @xmath151 . , and ( b ) @xmath159 , for several values of @xmath150 ( @xmath151 ) , for @xmath160.,width=336 ] from eq . [ conditional ] , one can obtain the average @xmath161 and the mean squared fluctuation @xmath162 , where @xmath163 denotes an average with respect to @xmath164 : @xmath165 , \nonumber\end{aligned}\ ] ] where @xmath166=2 for the goe,@xcite while @xmath166=1/2 for the gse . [ as before , @xmath149=@xmath150 ( @xmath151 ) for the gse ( goe ) . ] from this , we see that fluctuations are suppressed in the gse compared to the goe . more generally , fluctuations are largest in the goe ( compared with the gue@xcite and the gse , eq . [ conditionalavg ] ) and , hence , correlations are the weakest . we now consider the distribution of the product @xmath167 , @xmath168 . @xmath169 determines a number of experimentally relevant quantities , such as the form factor in resonant scattering in complex nucleii,@xcite amplitudes in tunneling measurements , and the conductance amplitude distribution through small quantum dots.@xcite from eq . [ jointsymp ] , we obtain for the gse @xmath170 where @xmath171 is a modified bessel functions of order - zero;@xcite in the goe , we obtain @xmath172 in the gse for several values of @xmath150 . for comparison , @xmath173 is also shown for @xmath174 . inset : comparison of numerical and rmt results for @xmath175.,height=268 ] figure [ fig : twopoint ] shows results for @xmath169 for several values of @xmath150 ( @xmath151 ) for the gse ( goe ) . we see that the maximum of @xmath176 decreases with increasing @xmath150 with the tail becoming slightly longer . for comparison , @xmath173 is shown for different values of @xmath151 . we see that correlations play a more significant role in the gse indeed , except for a very small region near @xmath177=0 , @xmath173 is essentially indistinguishable from the result with @xmath151@xmath40 . this is a consequence of the fact that fluctuations are largest in the goe and correlations are the weakest . shown in the inset are numerical results for @xmath176 for @xmath80=@xmath178 in comparison with the rmt result , eq . [ proddistribute ] . to summarize , we have investigated the statistical properties of wave functions in ( two - dimensional ) chaotic nanostructures with spin - orbit interactions , focussing particularly on spatial correlations of eigenfunctions . numerical results obtained for a chaotic stadium billiard were compared with ( analytic ) results from rmt . it was found that excellent agreement between rmt and microscopic simulations are obtained in a mean - field " description of the goe - gse crossover . a key observation from our results is that correlations of wave function amplitudes are suppressed with soc . interestingly , however , these correlations with soc play a more significant role in the two - point distribution function(s ) . our results have implications for a number of systems of current interest . indeed , the effects of soc have been observed in transport through quantum dots.@xcite these effects could also be observed in quantum corrals defined on au and ag ( 111 ) surfaces,@xcite where large soc has been observed recently,@xcite especially as scanning tunneling microscopy techniques have exquisite control of positioning and correlation measurements . we acknowledge helpful conversations with h. u. baranger and p. w. brouwer . ehk acknowledges the warm hospitality of the instituto de fsica torica ( madrid ) , where most of this work was performed . this work was supported in ohio by nsf - dmr mwn / ciam and nsf - pire grants . a. m. chang , h. u. baranger , l. n. pfeiffer , k. w. west , and t. y. chang , phys . rev . lett . * 76 * , 1695 ( 1996 ) ; j. a. folk , s. r. patel , s. f. godijn , a. g. huibers , s. m. cronenwett , and c. m. marcus , phys . rev . lett . * 76 * , 1699 ( 1996 ) . c. gmachl , f. capasso , e. e. narimanov , j. u. nockel , a. d. stone , j. faist , d. l. sivco , a. y. cho , science * 280 * , 1556 ( 1998 ) ; j. u. nockel , a. d. stone , g. chen , h. l. grossman , and r. k. chang , optics lett . * 21 * , 1609 ( 1996 ) . e. n. bulgakov and a. f. sadreev , jetp lett . * 78 * , 443 ( 2003 ) ; y. alhassid and t. rupp , cond - mat/0312691 ; g. murthy and r. shankar , phys . b * 75 * , 075327 ( 2007 ) ; o. zelyak and g. murthy , phys . rev . b * 80 * , 205310 ( 2009 ) .	we investigate the statistical properties of wave functions in chaotic nanostructures with spin - orbit coupling ( soc ) , focussing in particular on spatial correlations of eigenfunctions . numerical results from a microscopic model are compared with results from random matrix theory in the crossover from the gaussian orthogonal to the gaussian symplectic ensembles ( with increasing soc ) ; one- and two - point distribution functions were computed to understand the properties of eigenfunctions in this crossover . it is found that correlations of wave function amplitudes are suppressed with soc ; nevertheless , eigenfunction correlations play a more important role in the two - point distribution function(s ) , compared to the case with vanishing soc . experimental consequences of our results are discussed .
in the study of numeration systems with non - integer base , after the golden ratio , probably the most attention was given to its first generalization , the so - called tribonacci constant @xmath5 , the real root of @xmath4 . multiplication in systems corresponding to was studied grabner et al . @xcite and later messaoudi @xcite , providing bounds on the number of fractional digits of the product of two similar results for addition were obtained in @xcite . an important characteristics of a numeration system is the so - called finiteness property ( f ) , namely that the set of finite expansions forms a ring . property ( f ) for @xmath5-expansions follows from a general result by frougny and solomyak @xcite . this has an interesting consequence for proving connectedness of certain discrete planes @xcite . numeration systems with negative non - integer base @xmath0 received a non - negligible attention since the paper [ 8 ] of ito and sadahiro in 2009 . certain arithmetic aspects seem to be analogous to those for positive base systems @xcite . however , the construction of concrete algorithms is much more complicated in the former case . this might be due to the fact that while @xmath1-expansions are lexicographicaly greatest among all the representations of a given number , the @xmath6-expansions are not extremal with respect to neither lexicographical nor alternate lexicographical ordering , which is more natural for negative base , see @xcite . for example , in positive base systems , property ( f ) has been proven for large classes of pisot numbers @xcite . on the other hand , so far the only numbers known to possess analogous property ( -f ) are negative bases @xmath7 where @xmath1 is a root of @xmath8 , @xmath9 , @xmath10 , as shown in @xcite . the main results of this paper are proving property ( -f ) for zeros @xmath11 of @xmath12 , and an explicit description of a finite state transducer performing addition and subtraction in negative tribonacci base @xmath13 . the existence of such a transducer was proved in [ 5 ] for any negative pisot base . however , the proof is not constructive , and does not provide any simple procedure to construct such a transducer for a specific base . this result has implications for the possibility of modelling certain related sets in the frame of the so called cut - and project scheme , see @xcite . we also give a precise bound on the number of fractional digits arising from addition and subtraction of @xmath14-integers . the _ @xmath6-expansions _ considered by ito - sadahiro in @xcite are an analogue of introduced by rnyi in @xcite , which use a negative base . instead of defining the expansions of numbers from @xmath15 , the unit interval @xmath16 with @xmath17 was chosen . for @xmath0 , any @xmath18 has a unique expansion of the form @xmath19 defined by @xmath20 for any @xmath18 we obtain an infinite word from @xmath21 . another analogous concept is the _ @xmath6-admissibility _ , which characterizes all digit strings over @xmath22 that are @xmath6-expansion of some number . the lexicographic condition , similar to the one by parry , was also proved in @xcite . ito and sadahiro proved that a digit string @xmath23 is _ @xmath6-admissible _ ( or , if no confusion is possible , just admissible ) if and only if it fulfills the lexicographic condition @xmath24 here , the limit is taken over the product topology on @xmath25 and @xmath26 stands for _ alternate lexicographic ordering _ defined as follows : @xmath27 in analogy with @xmath1-numeration , the alternate ordering corresponds to the ordering on reals in @xmath16 , i.e. @xmath28 . the reference digit strings @xmath29 and @xmath30 play the same role for @xmath6-expansions as rnyi expansions of unity for @xmath1-expansions . while @xmath29 is obtainable directly from the definition , the following rule ( proved in @xcite ) can be used for determining @xmath30 : @xmath31 when @xmath32 is zero of @xmath12 , we have @xmath33 the lexicographic condition is @xmath34 therefore the sequences from @xmath25 with substrings @xmath35 , or @xmath36 , are not admissible . conversely , any sequence not containing those substrings is admissible . we can now recall the definition of @xmath6-expansions for all reals . [ de_minusbetarozvoje ] let @xmath0 , @xmath37 . let @xmath38 be minimal such that @xmath39 and @xmath40 . then the @xmath6-expansion of @xmath41 is defined as @xmath42 similarly as in a positive base numeration , the set of _ @xmath6-integers _ can be defined using the notion of @xmath43 . since the base is negative , we can represent any real number without the need of a minus sign . [ de_minusbetacela ] let @xmath32 . then the sets of @xmath6-integers and of numbers with finite are defined as @xmath44 @xmath45 we say that @xmath1 has _ property @xmath46 _ if @xmath47 is a ring , in other words , if @xmath48.$ ] we will be interested the length of the _ fractional part _ of an expansion . it means that if @xmath49 with @xmath50 we denote the length of the fractional part of @xmath41 as @xmath51 we set @xmath52 if @xmath53 . the suffix @xmath54 might be omitted . it was shown in @xcite that the number system with negative base @xmath7 possesses interesting properties when @xmath32 is set to be the zero of @xmath55 in particular , @xmath56 let us recall that an analogous result for @xmath1-expansions comes from ch . frougny @xcite . we will also often work with _ @xmath6-representations_. note that a representation differs from the expansion by the fact that it does not have to be admissible or even over the same alphabet . in other words , the representation @xmath57 corresponds to the number @xmath58 for @xmath59 which can not be escaped by reading zero , width=377 ] in section [ transducers ] , transducers @xmath60 and @xmath61 are defined . the former acts on the set @xmath62 and if the input is a digit - wise sum of two @xmath14-integers @xmath63 then the output sequence is a representation of @xmath64 over @xmath65 the output sequence is not necessarily an expansion . the ito - sadahiro @xmath6-expansions are not a redundant numeration system , thus addition is not computable by a finite state transducer while respecting the admissibility condition . also , @xmath60 is not defined on the whole set @xmath66 namely , the substring @xmath67 may not be recognized ( see figure [ subgraph ] ) . nevertheless , reading @xmath67 corresponds to the occurence of forbidden factors @xmath68 in both @xmath41 and @xmath69 . thus , it is sufficient to assume that the string representing either @xmath41 or @xmath69 is admissible . the transducer @xmath61 transforms a @xmath14-representation over the alphabet @xmath70 to the alphabet @xmath71 . the symbol @xmath72 stands for @xmath73 then , given a representation of @xmath41 over @xmath71 , the representation of @xmath74 over @xmath71 can be computed as @xmath75 where @xmath76 stands for the string @xmath41 whose digits are reversed , i.e. an image under the morphism @xmath77 even though the transducers act on sequences , the output sequence can be regarded as a representation . for , from the contruction it will be clear where the fractional point lies . we denote by @xmath78 and @xmath79 the lengths of fractional parts of the output i.e. if @xmath80 @xmath81 then @xmath82 note that @xmath83 as follows from . before we prove theorem [ teorem ] , we present a technical lemma whose proof can be found in section [ transducers ] . [ lemmatecko ] let @xmath84 , @xmath85 be a digit - wise sum of two 1 . we have @xmath86 2 . if @xmath87 , then @xmath88 3 . if @xmath89 then @xmath90 has suffix @xmath91 . 4 . if @xmath92 , then @xmath93 5 . let @xmath92 and @xmath94 then @xmath95 has suffix @xmath91 . [ teorem ] let @xmath96 then @xmath97 the fact that @xmath98 is a direct consequence of item 1 of lemma [ lemmatecko ] . let us focus on @xmath99 we distinguish the cases corresponding to the values @xmath100 obviously , we have @xmath101 and if @xmath102 , then by lemma [ lemmatecko ] , item 1 , it holds @xmath103 1 . if @xmath104 then @xmath105 . by lemma [ lemmatecko ] , item 2 , we have @xmath106 , thus @xmath103 2 . let @xmath107 then by item 3 of the lemma , @xmath95 has suffix @xmath108 and we can write @xmath109 we now apply @xmath60 on @xmath110 . it holds @xmath111 and in case @xmath112 we directly obtain @xmath113 + if @xmath114 , then @xmath115 and thus @xmath116 i.e. @xmath117 here we used the representation of zero @xmath118 , that follows from the minimal polynomial for @xmath5 . + in case @xmath119 we have by item 2 of the lemma that @xmath120 or @xmath121 , and therefore @xmath122 + if @xmath123 then either @xmath124 or one of the following happens . 1 . @xmath125 or 2 . @xmath126 using the rewriting rules @xmath127 one can verify that @xmath128 also holds for all @xmath129 . if @xmath130 then the statement clearly holds for @xmath131 the cases @xmath132 can be solved analogously to the previous case . the maximum value of fractional digits arising from arithmetical operations is sometimes denoted as @xmath133 ( addition and subtraction ) and @xmath134 ( multiplication ) , cf . @xcite . in fact , theorem [ teorem ] gives @xmath135 the inequality @xmath136 is shown in example [ priklad ] . altogether , we have @xmath137 the aim of this section is to demonstrate lemma [ lemmatecko ] , which was crucial in proving theorem [ teorem ] . the main element for the demonstration are transducers @xmath60 and @xmath61 which are defined below . the idea of the construction of the adding automaton is the following : first , do a digit - wise sum of two numbers . then , read this representation over the alphabet @xmath138 from left to right , and , using the rewriting rule @xmath139 rewrite it to the alphabet @xmath65 we show that we can obtain a letter - to - letter transducer , i.e. a transducer that reads and outputs only one letter at a time . however , for simplicity , we present a transducer that is not letter - to - letter , but can be modified to such . we define the tranducer @xmath140 where * @xmath141 is the set of states ; * @xmath142 is the initial state ; * @xmath143 is the input alphabet ; * @xmath144 is the output alphabet ; * @xmath145 are the labeled edges defined in the list below . we will denote the edge @xmath146 as the transition @xmath147 the transducer can be seen as a sliding window of length 4 ( or 5 in some cases ) that changes the string inside of itself and moves to the right while putting its left most symbol to the output . one can verify that the transitions defined below do not change the numerical value of the input string since the right side is obtained by adding a representations of zero . that is , if @xmath148 then @xmath149 can be obtained from @xmath150 by adding the strings @xmath151 or @xmath152 ( possibly two times ) . moreover , with one exception , the transitions from any state are defined for any input letter from @xmath153 the exception is the state @xmath154 that can not be escaped by reading symbol @xmath155 nevertheless , the only path to the state @xmath154 leads from @xmath156 by reading the string @xmath157 on the input ( see figure [ subgraph ] ) . reading @xmath67 would mean that both @xmath41 and @xmath69 contain forbidden string @xmath68 . therefore we assume that at least one summand does not contain @xmath158 [ priklad ] we will show how @xmath60 acts on the number @xmath159 first , we add the leading factor @xmath160 and the tailing factor @xmath161 to the representation . below is the sliding window representation and on the right side are the corresponding transitions of @xmath162 input and output symbols are in bold . @xmath163&1&2&0&0&0&0&0&0&0&0&0&0&000\|2\rightarrow 0\|002\\ \b0&[0&0&2&\b1]&2&0&0&0&0&0&0&0&0&0&0&002\|1\rightarrow 1\|112 \\ 0&\b1&[1&1&2&\b2]&0&0&0&0&0&0&0&0&0&0 & 112\|2\rightarrow 1\|122 \\ 0&1&\b1&[1&2&2&\b0]&0&0&0&0&0&0&0&0&0 & 122\|0\rightarrow 1\|220\\ 0&1&1&\b1&[2&2&0&\b0]&0&0&0&0&0&0&0&0 & 220\|0\rightarrow 1\|11\i\\ 0&1&1&1&\b1&[1&1&\i&\b0]&0&0&0&0&0&0&0 & 11\i\|0\rightarrow0\|00\i\\ 0&1&1&1&1&\b0&[0&0&\i&\b0]&0&0&0&0&0&0 & 00\i\|0\rightarrow0\|0\i0\\ 0&1&1&1&1&0&\b0&[0&\i&0&\b0]&0&0&0&0&0 & 0\i0\|0\rightarrow1\|0\i1\\ 0&1&1&1&1&0&0&\b1&[0&\i&1&\b0]&0&0&0&0 & 0\i1\|0\rightarrow1\|001\\ 0&1&1&1&1&0&0&1&\b1&[0&0&1&\b0&\b0]&0&0 & 001\|00\rightarrow00\|100\\ 0&1&1&1&1&0&0&1&1&\b0&\b0&[1&0&0&\b0]&0 & 100\|0\rightarrow1\|000\\ 0&1&1&1&1&0&0&1&1&0&0&\b1&[0&0&0&\b0 ] & 000\|0\rightarrow0\|000\\ \end{array}\ ] ] hence it holds that @xmath164 the latter representation is admissible from which it follows that @xmath165 the transducer @xmath60 is defined by the following list of transitions : = 13 = 13 ( -#1)#1 `` # 1''#2- > `` # 3''#4``#5\|#6 '' ] ; @xmath166 3 automat.txt induced by reading zeros at the input ] now we define the letter - to - letter transducer @xmath167 , where * @xmath168 is the set of states ; * @xmath142 is the initial state ; * @xmath169 is the input alphabet ; * @xmath144 is the output alphabet ; * the set of transitions is defined below 3 @xmath170 + @xmath171 + @xmath172 + @xmath173 + @xmath174 + @xmath175 + @xmath176 + @xmath177 + @xmath178 + @xmath179 + @xmath180 + @xmath181 + @xmath182 + @xmath183 + @xmath184 + @xmath185 + @xmath186 + @xmath187 + @xmath188 + @xmath189 + @xmath190 + @xmath191 + @xmath192 + @xmath193 we will provide demonstration for items 1 - 5 of the lemma . 1 . let us assume that the transducer is in the state @xmath194 and the string @xmath54 is on the input . this corresponds to the situation when we start reading the fractional digits . only the paths starting in the states @xmath195 @xmath196 result in six fractional digits ( see figure [ tree ] ) . note that in this case , the last nonzero digit is @xmath197th on the output since first three output digits belong to the integer part of the output . the states listed in the proof of item 1 are accessible only by reading the digit @xmath198 . 3 . from figure [ tree ] , one can see that if @xmath91 should not be suffix of @xmath199 then the path to the state @xmath160 must go through @xmath200 . if the path starts in @xmath201 , only one fractional digit is produced . the same holds if the path goes through @xmath202 , since we must read @xmath203 in the transition @xmath204 this case corresponds to the rewriting @xmath205 we also obtain at most one fractional digit . 4 . consider transducer @xmath61 and similarly as in the proof of item 1 , inspect the paths leading to @xmath206 precisely , it takes at most six zeros on the input to reach the state @xmath160 . 5 . by the same argumentation as in the proof of item 3 , one can see that @xmath95 has the suffix @xmath91 if @xmath207 . let us study property ( -f ) of the zeros of @xmath208 such numbers belong to the class , sometimes called confluent parry numbers . let us recall that in this case @xmath209 and @xmath210 for any sequence @xmath211 with finitely many nonzero elements . it was shown in @xcite that the zeros of @xmath212 do not possess property ( -f ) , hence we focus only on the case @xmath213 property ( -f ) for @xmath5 already follows from section [ transducers ] . for if @xmath54 is a suffix of the input sequence , the transducer reaches the state @xmath160 where it has a cycle @xmath214 the closeness of @xmath47 under multiplication follows trivially from the closeness under addition . for any @xmath215 , we can write @xmath216 where @xmath217 is the number of nonzero digits in a representation of @xmath69 . since it holds @xmath218 for all @xmath219 it suffices to show that @xmath220 for any @xmath221 note that @xmath222 , and @xmath223 trivially if @xmath224 hence we only need to show that @xmath225 has a finite representation . in order to show that @xmath223 , we again construct an automaton rewriting a representation of @xmath226 to the alphabet @xmath227 the following reasoning is helpful for an easy construction . we make use of the principle of the `` sliding window '' moving from left to right , see example [ priklad ] . in each step , a representation of zero is added . let us recall that it follows from the minimal polynomial of @xmath1 , that the digitwise addition of strings @xmath228 does not change the value of a representation . moreover , we do not consider output digits as long as they belong to @xmath22 . the initial setting of the window will be such that it contains the digit @xmath229 . obviously , the automaton can stop anytime when the digits in the window belong to @xmath22 . the states of the automaton are divided into several sets @xmath230 listed below , not necessarily disjoint . in each set of states , we consider all the choices @xmath231 that satisfy additional conditions . these conditions are of two types . conditions may not be satisfied , however , if they are not fulfilled , we just found a finite representation over @xmath22 and the automaton stops . the other conditions are either assumptions of the initial configuration , or inherited from the previous states . we also assume that the input digit is any member of @xmath153 for a given representation of @xmath232 we have the following initial states : 1 . if @xmath233 , the initial state lies in @xmath234 . 2 . if @xmath235 , the initial state lies in @xmath236 . 3 . if @xmath237 then it necessarily holds @xmath238 , otherwise @xmath41 is a non - admissible representation of @xmath41 . in case @xmath239 we have @xmath240 and we found a finite representation of @xmath241 in other case we have @xmath242 and we rewrite @xmath243 which corresponds to a state in @xmath244 let @xmath32 be the zero of @xmath245 consider the representation @xmath246 first , according to item 3 above , we rewrite @xmath247 , hence we start in the state @xmath248 @xmath249&0&1&2&0&0&q_\mathrm i\xrightarrow\m q_\mathrm{ii}\subseteq q_4\\ 2&2&0&1&[1&3&\overline 2&\b0]&1&2&0&0&q_4\xrightarrow{\m } q_5\\ 2&2&0&1&0&[1 & 2&{\overline 2}&\b1]&2&0&0&q_5\xrightarrow{\m } q_7\\ 2&2&0&1&0&0 & [ 0&0&\i&\b2]&0&0&q_7\xrightarrow{0000 } q_8\\ 2&2&0&1&0&0 & 0&[0&\i&2&\b0]&0&q_8\xrightarrow{\p } q_9\\ 2&2&0&1&0&0 & 0&1&[1&0&2&\b0]&q_9 : ( y = m\wedge z=0)\rightarrow\mathrm{\bf{stop}}\\ \end{array}\ ] ] note that @xmath69 and @xmath110 are not the values in the window , they are the values in the original representation . now we define the transducer by the set of states @xmath230 together with the transitions and corresponding rewriting rules . note that several times it holds @xmath250 for instance for @xmath251 we treat them separately for the sake of simplicity of @xmath252 one can see that the output stays in @xmath22 . it remains to show that the automaton stops for any entry with suffix @xmath54 . indeed , there is a unique loop @xmath253 , and ultimately , the condition @xmath254 in @xmath255 is not fulfilled . we have thus proved the following theorem . let @xmath32 be the zero of @xmath256 then @xmath47 is a ring if and only if @xmath257 we extended a class of numbers that are known to possess property ( -f ) . it seems that this property is satisfied also by the zeros of of arbitrary odd degree , in case that @xmath213 an approach described in section [ sekcepet ] might lead to obtaining such result . it would be nice to find a more general class satisfying property ( -f ) , somehow comparable for instance with the bases given in @xcite or @xcite . the author would like to thank to zuzana maskov for many useful comments . this work was supported by the grant agency of the czech technical university in prague , grant no . sgs14/205/ohk4/3t/14 and czech science foundation grant 13 - 03538s .	we study arithmetical aspects of ito - sadahiro number systems with negative base . we show that the bases @xmath0 , where @xmath1 is zero of @xmath2 possess the so - called finiteness property . for the tribonacci base @xmath3 zero of @xmath4 , we present an effective algorithm for addition and subtraction . in particular , we present a finite state transducer performing these operations . as a consequence of the structure of the transducer , we determine the maximal number of fractional digits arising from addition or subtraction of two
there has been a lot of recent interest in the chaotic dynamics of nonlinear systems that possess invariant subspaces ; a number of quite subtle dynamical effects come to light in examining the interaction of attractors with the invariant subspaces . moreover a good understanding of these effects is essential in interpreting and predicting dynamics of simulations and experiments where the presence of discrete spatial symmetries implies the existence of invariant subspaces . a fundamental bifurcation in such a setting is the blowout bifurcation where a chaotic attractor within an invariant subspace loses stability to perturbations transverse to the invariant subspace . in doing so , can either create a nearby ` branch ' of chaotic attractors or lose stability altogether . in contrast with this , the presence of invariant subspaces can lead to the existence of what have been called robust heteroclinic cycles between equilibria , that is , heteroclinic cycles that are persistent under small perturbations that preserve the symmetry . these cycles may or may not be attracting @xcite . recently it has been recognised that cycles to more complicated invariant sets can also occur robustly in symmetric systems , in particular to chaotic invariant sets ; this behaviour was named ` cycling chaos ' in a recent paper by dellnitz _ _ and has been further investigated by field @xcite and ashwin @xcite . in this paper we find there is a connection between these dynamical properties ; we show a scenario where a blowout bifurcation creates an attracting ` cycling chaotic ' state in a bifurcation that is analogous to a saddle - node homoclinic bifurcation with equilibria replaced by chaotic invariant sets . we also investigate how the attracting cycling chaotic state that is created in the blowout bifurcation loses stability at a resonance of liapunov exponents . ( a _ resonance bifurcation _ in its simplest form occurs when a homoclinic cycle to an equilibrium loses attractiveness ; this occurs when the real parts of eigenvalues of the linearisation become equal in magnitude . ) in spite of the system being neither a skew product nor being a homoclinic cycle to a chaotic set as in @xcite we see similar behaviour and can predict the loss of stability by looking at a rational combination of liapunov exponents . we find this scenario of a blowout bifurcation to cycling chaos is a mechanism for transition from stable two - dimensional to fully three - dimensional magnetoconvection . the model we study is a galerkin truncation for magnetoconvection in a region with square plan , subject to periodic boundary conditions on vertical boundaries and idealised boundary conditions on horizontal boundaries . phenomenologically speaking we see a change from a chaotically varying two - dimensional flow ( with trivial dependence on the third coordinate , and which comes arbitrarily close to a trivial conduction state ) to an attracting state where trajectories spend increasingly long times near one of two symmetrically related two - dimensional flows interspersed with short transients . we explain and investigate this transition in terms of a blowout bifurcation of a chaotic attractor in an invariant subspace . in the paper of ott and sommerer that coined the phrase ` blowout bifurcation ' , two scenarios are identified . either the blowout was _ supercritical _ in which case it leads to an _ on - off intermittent _ state , or it is _ subcritical _ and there is no nearby attractor after the bifurcation . we find an additional robust possibility for bifurcation at blowout . near this transition the three - dimensional flow patterns show characteristics of intermittent cycling between two symmetrically related ` laminar ' states corresponding to two - dimensional flows , but the time spent near the laminar state is , on average , infinite . this suggests that the blowout is supercritical , but in a weaker sense which we make precise . namely , we say a blowout is _ set supercritical _ if there is a branch of chaotic attractors after the blowout whose limit contains the attractor in the invariant subspaces before the blowout . in particular there may be other invariant sets contained in this limit and so any natural measures on the bifurcating branch of attractors ( if they exist ) need not limit to the natural measure of the system on the invariant subspace . we also show that the attractors corresponding to two - dimensional flows are not liapunov stable , but are milnor attractors near the transition to three dimensions , and so in particular we expect the presence of noise to destabilise two - dimensional attractors near blowout by a _ bubbling _ type of mechanism @xcite . we find in our example that the state of cycling chaos is attracting once it has been created : trajectories cycle between neighbourhoods of the chaotic sets within the invariant subspaces , and the time between switches from one neighbourhood to the next increases geometrically as trajectories get closer and closer to the invariant subspaces . by estimating the rate of increase of switching times , we are able to show that cycling chaos ceases to be attracting in a resonance bifurcation . one remarkable aspect of this study is that we are able to predict the parameter values at which the blowout bifurcation and the resonance occur , requiring only a single numerical average over the chaotic set within the invariant subspace . the paper is organised as follows : in section [ secmodel ] we introduce the ode model for magnetoconvection , discuss its symmetries and corresponding invariant subspaces . this is followed by a description of the creation , persistence and loss of stability of the cycling chaos on varying a parameter in numerical simulations in section [ secnumerics ] . section [ secmap ] shows how one can , under certain assumptions , derive a map model of the dynamics of the ode that has the same dynamical behaviour . section [ secblowout ] is a theoretical analysis of the blowout bifurcation that creates the cycling chaotic attractor and is followed in section [ secresonance ] by a theoretical analysis of its loss of stability . finally , section [ secdiscuss ] discusses some of the implications of this work on the chaotic dynamics of symmetric systems . the model we study is an ode on @xmath1 described by the following equations @xmath2 these odes have been derived as an asymptotic limit of a model of three - dimensional incompressible convection in a plane layer , with an imposed vertical magnetic field ; for further details and details of its derivation , see @xcite . in the context of this model , @xmath3 and @xmath4 represent the amplitudes of convective rolls with their axes aligned in the @xmath5 and @xmath6 ( horizontal ) directions respectively , @xmath7 and @xmath8 represent modes that cause the rolls to tilt , and @xmath9 and @xmath10 represent shear across the layer in the @xmath6 and @xmath5 directions . the modes @xmath11 and @xmath12 represent the horizontal component of the magnetic field in the @xmath6 and @xmath5 directions , and @xmath13 represents the horizontally averaged temperature . the model has are five primary parameters : @xmath14 is proportional to the imposed temperature difference across the layer , with @xmath15 at the initial bifurcation to convection ; @xmath16 is related to the horizontal spatial periodicity length , but is an arbitrary small parameter in the model of @xcite ; @xmath17 and @xmath18 are dimensionless viscous and magnetic diffusion coefficients ; and @xmath19 is proportional to the square of the imposed magnetic field . note that @xmath17 , @xmath18 and @xmath19 are scaled by factors of @xmath20 , @xmath20 and @xmath21 from their usage in @xcite . two secondary parameters that we use are @xmath22 and @xmath23 . in the parameter regime of interest , all parameters are non - negative . consider @xmath24 acting on the plane with unit cell @xmath25 in the usual way , with the torus @xmath26 acting by translations on the plane , and @xmath27 by reflections in the axes and rotation through @xmath28 . we define the following group elements @xmath29 note that @xmath30 , @xmath31 and any reflection @xmath32 can be used to generate the group @xmath24 . .selected fixed point subspaces @xmath33 of the action of @xmath34 on @xmath1 together with name , representative point and dimension of @xmath33 . there are others ( e.g. @xmath35 ) but these are not important for the dynamics we discuss here . [ cols="<,<,<,^",options="header " , ] ( 50,60 ) ( 25,0)(0,0)[bc]@xmath36 ( 18,14)(0,0)[bc]@xmath37 ( 32,14)(0,0)[bc]@xmath38 ( 10,28)(0,0)[bc]@xmath39 ( 25,28)(0,0)[bc]@xmath40 ( 40,28)(0,0)[bc]@xmath41 ( 3,42)(0,0)[bc]@xmath42 ( 18,42)(0,0)[bc]@xmath43 ( 32,42)(0,0)[bc]@xmath44 ( 47,42)(0,0)[bc]@xmath45 ( 25,56)(0,0)[bc]@xmath46 ( 18,12)(2,-3)5 ( 32,12)(-2,-3)5 ( 10,26)(2,-3)5 ( 25,26)(2,-3)5 ( 40,26)(-2,-3)5 ( 37,26)(-2,-1)15 ( 3,40)(2,-3)5 ( 18,40)(2,-3)5 ( 33,40)(2,-3)5 ( 47,40)(-2,-3)5 ( 5,40)(4,-1)30 ( 22,40)(2,-1)15 ( 22,54)(-2,-1)15 ( 24,54)(-2,-3)5 ( 26,54)(2,-3)5 ( 28,54)(2,-1)15 we consider the subgroup @xmath47 since @xmath34 contains the subgroup @xmath48 , generated by @xmath49 and @xmath50 , of translations @xmath26 , it follows that @xmath34 is isomorphic to a semidirect product @xmath51 ( @xmath52 ) . the ode ( [ eqode ] ) is equivariant under the group @xmath34 of symmetries acting on @xmath1 by @xmath53 this action gives rise to a number of isotropy types , shown in table [ tabiso ] . figure [ figlattice ] gives a partial isotropy lattice for this group action . dynamics in @xmath46 always decays to the trivial equilibrium point , corresponding to the absence of convection . we refer to dynamics in @xmath39 and @xmath40 as two - dimensional , since these correspond to two - dimensional convection in the original problem ( though @xmath54 is 5 ) . dynamics in @xmath42 and @xmath43 corresponds to mirror symmetric two - dimensional rolls with their axes aligned along the @xmath5 and @xmath6 directions ; we refer to equilibrium points in these subspaces as @xmath6-rolls and @xmath5-rolls respectively . in @xmath39 and @xmath40 , convection is two - dimensional but not mirror symmetric , and is referred to as tilted rolls . dynamics in @xmath37 and @xmath38 corresponds to three - dimensional convection that is still invariant under one mirror reflection ( tilted rolls with a cross - roll component ) . otherwise we say the dynamics is fully three dimensional . in a slight break from convention we say the fixed point subspaces @xmath33 as having isotropy subgroup @xmath55 rather than considering the isotropy subgroups as the fundamental objects . we present numerical simulations of the odes that demonstrate two aspects of cycling chaos that we seek to explain : how cycling chaos can be created in a blowout bifurcation , and how cycling chaos can cease to be attracting . we concentrate on parameter values that are known to have lorenz - like chaotic dynamics within @xmath39 and @xmath40 : @xmath56 ( and hence @xmath57 and @xmath58 ) , although we note that qualitatively similar attractors are found for a large proportion of nearby parameter values . these parameter values correspond to those in figures 15(c ) and 20(a ) in @xcite . the numerical method used was a bulirsch stoer adaptive integrator @xcite , with a tolerance for the relative error set to @xmath59 for each step . we use the parameter @xmath16 as a normal parameter ( see @xcite ) for the dynamics in @xmath39 and @xmath40 ; that is , it controls instabilities transverse to @xmath39 and @xmath40 in the directions @xmath37 and @xmath38 without altering the dynamics inside @xmath39 or @xmath40 . in figure [ figphaseportraits ] we show a typical example of timeseries when there is attracting cycling chaos ( with @xmath60 ) . the system starts with @xmath3 oscillating chaotically while @xmath4 is quiescent , switches to a state where @xmath4 oscillates chaotically while @xmath3 is quiescent , and so on . a more careful examination reveals that after a switch the trajectory remains close to a fixed point in @xmath42 or @xmath43 for an increasing length of time . physically , this corresponds to chaotic two - dimensional convection that switches between rolls aligned in the @xmath6 and @xmath5 directions . figure [ figphaseportraits](c ) shows the chaotic trajectories projected onto the @xmath61 plane , while ( d ) illustrates switching between @xmath39 ( the ` horizontal ' plane ) and @xmath40 ( the ` vertical ' plane ) . note how the chaotic behaviour in figure [ figphaseportraits](a ) and ( b ) repeats : trajectories spend longer and longer near the unstable manifolds of the @xmath6-roll and @xmath5-roll equilibrium points and take longer and longer between each switch . this is illustrated further in figure [ figslowingdown ] , which shows intersections of a trajectory with the poincar surface @xmath62 close to the trivial solution . there are two phases evident in the cycle : the order one chaotic behaviour of @xmath3 near @xmath39 ( while @xmath4 grows exponentially ) , and the exponential growth of @xmath63 as the trajectory moves away from @xmath40 ( while @xmath4 behaves chaotically ) . the time between switches increases monotonically and the rate of increase varies with @xmath16 , the normal parameter . in these numerical simulations , the switching time saturates when certain components of the solution come close to the machine accuracy of the computations ( about @xmath64 for double precision ) . we argue that this is evidence for attracting cycling chaos : trajectories approach a structurally stable heteroclinic cycle between chaotic sets . in figure [ figcycle ] we show a schematic picture of the heteroclinic cycle . we recall that the fixed point spaces @xmath42 and @xmath43 are 2 dimensional , @xmath41 is 3 dimensional , @xmath39 and @xmath40 are 5 dimensional and @xmath37 and @xmath38 are 6 dimensional invariant subspaces in the 9 dimensional phase space . the system starts near the @xmath6-roll equilibrium point in @xmath42 , which is unstable to shear ( @xmath9 ) ; the parameters are such that the unstable manifold of @xmath6-rolls comes close to the trivial solution and returns to a neighbourhood of @xmath6-rolls . this global near - connection within @xmath39 is the source of the chaotic behaviour . we refer to the chaotic sets in @xmath39 and @xmath40 as @xmath65 and @xmath66 respectively ; these contain the relevant roll equilibrium points and they both contain the trivial solution , so there are structurally stable connections from the trivial solution to the roll equilibrium points and from those to @xmath65 and @xmath66 . within @xmath37 , @xmath65 is unstable to cross - rolls ( @xmath4 ) since the trivial solution is equally unstable in the @xmath3 and @xmath4 directions . eventually @xmath4 grows large enough that there is a switch to the @xmath5-roll equilibrium point in @xmath43 , at which point @xmath10 starts to grow . thus the cycle connects invariant sets in the following fixed point subspaces : @xmath67 between the equilibrium points in @xmath42 and @xmath43 , and @xmath65 and @xmath66 within @xmath39 and @xmath40 , with the structurally stable connections needed to complete the cycle from @xmath65 and @xmath66 to the @xmath5-roll and @xmath6-roll equilibrium points lying within @xmath37 and @xmath38 . note that our scenario is certainly a simplification of the full set of heteroclinic connections ; there are other fixed points contained in the closure of the smallest attracting invariant set , in particular the origin is contained within the cycle . we turn now to the question of how the cycling chaos is created . with @xmath68 we find attracting two - dimensional chaos ( figure [ figblowout]a , b ) , which loses stability around @xmath69 ( c ) in a blowout bifurcation , and for @xmath70 ( d ) , there is exponential growth away from @xmath39 into @xmath37 . within @xmath37 , the @xmath5-roll equilibrium points are sinks , establishing the structurally stable connection from the chaos in @xmath39 to the equilibrium point in @xmath43 , and hence the creation of cycling chaos . as illustrated in figures [ figslowingdown ] , as @xmath16 decreases towards @xmath71 , trajectories spend progressively longer in each visit to @xmath39 and @xmath40 before switching to the conjugate chaotic invariant set . eventually , trajectories come arbitrarily close to the invariant subspaces @xmath39 and @xmath40 ( limited only by machine accuracy in the numerical simulations ) . figure [ figresonance ] shows how the time intervals between switches between @xmath65 and @xmath66 increases as the system approaches this heteroclinic cycle , and how the rate of approach to the heteroclinic cycle decreases as @xmath16 approaches @xmath72 . the intervals between switches grow by a factor of about @xmath73 , @xmath74 and @xmath75 per switch for @xmath76 , @xmath77 and @xmath78 respectively , and @xmath79 and @xmath80 for @xmath81 and @xmath71 . by @xmath82 , the heteroclinic cycle is no longer attracting , and for @xmath83 and @xmath84 , the system settles down to periodic behaviour that is bounded away from @xmath39 and @xmath40 , though the periodic orbits are actually quite close to these invariant subspaces . for these calculations , we imposed a cut - off of @xmath85 : the calculation ceased once any variable became smaller than this . in the next section , we derive a map that allows us to compute longer trajectories more accurately , and we demonstrate cycling chaos , its creation in a blowout bifurcation and its loss of attractiveness using this map . we analyse the blowout bifurcation in section [ secblowout ] , and argue in section [ secresonance ] that the cycling chaos created in that bifurcation ceases to be attractive at a resonance . in this section , we discuss a map that models the behaviour of the odes in the parameter regime described above , and that helps clarify what happens near blowout bifurcation and the resonance of the cycling chaos . we rely on a map derived for the two - dimensional dynamics by breaking up the flow into pieces near and between equilibrium points @xcite . within the @xmath39 subspace , the leading stable eigendirection of the origin ( that is , the eigendirection with negative eigenvalue closest to zero ) is in the @xmath86 plane , and the unstable direction of the origin is along the @xmath3 axis . all trajectories leaving the origin in that direction follow the structurally stable connection within @xmath42 to one or other of the @xmath6-roll equilibrium points . the one - dimensional unstable manifold of @xmath6-rolls lies within @xmath39 , and the chaotic attractor is associated with a global bifurcation at which that unstable manifold collides with the origin . near this global bifurcation , the dynamics within @xmath39 is modelled by an augmented lorenz map : @xmath87 defined as a map from the surface of section @xmath88 to itself . details of the derivation are given in @xcite , but briefly , @xmath32 is a parameter related to @xmath14 in the odes , with @xmath89 at the global bifurcation ; @xmath90 is a small positive constant that we scale to one ; @xmath91 if @xmath92 and @xmath93 if @xmath94 ; @xmath95 depends on the ratio of various stable and unstable eigenvalues at the origin and the @xmath6-roll equilibrium points ( with @xmath96 ) ; and @xmath97 is a ( negative ) constant . it is a straightforward matter to include the effect of a small perturbation in the @xmath4 and @xmath10 directions . near the @xmath39 subspace , @xmath4 and @xmath10 will grow linearly at a rate that depends on @xmath3 . this means that we get a mapping of the form @xmath98 where @xmath99 and @xmath100 are constants and the exponents @xmath101 and @xmath102 again depend on ratios of eigenvalues , with @xmath103 and @xmath104 . the exponent @xmath101 is negative since a small value of @xmath3 implies that the trajectory spends longer near the origin , so @xmath4 has a longer time to grow . similarly , near @xmath40 we get the mapping @xmath105 as long as @xmath3 is much smaller than @xmath4 . these maps are valid provided that the trajectory remains close to the @xmath39 or @xmath40 subspaces . we model the switch from behaviour near @xmath39 to near @xmath40 by assuming that ( [ eqnearpx ] ) holds provided that @xmath106 ; otherwise , the trajectory leaves the neighbourhood of the origin along the @xmath4-axis , visits a @xmath5-roll equilibrium point and returns to the surface of section @xmath107 near the origin : @xmath108 where , as above , @xmath109 , @xmath110 and @xmath111 are constants and @xmath112 , @xmath113 and @xmath114 are ratios of eigenvalues . similarly , ( [ eqnearpy ] ) holds if @xmath115 ; otherwise the trajectory switches from @xmath40 to @xmath39 : @xmath116 then in the full map , as long as @xmath106 , the trajectory behaves chaotically under ( [ eqnearpx ] ) ( near the @xmath39 subspace ) , with @xmath3 obeying a lorenz map and @xmath4 growing or decaying according to the value of @xmath3 . if @xmath4 grows sufficiently that @xmath117 , there is one iterate of map ( [ eqswitchpxpy ] ) , which makes @xmath3 small as the system switches from @xmath39 to @xmath40 , followed by many iterates of ( [ eqnearpy ] ) , and one iterate of ( [ eqswitchpypx ] ) as the system switches back to @xmath39 . the @xmath118 exponents can be determined from the eigenvalues of the trivial solution and of @xmath6-rolls . the relevant eigenvalues of the origin are ( in the notation of @xcite ) the growth rate @xmath14 of @xmath3 and @xmath4 and the slowest decay rate @xmath119 of @xmath86 , determined by the eigenvalue of @xmath120\ ] ] closest to zero . the relevant eigenvalues of @xmath6-rolls are the growth rate @xmath121 and the slowest decay rate @xmath122 of @xmath123 , determined by the eigenvalues of @xmath124.\ ] ] the other important eigenvalues at the @xmath6-roll equilibrium point are the decay rate of @xmath4 ( @xmath125 ) and the decay rate of @xmath10 ( @xmath119 ) . then we have @xmath126 from fitting the map to trajectories of the odes with @xmath69 ( see figure [ figblowout]b ) , we find values for the constants : @xmath127 the eigenvalues that do not depend on @xmath16 are : @xmath128 with these eigenvalues , we have @xmath129 , @xmath130 , @xmath131 and @xmath132 , while @xmath101 ( negative ) and @xmath112 depend on @xmath16 . we seek to reproduce in the map what we have observed in the odes : the blowout bifurcation that creates cycling chaos , and the loss of attractiveness of cycling chaos . the first of these ( figure [ figmapone ] ) is straightforward : typical lorenz - like chaos is shown in figure [ figmapone](a ) , while the change from decay towards @xmath39 to growth away from @xmath39 at @xmath133 is shown in ( b d ) . cycling chaos is found after the blowout bifurcation ( figure [ figmaptwo]a ) : the system switches between chaos in @xmath39 and chaos in @xmath40 , getting closer and closer to the invariant subspaces after each switch and spending longer between switches ( figure [ figmaptwo]b ) . the rate of increase of the intervals between switches depends on @xmath16 , and is about a factor of 1.01 per switch for @xmath134 . decreasing @xmath16 to @xmath135 ( figure [ figmaptwo]c , d ) results in growth away from cycling chaos and saturation to a periodic orbit . these calculations required formulating the map ( [ eqnearpx][eqswitchpypx ] ) in terms of the logarithms of the variables in order to cope with the large ( @xmath136 ) dynamic range . the one place in which accuracy is inevitably lost is in the switch from @xmath39 to @xmath40 ( and back ) , using ( [ eqswitchpxpy ] ) : the @xmath137 term is swamped by the order one @xmath32 . as a result of this , the chaotic trajectories start in exactly the same way each time the system switches from one invariant subspace to the other . in terms of the odes , the trajectory entering a neighbourhood of @xmath39 close to @xmath6-rolls shadows the unstable manifold of that equilibrium point ( lying inside @xmath39 and leading to the chaotic set @xmath65 ) until the switch to @xmath40 . this is in agreement with the ode behaviour shown in figure [ figphaseportraits ] . we briefly review some definitions . if @xmath138 is a compact flow - invariant subset then we define the unstable set @xmath139 and the stable set ( or _ basin of attraction _ ) @xmath140 where @xmath141 ( resp . @xmath142 ) is the limit set of a trajectory of the ode passing through @xmath6 in the limit @xmath143 ( resp . @xmath144 ) . we say a compact invariant set @xmath138 is an _ attractor _ in the sense of milnor if @xmath145 where @xmath146 is lebesgue measure on @xmath1 . it is said to be a _ minimal milnor attractor _ if there are no proper compact invariant subsets @xmath147 with @xmath148 @xcite . as shown in , milnor attractors can occur in a robust manner if the attractor lies within an invariant subspace . suppose @xmath149 is an invariant subspace and @xmath138 is a compact invariant set in @xmath149 such that @xmath138 is a minimal milnor attractor for the flow restricted to @xmath149 and such that @xmath138 has a natural ergodic invariant measure @xmath14 for the flow restricted to @xmath149 . it is possible to show ( under certain additional hypotheses ) that @xmath138 as an attractor in the full system if and only if @xmath150 , where @xmath151 is the most positive liapunov exponent for @xmath14 in a direction transverse to @xmath149 ( see @xcite for a detailed discussion ) . if we have access to a normal parameter @xcite such that we can vary the normal dynamics without changing the dynamics in @xmath149 , we can vary @xmath151 through zero and observe loss of attractiveness of @xmath138 at what ott and sommerer have termed a _ blowout bifurcation _ . ott and sommerer identify two possible scenarios at blowout . at a _ supercritical blowout _ the attractor bifurcates to a family of attractors displaying on - off intermittency , with trajectories that come arbitrarily close to @xmath138 and thus linger near @xmath138 for long times ( but with a well - defined mean length of lingering or ` laminar phase ' ) . at _ subcritical blowout _ there are no nearby attractors after loss of stability of @xmath138 . note that @xcite discusses the question of how to distinguish these cases . however , what we see in this paper is that there is at least one other possibility at blowout bifurcation that is also robust to normal perturbations , namely a bifurcation to a cycling state or a robust heteroclinic cycle between chaotic invariant sets . this can be seen to be set supercritical but not supercritical , in the following sense . by reparametrising if necessary we can assume that we have a normal parameter @xmath152 so @xmath138 undergoes a blowout bifurcation at @xmath153 . by the argument above , @xmath138 is an attractor only if @xmath154 . if there is a family of minimal milnor attractors @xmath155 , @xmath156 such that for all @xmath157 @xmath158 then we say that the blowout is _ set supercritical_. note that as discussed in and implied by the very word ` blowout ' , we can not typically expect the limit of the attractors to just be the set @xmath138 . if in addition @xmath155 supports a family of natural ergodic invariant measures @xmath159 ( @xmath156 ) and @xmath160 as @xmath161 then we say the blowout is _ measure supercritical _ or just _ supercritical_. ( convergence is the in the @xmath162 topology on probability measures . ) this definition of criticality was used in @xcite . if the blowout is not set supercritical then we say the blowout is _ subcritical_. it seems that one will often get bifurcations in symmetric systems that are set supercritical but not supercritical . for example one can generically get a blowout from a group orbit of attractors ( under a finite group ) which yields a single attractor that limits onto the whole group orbit . moreover , in the cycling chaos studied here , the attractor after blowout includes not only the chaotic invariant sets , but also fixed points involved in the heteroclinic cycle . as it is a cycle , it does not possess a natural ergodic invariant measure @xcite and averages of observables from the system in this state will typically not converge but continue to oscillate more and more slowly . the blowout scenario described above holds only for variation of normal parameters ; in general the variation of a parameter will affect both the normal dynamics and the dynamics within the invariant subspace . if the dynamics in the invariant subspace is chaotic , we can expect to see a large number of bifurcations happening within the invariant subspace and these will cause the blowout to be spread over an interval of parameter values ; there is no reason why @xmath163 should vary continuously with a parameter that varies @xmath164 in a very discontinuous manner . nevertheless , for the numerical results presented in section [ secnumerics ] the dynamics in the invariant subspace vary in quite regular way . this is because in our system the parameter @xmath16 is a normal parameter for the attractor @xmath65 ; for more discussion of normal parameters , see . by mechanisms described in @xcite the pure @xmath6 and @xmath5 chaotic dynamics corresponding to dynamics within @xmath39 and @xmath40 respectively can become chaotic by means of a symmetric global bifurcation that generates lorenz - like attractors approaching the equilibrium solution with full symmetry and @xmath6-roll or @xmath5-roll equilibrium solutions in @xmath42 or @xmath43 . now suppose there exist chaotic attractors @xmath65 and @xmath66 contained in @xmath39 and @xmath40 ( on average they have more symmetry ) , from here on we will mostly discuss @xmath65 but note that as @xmath66 is a conjugate attractor , the same will hold for @xmath66 . these attractors contain a saddle equilibrium @xmath165 in @xmath46 ( the trivial solution ) and so in particular they can not be liapunov stable attractors . this is because @xmath166 must be non - trivial manifold ( otherwise @xmath165 is an attractor ; thus @xmath167 is also non - trivial as @xmath168 is conjugate to @xmath169 . therefore @xmath170 and so @xmath65 can not be liapunov stable . however it is possible that @xmath65 is an attractor in the sense of milnor ; this will imply that the basin will be locally riddled in the sense of @xcite . we assume that @xmath65 and @xmath66 are minimal milnor attractors containing @xmath165 . we also assume that they have natural ergodic invariant measures @xmath164 and @xmath171 supported on them . we now concentrate on @xmath65 . whether @xmath65 is an attractor depends on its the spectrum of normal liapunov exponents . note that the zero liapunov exponent corresponding to time translation always corresponds to a perturbation tangential to the invariant subspace . if @xmath172 where @xmath173 is the most positive normal liapunov exponent for the measure @xmath14 , it is possible to show that @xmath65 satisfies ( [ eqmattr ] ) @xcite , and hence that @xmath65 is a milnor attractor . it is comparatively easy to compute @xmath163 in that case , as the largest transverse liapunov exponent of @xmath65 in the parameter regime discussed corresponds to directions in the @xmath37 direction , where linearised perturbations are described by @xmath174 thus we can see that @xmath175 where @xmath176 . from the equation for @xmath177 it is possible to show that @xmath178 along bounded trajectories of the ode . for the given parameter values it is possible to approximate @xmath179 and so @xmath180 . thus @xmath181 implying that the blowout bifurcation occurs at approximately @xmath69 , which is in good agreement with the simulations ( figure [ figblowout]b ) . note that whenever @xmath165 is hyperbolic and within the attractor @xmath65 there exists at least one ergodic invariant measure @xmath182 ( in particular that one supported on @xmath165 ) such that @xmath183 . in particular , this means that the basins of the attractors @xmath65 and @xmath66 are riddled for all parameter values in our problem ! likewise , in the map ( [ eqnearpx ] ) near @xmath39 , the logarithm of @xmath4 obeys @xmath184 where @xmath101 is a function of @xmath16 . the most positive normal liapunov exponent in this case is then @xmath185 where the average is taken over the lorenz attractor . averaging over @xmath186 iterates of ( [ eqlorenz ] ) yields @xmath187 and , solving ( [ eqlambdae ] ) for @xmath16 , we obtain @xmath188 at the blowout bifurcation , in agreement with the data in figure [ figmapone](b ) . as @xmath163 passes through @xmath189 , @xmath65 loses stability and becomes a chaotic saddle , and in doing so it creates a continuum of connections from @xmath65 to the fixed point ( @xmath5-rolls ) in @xmath43 . these connections are robust to @xmath34-equivariant perturbations , as @xmath5-rolls are sinks within @xmath37 , and so there is a robust cycle alternating between the equilibrium points in @xmath42 and @xmath43 and the chaotic sets in @xmath39 and @xmath40 . we observed in sections [ secnumerics ] and [ secmap ] that this cycling chaos is attractive once it is created ; we turn to the stability of cycling chaos in the next section . it is clear from figures [ figslowingdown ] , [ figresonance ] and [ figmaptwo ] that the key to understanding the loss of stability of cycling chaos lies in obtaining the rate at which trajectories approach that state . it is possible to estimate this rate for the map , and we use information gained in this calculation to carry out the same estimate in the odes , and thus are able to obtain the values of @xmath16 at which the loss of stability occurs , in the map and in the odes . what is remarkable is that , once a single average over @xmath65 has been computed numerically , the value of @xmath16 at the bifurcation point can be obtained analytically . we suppose that ( in the map ) the system arrives near @xmath39 with given values of @xmath190 , iterates using ( [ eqnearpx ] ) @xmath191 times ( @xmath192 ) , ending up in a state @xmath193 with @xmath194 . there follows one iterate of ( [ eqswitchpxpy ] ) , leaving the system near @xmath40 in a state @xmath195 ( we are ignoring changes of sign of the variables ) . we need to establish an estimate of @xmath196 , a measure of the distance from @xmath40 , given that @xmath197 were small when the system started close to @xmath39 . properly , the value of @xmath198 after @xmath191 iterates will depend on the values of @xmath3 over those @xmath191 iterates , but if @xmath191 is large , we approximate the detailed history of @xmath3 by its average and obtain @xmath199 where @xmath200 are liapunov exponents in the expanding and contracting directions around @xmath39 . @xmath201 is precisely the liapunov exponent that went through zero at the blowout bifurcation in ( [ eqlambdae ] ) . note that we have effectively approximated the chaotic set @xmath65 by an equilibrium point . the trajectory escapes from the neighbourhood of @xmath39 once @xmath194 ; since @xmath3 is typically of order one ( compared to the tiny initial value of @xmath4 ) , we assume that the escape takes place when @xmath202 , so obtaining @xmath203 with @xmath204 and @xmath198 both of order one . one iterate of ( [ eqswitchpxpy ] ) now yields @xmath205 these expressions will be dominated by @xmath206 once the trajectory is very close to the heteroclinic cycle , so , neglecting terms of order one , we obtain @xmath207 \left(\matrix{\log\|y_0\|\cr \log\|y_2\|\cr}\right ) + { \cal o}(1)\ ] ] for large negative values of @xmath208 and @xmath209 . a conjugate map will describe the return from @xmath40 to @xmath39 . one eigenvalue of the matrix is zero because of the way we approximated the behaviour near the chaotic set ; the other eigenvalue is @xmath210 which we refer to as the _ stability index_. the zero eigenvalue forces @xmath211 , so after one iterate of the composite map ( [ eqlinearisedmap ] ) , the dynamics will obey @xmath212 clearly if @xmath213 , @xmath214 and @xmath208 will tend to @xmath215 and the trajectory will asymptote to attracting cycling chaos . conversely , if @xmath216 , cycling chaos is unstable and trajectories will leave the domain of validity of the approximations we have made . we can also deduce from ( [ eqnumberits ] ) that the number of iterates between each switch will increase by a factor of @xmath30 per switch . at the point at which cycling chaos is created ( as @xmath201 increases through zero ) , we see that @xmath30 is much greater than one , provided that @xmath217 is negative and @xmath112 is positive , both of which are true in the examples we have discussed . we deduce that cycling chaos is attracting near to its creation at the blowout bifurcation . from the condition that @xmath218 at the loss of stability of the chaotic cycle , we determine that this bifurcation occurs in the map at @xmath219 , in agreement with the data in figure [ figmaptwo ] . returning to the odes , we observe that in ( [ eqrhomap ] ) @xmath112 and @xmath113 are ratios of eigenvalues of @xmath6-rolls ( proportional to decay rates of @xmath4 and @xmath10 ) , while @xmath201 and @xmath217 are the growth rate of @xmath4 and the decay rate of @xmath10 near the chaotic set @xmath65 . in the odes , the linearisation of ( [ eqode ] ) about @xmath65 yields @xmath119 for the decay rate of @xmath10 , while the growth rate of @xmath4 is given by ( [ eqlambdae ] ) . hence we have the stability index @xmath220 for the odes , where @xmath201 is given by ( [ eqlambdae ] ) . note that @xmath112 and @xmath201 are both functions of @xmath16 . the condition @xmath218 is readily solved for @xmath16 , and has solution @xmath221 . at @xmath221 the odes are still approaching cycling chaos , with switching times increasing by a factor of above 1.02 per switch ( see figure [ figresonance ] ) . however , the odes have not yet reached their asymptotic rate of slowing down , resolving this small discrepancy . on decreasing @xmath16 below @xmath71 , the stability index @xmath30 increases above unity and the cycling chaos will no longer be attracting . this loss of stability can broadly be classed as a resonance of liapunov exponents . observe that the resonance will be located at different @xmath16 for different invariant measures supported on @xmath222 and so we expect the presence of riddling and associated phenomena found in @xcite at a resonance of a simpler model displaying cycling chaos . we observe that for @xmath16 below the resonance bifurcation , the system exhibits behaviour that is numerically indistinguishable from periodic : there appear to be a large number of coexisting apparently periodic orbits . we hypothesize that the resonance creates a branch of ` approximately periodic attractors ' , i.e. , attractors that have a well - defined finite mean period of passage around the cycle , going to infinity at the resonance @xcite . these might lock onto long periodic orbits for progressively smaller @xmath16 , as found in the numerical simulations . for this example , the approximately periodic attractor branches set supercritically from the cycling chaos ; however one presumes that subcritical branching is also possible . research is presently in progress on understanding the more detailed branching behaviour at this bifurcation . this study has shown that one possible , apparently generic scenario for loss of stability of a chaotic attractor in an invariant subspace on varying a normal parameter is as follows : there is a blowout bifurcation that creates an attracting , robust heteroclinic cycle between chaotic invariant sets ( cycling chaos ) . the bifurcation is set supercritical but not supercritical , i.e. , the bifurcated attractors contain the attractor for the system in the invariant subspace , but unlike in a supercritical bifurcation ( to an on - off intermittent state ) the length of laminar phases increases unboundedly along a single trajectory even at a finite distance from the blowout . this cycling chaotic state can be modelled well by the network shown in figure [ figcycle ] although in reality the network is complicated by the facts that ( a ) there are other fixed points contained in the closure of unstable manifolds and ( b ) the fixed points in @xmath223 and @xmath224 are actually contained in the chaotic sets @xmath65 and @xmath66 rather than being isolated . we suspect this may have the consequence that there is no poincar section to the flow and so the cycle is ` dirty ' in the terminology of @xcite . nevertheless , the normal liapunov spectrum of the invariant chaotic set seems to determine the attraction or not of the cycle . the attracting cycling chaos is observed to lose stability via a mechanism that resembles a resonance of eigenvalues in an orbit heteroclinic to equilibria . such a resonance has been seen to occur in special classes of systems with skew product structure @xcite , in analogy to the branching of periodic orbits at a homoclinic resonance investigated by . throughout this investigation , it has been necessary to monitor carefully several numerical effects . in particular , for trajectories that display asymptotic slowing down characteristic of cycling chaos there will be a point at which rounding errors cause the dynamics either to transfer to the invariant subspace , or keep the dynamics a finite distance from the invariant subspace . in the context of physical systems there will always be imperfections in the system and noise that will destroy the invariant subspaces . nevertheless the perfect symmetry model will be very useful in describing what one expects to see in such imperfect situations . it still remains to prove rigorously that the observed scenario is generic and so of interest to other , less specific models and in particular pde models of which this is a truncation . we could like to emphasise that the behaviour we see occurs for a reasonably large region of physically relevant parameters in the ode model and moreover we are unaware of any other classification which explains and predicts the observed dynamics to the degree that we have done here . in principal cycling chaos can be seen in ode models down to dimension 4 , though not smaller than this ; thus this dynamics should be seen as something that will not be created at a generic bifurcation from a trivial state , but rather in a more complicated dynamical regime far from primary bifurcation . we have discussed a possible route to cycling chaos through a blowout bifurcation , and how cycling chaos might cease to be attracting , both in general terms and in the context of a specific model . our general results ought to be applicable to a variety of other problems . in particular , behaviour that might be understood in terms of cycling chaos has been seen by knobloch _ et al . _ @xcite in an ode model of the dipole quadrupole interaction in the solar dynamo . in their model , there is a weakly broken symmetry between the dipole and quadrupole subspaces , and the system switches between the two subspaces , favouring one over the other since they are not equivalent . finally , one comment that deserves to be made is that the choice of @xmath16 as the parameter allows an important simplification because of this parameter is normal for dynamics within @xmath39 and @xmath40 . one assumes that similar behaviour will be observed for non - normal parameters with the exception that the chaos in the invariant subspace will be fragile and destroyed by many arbitrarily small perturbations ; see . we acknowledge very interesting conversations with mike field , marty golubitsky and edgar knobloch concerning this work . the research of pa was partially supported by a nuffield ` newly appointed science lecturer ' award and epsrc grant gr / k77365 . amr is grateful for support from the royal astronomical society .	we examine a model system where attractors may consist of a heteroclinic cycle between chaotic sets ; this ` cycling chaos ' manifests itself as trajectories that spend increasingly long periods lingering near chaotic invariant sets interspersed with short transitions between neighbourhoods of these sets . this behaviour can be robust ( i.e. , structurally stable ) for systems with symmetries and provides robust examples of non - ergodic attractors in such systems ; we examine bifurcations of this state . we discuss a scenario where an attracting cycling chaotic state is created at a blowout bifurcation of a chaotic attractor in an invariant subspace . this is a novel scenario for the blowout bifurcation and causes us to introduce the idea of _ set supercriticality _ to recognise such bifurcations . the robust cycling chaotic state can be followed to a point where it loses stability at a resonance bifurcation and creates a series of large period attractors . the model we consider is a @xmath0th order truncated ordinary differential equation ( ode ) model of three - dimensional incompressible convection in a plane layer of conducting fluid subjected to a vertical magnetic field and a vertical temperature gradient . symmetries of the model lead to the existence of invariant subspaces for the dynamics ; in particular there are invariant subspaces that correspond to regimes of two - dimensional flows . stable two - dimensional chaotic flow can go unstable to three - dimensional flow via the cross - roll instability . we show how the bifurcations mentioned above can be located by examination of various transverse liapunov exponents . we also consider a reduction of the ode to a map and demonstrate that the same behaviour can be found in the corresponding map . this allows us to describe and predict a number of observed transitions in these models . 05.45+b 47.20.ky 47.65heteroclinic cycle , symmetry , chaotic dynamics , magnetoconvection .
in recent work , we have reported near - exact @xcite and exact @xcite solutions of the singlet ground state of two electrons , interacting _ via _ a coulomb potential , but trapped on the surface of a sphere . this model was first used by berry and co - workers in the 1980 s to provide insight into angular correlation in two - electron systems @xcite . it has proven useful for understanding the electronic polarity of nanoclusters and for explaining the giant polarizability of na@xmath0f@xmath1 and spontaneous dipole formation on niobium clusters @xcite . within the adiabatic connection in density functional theory ( dft ) @xcite , seidl carefully studied this system @xcite in order to test the isi ( interaction - strength interpolation ) model @xcite , deriving values of the energy by numerical integration . furthermore , it has been shown that this kind of spherical constraint applied to the moshinsky atom @xcite leads to a solvable schrdinger equation @xcite . berry and collaborators also considered an extension in which each particle is confined to a different , concentric sphere @xcite and used this model to simulate the rovibrational spectra of the water molecule in both the ground @xcite and excited states @xcite . more recently , the model has been applied to quantum - mechanical calculations of large - amplitude light - atom dynamics in polyatomic hydrides @xcite . it seems timely therefore to generalize our earlier studies @xcite to the case of two electrons located on the surface of two concentric spheres of different radii . to be consistent with our previous work @xcite , we will focus on the singlet ground state , which allows us to confine our attention to the symmetric spatial part of the wave function and ignore the spin coordinates . however , when the two radii are not equal , the spin coordinates are irrelevant , and one can easily generalize the present results to the triplet state by antisymmetrizing the spatial wave function . symmetric and asymmetric hartree - fock ( hf ) solutions are discussed in section [ sec : hf ] and the strengths and weaknesses of mller - plesset ( mp ) perturbation theory @xcite in section [ sec : mp ] . we consider asymptotic solutions for the large - spheres regime in section [ sec : larger ] and , in section [ sec : ci ] , we study the convergence behavior of the variational configuration interaction ( ci ) scheme . finally , by investigating the shape of the position intracule and the corresponding coulomb hole ( section [ sec : p ] ) , we report the existence of a secondary coulomb hole , shedding light on long - range correlation effects in two - electron systems . atomic units are used throughout . our model consists of two concentric spheres of radii @xmath2 , each bearing one electron . the position of the @xmath3-th electron is defined by the spherical angles @xmath4 , the interelectronic angle @xmath5 by @xmath6 and the interelectronic distance by @xmath7 where @xmath8 the hamiltonian of the system @xmath9 is simply @xmath10 where @xmath11 is the kinetic energy operator and @xmath12 is the coulomb operator . it is sometimes convenient to recast @xmath9 in term of the interelectronic angle @xmath5 . introducing the dimensionless parameter @xmath13 and using , one finds @xmath14 which shows the different scaling behavior of the kinetic and electrostatic terms . frontier between the shf and ahf solutions with respect to @xmath15 and @xmath16.,scaledwidth=48.0% ] @xmath17 ( dotted ) , @xmath18 ( dashed ) and @xmath19 ( solid ) as a function of @xmath16 for @xmath20.,scaledwidth=48.0% ] for @xmath21 , the restricted hartree - fock ( hf ) wave function and energy take @xcite the simple forms @xmath22 for @xmath23 , the electrons occupy different orbitals and an unrestricted hf treatment is required . however , the high symmetry of the system implies that there is a solution in which each orbital is constant over its sphere , and the resulting wave function and energy are @xmath24 we call this the `` symmetric hartree - fock '' ( shf ) solution , because the orbitals are spherically symmetric and we note that the shf energy depends only on the radius of the larger sphere . for certain values of @xmath15 and @xmath16 , a second , lower - energy hf solution arises @xcite , in which the two electrons tend to localize on opposite sides of the spheres . we call this the `` asymmetric hartree - fock '' ( ahf ) solution for the orbitals possess cylindrical , not spherical , symmetry . to obtain the ahf wave function @xmath25 we expand the orbitals as @xmath26 in the basis of zonal harmonics @xcite @xmath27 the fock matrix elements for the two orbitals are @xmath28 where @xmath29 is the kronecker symbol and @xmath30 are the two - electron integrals expressed in terms of the wigner @xmath31-symbols @xcite @xmath32 the summation in runs from @xmath33 to @xmath34 because of selection rules @xcite . the ahf energy is @xmath35\\ + \frac{1}{2 } \sum_{\ell_1,\ell_2=0}^\infty \left ( c_{\ell_1 } c_{\ell_2 } f_1^{\ell_1 \ell_2 } + d_{\ell_1 } d_{\ell_2 } f_2^{\ell_1 \ell_2 } \right).\end{gathered}\ ] ] for all of the radii considered , truncating the expansions in at @xmath36 yields @xmath37 with an accuracy of @xmath38 . the asymptotic limits of the hf energies satisfy @xmath39 and the limiting ahf energy corresponds to the coulomb interaction between two electrons that are fully localized on opposite side of their respective spheres . such systems are known as wigner molecules @xcite and have been observed in a variety of similar systems @xcite . the creation of localized orbitals leads to decreased coulombic repulsion but increased kinetic energy , and an asymmetric solution therefore exists only when the former outweighs the latter . by considering an orbital basis consisting of only @xmath40 and @xmath41 , it can be shown that this occurs only when @xmath42 and @xmath43 , where @xmath44 . figure [ fig : svsa ] illustrates this graphically . figure [ fig : ahf ] shows the shf , ahf and exact energies as functions of @xmath16 for several values of @xmath15 . the difference @xmath45 decreases as @xmath46 increases , indicating that the ahf energy is asymptotically correct . mp2 ( dashed ) , mp3 ( dotted ) and exact ( solid ) correlation energies ( relative to shf ) as a function of @xmath16 for @xmath47.,scaledwidth=48.0% ] in mller - plesset ( mp ) perturbation theory @xcite , the total hamiltonian is partitioned into a zeroth - order hamiltonian @xmath48 and a perturbative correction @xmath49 . the unperturbed orbitals are spherical harmonics on each sphere and therefore , from section [ subsec : shf ] , we have @xmath50 and @xmath51 . the @xmath52-th excited eigenfunction and eigenvalue of @xmath53 with @xmath54 symmetry are @xcite @xmath55 in intermediate normalization , the first - order correction to the wave function is @xmath56 where it can be shown that @xmath57 this yields the normalized first - order wave function @xmath58 ^ 2 \sum_{\ell=1}^\infty \frac{{\lambda}^{2\ell}}{\ell^2(\ell+1)^2(2\ell+1)}}}.\ ] ] using , one finds that the second - order energy @xmath59 ( where @xmath60 is the gauss hypergeometric function @xcite ) depends only on the ratio of the radii . when the radii are equal , @xmath61 takes the value @xmath62 which has been discussed by seidl @xcite and us @xcite . when the radii are very different ( _ i.e. _ @xmath63 ) , the hf treatment is accurate and the second - order energy is @xmath64 where @xmath65 . although can be identified as the dispersion energy , it does not exhibit the usual @xmath66 behavior . analogous results have also been reported for other systems @xcite . using , one finds the third - order energy @xmath67 \notag \\ & = r_2 \left[\frac{2{\lambda}^8}{9 } - \frac{12{\lambda}^{10}}{35 } + o({\lambda}^{12 } ) \right],\end{aligned}\ ] ] where @xmath68 is the dilogarithm function @xcite . when the radii are equal , @xmath69 takes the value @xmath70 that we have given previously @xcite . when the radii are very different and @xmath16 is not too large , @xmath69 is tiny and @xmath61 is a good approximation to the total correlation energy . the mp2 and mp3 correlation energies , defined by @xmath71 are shown in fig . [ fig : mpn ] . for @xmath72 , the mp2 and mp3 energies are accurate for all @xmath16 . for larger @xmath15 , the discrepancy between the mp and exact energies is noticeable for small @xmath16 , but remains small for large @xmath16 . the mp3 energy is usually better than the mp2 energy . however , as we have shown previously @xcite , the mp expansion appears to diverge when the radii are similar and not small @xcite . @xmath73 ( dashed ) and @xmath74 ( dotted ) correlation energies ( relative to shf ) as a function of @xmath16 for @xmath75 . the exact correlation energy ( solid ) is also shown.,scaledwidth=48.0% ] in the large - spheres ( ls ) regime , the electrons reduce their coulomb repulsion by localizing on opposite sides of their spheres , oscillating around their equilibrium positions with angular frequency @xmath76 ( zero - point oscillations ) . the same phenomenon has been observed by seidl and collaborators @xcite . in this case , the supplementary angle @xmath77 becomes the natural coordinate of the system . using the taylor expansions , @xmath78 and @xmath79 the hamiltonian becomes @xmath80.\end{gathered}\ ] ] the lowest eigenfunction of is @xmath81,\ ] ] and the associated eigenvalue is @xmath82 the first term of represents the classical interaction of two electrons separated by a distance @xmath83 , and the second one is the energy associated with the zero - point oscillations of angular frequency @xmath84 the first anharmonic correction @xmath85 arises from the next two terms of the taylor expansion of @xmath86 and the coulomb operator . defining @xmath87 , the anharmonic correction energy is @xmath88 the ls0 and ls1 correlation energies are shown in fig . [ fig : larger ] with respect to @xmath16 for three values of @xmath15 . for the large values of @xmath15 , both curves agree very well with the exact correlation energies , while for the smaller values of the radius of the first sphere , ls1 systematically improves the results compared to ls0 . [ cols="^,^,^,^,^",options="header " , ] in the weak interaction limit , the only hf solution is the symmetric one ( section [ subsec : shf ] ) and correlation effects are well - described by the mp approximation ( section [ sec : mp ] ) . figures [ fig : p - mp - r](a ) , [ fig : p - mp - r](c ) and [ fig : p - mp - r](e ) show the coulomb holes derived from the mp1 , mp2 and ci wave functions for three small values of @xmath89 . for such radii , the mp - based and exact position intracules are very similar , and become identical as @xmath90 . the holes are negative for small @xmath91 and positive for larger @xmath91 , implying that correlation decreases the likelihood of finding the two electrons close together and increases the probability of their being far apart @xcite . to illustrate the spatial distributions of the electrons , we have plotted the mp1 coulomb holes on the surface of a sphere ( fig . [ fig : densityplot - mp1 ] ) for the three same values of the @xmath89 . the evolution of @xmath92 with respect to the increase of @xmath16 is shown in figs . [ fig : p - mp - r](b ) , [ fig : p - mp - r](d ) and [ fig : p - mp - r](f ) . as @xmath16 increases , the difference between the mp1 and exact holes decreases and they match perfectly as @xmath93 . table [ tab : dp - mp ] shows that the first - order correction reduces the probability of small @xmath91 values too much , and that the second - order correction partly corrects this , at least for small of values of @xmath94 . as a consequence , the strength of the mp1 hole is always larger than the true one , but exhibits the right asymptotic behavior for small @xmath94 . holes based on the exact wave function for @xmath95 and various @xmath16 ( from 10 to 90 ) . the inset graph represents the transition between the ahf and shf solution ( @xmath96).,scaledwidth=48.0% ] in the strong interaction limit , the coulomb repulsion dominates the kinetic energy , an ahf solution exists ( section [ subsec : ahf ] ) and the electrons oscillate around their equilibrium positions ( section ) . for @xmath97 , a secondary coulomb hole appears in the exact @xmath98 ( fig . [ fig : dp - wrt - r ] ) , revealing that correlation _ decreases _ the probability of finding electrons at large separations . this implies that the ahf wave function over - localizes the electrons on opposite sides of their spheres , and that correlation then delocalizes them slightly . such secondary coulomb holes are not peculiar to our system ; they have also recently been observed in the he atom @xcite and the h@xmath99 molecule @xcite . for @xmath100 , the primary coulomb hole disappears completely , leaving only the secondary one ( table [ tab : dp - lr ] and fig . [ fig : dp - wrt - r ] ) . the secondary : primary strength ratio is larger than in the he atom and the equilibrium h@xmath99 molecule ( 1 - 2% ) and resembles that in the h@xmath99 molecule at a bond length of 3 a.u . @xcite . figure [ fig : dp - wrt - r2 ] shows the evolution of the exact hole for @xmath95 and @xmath16 ranging from 10 to 100 . the secondary hole vanishes when @xmath16 exceeds @xmath101 and the ahf solution collapses to the shf one . to compare the holes based on the ls wave function @xmath102 ( eq . ) and the exact one , we have plotted the difference between the exact and the ls holes ( @xmath103 in fig . [ fig : dp - lr ] ) . for @xmath89 , the agreement between the two holes is fairly good for large @xmath94 ( fig . [ fig : dp - lr](a ) ) . for the smaller values of the radius , it shows that the electronic zero - point oscillations tend to over - localize the electrons compared to the exact treatment . however , the secondary coulomb hole is less pronounced but still present in the ls approximation . moreover , one can see that the ls treatment slightly increases the likelihood of finding the two electrons close together . figure [ fig : dp - lr](b ) reports the modification of @xmath104 for a fixed value of the first sphere radius ( @xmath95 ) and various @xmath16 ( 10 , 30 and 50 ) . when @xmath16 is increasing , the first minimum disappears , and the main effect of the ls approximation is thus to over - localize the electrons on opposite side of the spheres . the 2d spatial distribution of the electrons is depicted in fig . [ fig : densityplot - ls ] , where we have represented the ls holes for various @xmath89 ( 5 , 10 and 50 ) on the surface of a sphere . ls holes for various @xmath89 plotted on the surface of a sphere . the holes with respect to @xmath5 are also represented.,scaledwidth=48.0% ] we have performed a comprehensive study of the singlet ground state of two electrons on the surface of spheres of radius @xmath15 and @xmath16 . symmetric and asymmetric hf solutions show that the symmetry - breaking process occurs only when @xmath105 and @xmath43 . mp2 and mp3 energy corrections reveal that mp theory is appropriate when both radii are small ( as previously known ) and , also , when @xmath106 . to derive asymptotic solutions of this problem , we have taken into account , in the harmonic and anharmonic approximations , the zero - point oscillations of the electrons around their equilibrium position . for any values of @xmath107 , the near - exact wave function and energy can easily be obtained by a ci expansion based on legendre polynomials because there is no cusp in the wave function . a study of the position intracules and coulomb holes reveals that , as in the helium atom and hydrogen molecule , there is a secondary coulomb hole in the large - sphere regime . indeed , as @xmath94 increases , the primary hole disappears and only the secondary one remains . this reflects an overlocalization of the electrons in the asymmetric hartree - fock solution . our results should be useful for the future development of accurate correlation functionals within density - functional theory @xcite and intracule functional theory @xcite and , also , for understanding secondary coulomb holes in more complex systems @xcite . thanks the nci national facility for a generous grant of supercomputer time and the australian research council ( grants dp0771978 and dp0984806 ) for funding . we also thank yves bernard for fruitful discussions and helpful comments on the manuscript .	we extend our analysis of two electrons on a sphere [ phys . rev . a * 79 * , 062517 ( 2009 ) ; phys . rev . lett . * 103 * , 123008 ( 2009 ) ] to electrons on concentric spheres with different radii . the strengths and weaknesses of several electronic structure models are analyzed , ranging from the mean - field approximation ( restricted and unrestricted hartree - fock solutions ) to configuration interaction expansion , leading to near - exact wave functions and energies . the mller - plesset energy corrections ( up to third - order ) and the asymptotic expansion for the large - spheres regime are also considered . we also study the position intracules derived from approximate and exact wave functions . we find evidence for the existence of a long - range coulomb hole in the large - spheres regime , and infer that unrestricted hartree - fock theory over - localizes the electrons .
let @xmath3 be an i.i.d . sequence of random variables with distribution @xmath4 , and denote by @xmath5 the distribution of a real - valued root @xmath6 under @xmath7 . in statistics and econometrics , it is often of interest to estimate certain quantiles of @xmath5 . two commonly used methods for this purpose are subsampling and the bootstrap . this paper provides conditions under which these estimators behave well uniformly over @xmath0 . more precisely , we provide conditions under which subsampling and the bootstrap may be used to construct estimators @xmath8 of the @xmath9 quantiles of @xmath5 and @xmath10 of the @xmath11 quantiles of @xmath5 , satisfying @xmath12 here , @xmath13 is understood to be @xmath14 , and @xmath15 is understood to be @xmath16 . for the construction of two - sided confidence intervals of nominal level @xmath17 for a real - valued parameter , we typically would consider @xmath18 , while for a one - sided confidence interval of nominal level @xmath19 we would consider either @xmath20 and @xmath21 , or @xmath22 and @xmath23 . in many cases , it is possible to replace the @xmath24 and @xmath25 in ( [ eqtwosided ] ) with @xmath26 and @xmath27 , respectively . these results differ from those usually stated in the literature in that they require the convergence to hold uniformly over @xmath0 instead of just pointwise over @xmath0 . the importance of this stronger notion of convergence when applying these results is discussed further below . as we will see , the result ( [ eqtwosided ] ) may hold with @xmath28 and @xmath29 , but it may fail if @xmath23 and @xmath30 , or the other way round . this phenomenon arises when it is not possible to estimate @xmath5 uniformly well with respect to a suitable metric , but , in a sense to be made precise by our results , it is possible to estimate it sufficiently well to ensure that ( [ eqtwosided ] ) still holds for certain choices of @xmath9 and @xmath31 . note that metrics compatible with the weak topology are not sufficient for our purposes . in particular , closeness of distributions with respect to such a metric does not ensure closeness of quantiles . see remark [ remarklevy ] for further discussion of this point . in fact , closeness of distributions with respect to even stronger metrics , such as the kolmogorov metric , does not ensure closeness of quantiles either . for this reason , our results rely heavily on lemma [ lemmaquant ] which relates closeness of distributions with respect to a suitable metric and coverage statements . in contrast , the usual arguments for the pointwise asymptotic validity of subsampling and the bootstrap rely on showing for each @xmath32 that @xmath33 tends in probability under @xmath7 to the @xmath34 quantile of the limiting distribution of @xmath35 under @xmath7 . because our results are uniform in @xmath36 , we must consider the behavior of @xmath35 and @xmath33 under arbitrary sequences @xmath37 , under which the quantile estimators need not even settle down . thus , the results are not trivial extensions of the usual pointwise asymptotic arguments . the construction of @xmath38 satisfying ( [ eqtwosided ] ) is useful for constructing confidence regions that behave well uniformly over @xmath0 . more precisely , our results provide conditions under which subsampling and the bootstrap can be used to construct confidence regions @xmath39 of level @xmath40 for a parameter @xmath41 that are uniformly consistent in level in the sense that @xmath42 our results are also useful for constructing tests @xmath43 of level @xmath44 for a null hypothesis @xmath45 against the alternative @xmath46 that are uniformly consistent in level in the sense that @xmath47 \leq\alpha.\ ] ] in some cases , it is possible to replace the @xmath48 and @xmath25 in ( [ equnifcr ] ) or the @xmath49 and @xmath50 in ( [ equniftest ] ) with @xmath51 and @xmath27 , respectively . confidence regions satisfying ( [ equnifcr ] ) are desirable because they ensure that for every @xmath52 there is an @xmath53 such that for @xmath54 we have that @xmath55 is no less than @xmath56 for all @xmath4 . in contrast , confidence regions that are only pointwise consistent in level in the sense that @xmath57 for each fixed @xmath4 have the feature that there exists some @xmath52 and @xmath58 such that @xmath59 is less than @xmath60 infinitely often . likewise , tests satisfying ( [ equniftest ] ) are desirable for analogous reasons . for this reason , inferences based on confidence regions or tests that fail to satisfy ( [ equnifcr ] ) or ( [ equniftest ] ) may be very misleading in finite samples . of course , as pointed out by @xcite , there may be no nontrivial confidence region or test satisfying ( [ equnifcr ] ) or ( [ equniftest ] ) when @xmath0 is sufficiently rich . for this reason , we will have to restrict @xmath0 appropriately in our examples . in the case of confidence regions for or tests about the mean , for instance , we will have to impose a very weak uniform integrability condition . see also @xcite , @xcite , leeb and ptscher ( @xcite ) , @xcite for related results in more complicated settings , including post - model selection , shrinkage - estimators and ill - posed problems . some of our results on subsampling are closely related to results in @xcite , which were developed independently and at about the same time as our results . see the discussion on page 431 of @xcite . our results show that the question of whether subsampling can be used to construct estimators @xmath38 satisfying ( [ eqtwosided ] ) reduces to a single , succinct requirement on the asymptotic relationship between the distribution of @xmath5 and @xmath61 , where @xmath62 is the subsample size , whereas the results of @xcite require the verification of a larger number of conditions . moreover , we also provide a converse , showing this requirement on the asymptotic relationship between the distribution of @xmath5 and @xmath61 is also necessary in the sense that , if the requirement fails , then for some nominal coverage level , the uniform coverage statements fail . thus our results are stated under essentially the weakest possible conditions , yet are verifiable in a large class of examples . on the other hand , the results of @xcite further provide a means of calculating the limiting value of @xmath63 in the case where it may not satisfy ( [ eqtwosided ] ) . to the best of our knowledge , our results on the bootstrap are the first to be stated at this level of generality . an important antecedent is @xcite , who studies the uniform asymptotic behavior of confidence regions for a univariate cumulative distribution function . see also @xcite , who analyzes the uniform asymptotic behavior of some tests that arise in the context of an autoregressive model . the remainder of the paper is organized as follows . in section [ sectiongeneral ] , we present the conditions under which @xmath64 satisfying ( [ eqtwosided ] ) may be constructed using subsampling or the bootstrap . we then provide in section [ sectionappl ] several applications of our general results . these applications include the multivariate mean , testing moment inequalities , multiple testing , the empirical process and @xmath2-statistics . the discussion of @xmath2-statistics is especially noteworthy because it highlights the fact that the assumptions required for the uniform asymptotic validity of subsampling and the bootstrap may differ . in particular , subsampling may be uniformly asymptotically valid under conditions where , as noted by @xcite , the bootstrap fails even to be pointwise asymptotically valid . the application to multiple testing is also noteworthy because , despite the enormous recent literature in this area , our results appear to be the first that provide uniformly asymptotically valid inference . proofs of the main results ( theorems [ theoremsubsample ] and [ theoremboot ] ) can be found in the ; proofs of all other results can be found in @xcite , which contains supplementary material . many of the intermediate results may be of independent interest , including uniform weak laws of large numbers for @xmath2-statistics and @xmath65-statistics [ lemmas s.17.3 and s.17.4 in @xcite , resp . ] as well as the aforementioned lemma [ lemmaquant ] . let @xmath3 be an i.i.d . sequence of random variables with distribution @xmath66 . denote by @xmath5 the distribution of a real - valued root @xmath6 under @xmath7 . the goal is to construct procedures which are valid uniformly in @xmath7 . in order to describe the subsampling approach to approximate @xmath5 , let @xmath67 be a sequence of positive integers tending to infinity , but satisfying @xmath68 , and define @xmath69 . for @xmath70 , denote by @xmath71 the @xmath72th subset of data of size @xmath62 . below , we present results for two subsampling - based estimators of @xmath5 . we first consider the estimator given by @xmath73 more generally , we will also consider feasible estimators @xmath74 in which @xmath75 is replaced by some estimator @xmath76 , that is , @xmath77 typically , @xmath78 , where @xmath79 is the empirical distribution , but this is not assumed below . even though the estimator of @xmath5 defined in ( [ equationsubdist ] ) is infeasible because of its dependence on @xmath7 , which is unknown , it is useful both as an intermediate step toward establishing some results for the feasible estimator of @xmath5 and , as explained in remarks [ remarkinvert ] and [ remarktesting ] , on its own in the construction of some feasible tests and confidence regions . [ theoremsubsample ] let @xmath67 be a sequence of positive integers tending to infinity , but satisfying @xmath68 , and define @xmath80 as in ( [ equationsubdist ] ) . then , the following statements are true : if @xmath81 , then @xmath82 holds for @xmath20 and any @xmath83 . if @xmath84 , then ( [ eqtwosidedsub ] ) holds for @xmath23 and any @xmath85 . if @xmath86 , then ( [ eqtwosidedsub ] ) holds for any @xmath87 and @xmath88 satisfying @xmath89 . [ remarkexactsub ] it is typically easy to deduce from the conclusions of theorem [ theoremsubsample ] stronger results in which the @xmath90 and @xmath25 in ( [ eqtwosidedsub ] ) are replaced by @xmath26 and @xmath27 , respectively . for example , in order to assert that ( [ eqtwosidedsub ] ) holds with @xmath90 and @xmath25 replaced by @xmath26 and @xmath27 , respectively , all that is required is that @xmath91 for some @xmath4 . this can be verified using the usual arguments for the pointwise asymptotic validity of subsampling . indeed , it suffices to show for some @xmath4 that @xmath5 tends in distribution to a limiting distribution @xmath92 that is continuous at the appropriate quantiles . see @xcite for details . [ remarkinvert ] as mentioned earlier , @xmath80 defined in ( [ equationsubdist ] ) is infeasible because it still depends on @xmath7 , which is unknown , through @xmath93 . even so , theorem [ theoremsubsample ] may be used without modification to construct feasible confidence regions for a parameter of interest @xmath41 provided that @xmath94 , and therefore @xmath80 , depends on @xmath7 only through @xmath41 . if this is the case , then one may simply invert tests of the null hypotheses @xmath95 for all @xmath96 to construct a confidence region for @xmath41 . more concretely , suppose @xmath97 and @xmath98 . whenever we may apply part ( i ) of theorem [ theoremsubsample ] , we have that @xmath99 satisfies ( [ equnifcr ] ) . similar conclusions follow from parts ( ii ) and ( iii ) of theorem [ theoremsubsample ] . [ remarktesting ] it is worth emphasizing that even though theorem [ theoremsubsample ] is stated for roots , it is , of course , applicable in the special case where @xmath100 . this is especially useful in the context of hypothesis testing . see example [ exmomineqsub ] for one such instance . next , we provide some results for feasible estimators of @xmath5 . the first result , corollary [ corrsubsample ] , handles the case of the most basic root , while theorem [ theoremsubsample2 ] applies to more general roots needed for many of our applications . [ corrsubsample ] suppose @xmath101 , where @xmath102 is a sequence of normalizing constants , @xmath41 is a real - valued parameter of interest and @xmath103 is an estimator of @xmath41 . let @xmath67 be a sequence of positive integers tending to infinity , but satisfying @xmath68 , and define @xmath104 then statements of theorem [ theoremsubsample ] hold when @xmath105 is replaced by @xmath106 . [ theoremsubsample2 ] let @xmath67 be a sequence of positive integers tending to infinity , but satisfying @xmath68 . define @xmath80 as in ( [ equationsubdist ] ) and @xmath74 as in ( [ equationfeassubdist ] ) . suppose for all @xmath52 that @xmath107 then , statements of theorem [ theoremsubsample ] hold when @xmath105 is replaced by @xmath108 . as a special case , theorem [ theoremsubsample2 ] can be applied to studentized roots . [ corrsubsample2 ] suppose @xmath109 where @xmath102 is a sequence of normalizing constants , @xmath41 is a real - valued parameter of interest , and @xmath103 is an estimator of @xmath41 , and @xmath110 is an estimator of some parameter @xmath111 . suppose further that : the family of distributions @xmath112 is tight , and any subsequential limiting distribution is continuous . for any @xmath52 , @xmath113 let @xmath67 be a sequence of positive integers tending to infinity , but satisfying @xmath68 and @xmath114 . define @xmath115 then statements of theorem [ theoremsubsample ] hold when @xmath105 is replaced by @xmath116 . [ remarkdec6 ] one can take @xmath117 in corollary [ corrsubsample2 ] . since @xmath118 effectively cancels out from both sides of the inequality in the event @xmath119 , such a root actually leads to a computationally feasible construction . however , corollary [ corrsubsample2 ] still applies and shows that we can obtain a positive result without the correction factor @xmath120 present in corollary [ corrsubsample ] , provided the conditions of corollary [ corrsubsample2 ] hold . for example , if for some @xmath118 , we have that @xmath121 is asymptotically standard normal under any sequence @xmath122 , then the conditions hold . in corollaries [ corrsubsample ] and [ corrsubsample2 ] , it is assumed that the rate of convergence @xmath123 is known . this assumption may be relaxed using techniques described in @xcite . we conclude this section with a result that establishes a converse for theorems [ theoremsubsample ] and [ theoremsubsample2 ] . [ theoremconverse ] let @xmath67 be a sequence of positive integers tending to infinity , but satisfying @xmath68 and define @xmath80 as in ( [ equationsubdist ] ) and @xmath74 as in ( [ equationfeassubdist ] ) . then the following statements are true : if @xmath124 , then ( [ eqtwosidedsub ] ) fails for @xmath20 and some @xmath125 . if @xmath126 , then ( [ eqtwosidedsub ] ) fails for @xmath23 and some @xmath127 . if @xmath128 , then ( [ eqtwosidedsub ] ) fails for some @xmath87 and @xmath88 satisfying @xmath89 . if , in addition , ( [ eqfeasclose ] ) holds for any @xmath52 , then statements above hold when @xmath129 is replaced by @xmath108 . as before , let @xmath3 be an i.i.d . sequence of random variables with distribution @xmath66 . denote by @xmath5 the distribution of a real - valued root @xmath130 under @xmath7 . the goal remains to construct procedures which are valid uniformly in @xmath7 . the bootstrap approach is to approximate @xmath131 by @xmath132 for some estimator @xmath79 of @xmath7 . typically , @xmath79 is the empirical distribution , but this is not assumed in theorem [ theoremboot ] below . because @xmath79 need not a priori even lie in @xmath133 , it is necessary to introduce a family @xmath134 in which @xmath79 lies ( at least with high probability ) . in order for the bootstrap to succeed , we will require that @xmath135 be small for some function ( perhaps a metric ) @xmath136 defined on @xmath137 . for any given problem in which the theorem is applied , @xmath138 , @xmath134 and @xmath139 must be specified . [ theoremboot ] let @xmath140 be a function on @xmath141 , and let @xmath79 be a ( random ) sequence of distributions . then , the following are true : suppose @xmath142 for any sequences @xmath143 and @xmath144 satisfying @xmath145 . if @xmath146 for any sequence @xmath147 , then @xmath148 holds for @xmath20 and any @xmath83 . suppose @xmath149 for any sequences @xmath143 and @xmath150 satisfying @xmath145 . if ( [ eqihatethishere ] ) holds for any sequence @xmath147 , then ( [ eqtwosidedboot ] ) holds for @xmath23 and any @xmath127 . suppose @xmath151 for any sequences @xmath152 and @xmath153 satisfying @xmath145 . if ( [ eqihatethishere ] ) holds for any sequence @xmath147 , then ( [ eqtwosidedboot ] ) holds for any @xmath87 and @xmath88 satisfying @xmath89 . [ remarkexactboot ] it is typically easy to deduce from the conclusions of theorem [ theoremboot ] stronger results in which the @xmath154 and @xmath25 in ( [ eqtwosidedboot ] ) are replaced by @xmath26 and @xmath27 , respectively . for example , in order to assert that ( [ eqtwosidedboot ] ) holds with @xmath90 and @xmath25 replaced by @xmath155 and @xmath27 , respectively , all that is required is that @xmath156 for some @xmath4 . this can be verified using the usual arguments for the pointwise asymptotic validity of the bootstrap . see @xcite for details . [ remarklevy ] in some cases , it is possible to construct estimators @xmath157 of @xmath5 that are uniformly consistent over a large class of distributions @xmath0 in the sense that for any @xmath52 @xmath158 where @xmath139 is the levy metric or some other metric compatible with the weak topology . yet a result such as ( [ eqlevyconv ] ) is not strong enough to yield uniform coverage statements such as those in theorems [ theoremsubsample ] and [ theoremboot ] . in other words , such conclusions do not follow from uniform approximations of the distribution of interest if the quality of the approximation is measured in terms of metrics metrizing weak convergence . to see this , consider the following simple example . let @xmath3 be an i.i.d . sequence of random variables with distribution @xmath159 . denote by @xmath160 the distribution of the root @xmath161 under @xmath162 , where @xmath163 . let @xmath79 be the empirical distribution of @xmath164 or , equivalently , @xmath165 . lemma s.1.1 in @xcite implies for any @xmath52 that @xmath166 whenever @xmath139 is a metric compatible with the weak topology . nevertheless , it follows from the argument on page 78 of @xcite that the coverage statements in theorem [ theoremboot ] fail to hold provided that both @xmath9 and @xmath31 do not equal zero . indeed , consider part ( i ) of theorem [ theoremboot ] . suppose @xmath20 and @xmath167 . for a given @xmath168 and @xmath169 , let @xmath170 . under @xmath171 , the event @xmath172 has probability @xmath173 . moreover , whenever such an event occurs , @xmath174 . therefore , @xmath175 . since the choice of @xmath176 was arbitrary , it follows that @xmath177 a similar argument establishes the result for parts ( ii ) and ( iii ) of theorem [ theoremboot ] . on the other hand , when @xmath139 is the kolmogorov metric , ( [ eqlevy ] ) holds when the supremum over @xmath178 is replaced with a supremum over @xmath179 for some @xmath169 . moreover , when @xmath180 is restricted to such an interval , the coverage statements in theorem [ theoremboot ] hold as well . before proceeding , it is useful to introduce some notation that will be used frequently throughout many of the examples below . for a distribution @xmath7 on @xmath181 , denote by @xmath182 the mean of @xmath7 , by @xmath183 the covariance matrix of @xmath7 , and by @xmath184 the correlation matrix of @xmath7 . for @xmath185 , denote by @xmath186 the @xmath187th component of @xmath182 and by @xmath188 the @xmath187th diagonal element of @xmath183 . in all of our examples , @xmath189 will be an i.i.d . sequence of random variables with distribution @xmath7 and @xmath79 will denote the empirical distribution of @xmath164 . as usual , we will denote by @xmath190 the usual sample mean , by @xmath191 the usual sample covariance matrix and by @xmath192 the usual sample correlation matrix . for @xmath193 , denote by @xmath194 the @xmath187th component of @xmath195 and by @xmath196 the @xmath187th diagonal element of @xmath197 . finally , we say that a family of distributions @xmath198 on the real line satisfies the standardized uniform integrability condition if @xmath199 = 0.\ ] ] in the preceding expression , @xmath200 denotes a random variable with distribution @xmath201 . the use of the term standardized to describe ( [ equi ] ) reflects that fact that the variable @xmath200 is centered around its mean and normalized by its standard deviation . [ examplesubmean ] let @xmath202 be an i.i.d . sequence of random variables with distribution @xmath4 on @xmath181 . suppose one wishes to construct a rectangular confidence region for @xmath182 . for this purpose , a natural choice of root is @xmath203 in this setup , we have the following theorem : [ theoremsubmean ] denote by @xmath204 the set of distributions formed from the @xmath187th marginal distributions of the distributions in @xmath0 . suppose @xmath0 is such that ( [ equi ] ) is satisfied with @xmath205 for all @xmath185 . let @xmath5 be the distribution of the root ( [ eqmaxroot ] ) . let @xmath67 be a sequence of positive integers tending to infinity , but satisfying @xmath206 and define @xmath80 by ( [ equationsubdist ] ) . then @xmath207\\[-8pt ] & & \qquad = 1 - \alpha_1 - \alpha_2\nonumber\end{aligned}\ ] ] for any @xmath87 and @xmath208 such that @xmath209 . furthermore , ( [ eqsubmean ] ) remains true if @xmath105 is replaced by @xmath108 , where @xmath74 is defined by ( [ equationfeassubdist ] ) with @xmath210 . under suitable restrictions , theorem [ theoremsubmean ] generalizes to the case where the root is given by @xmath211 where @xmath212 is a continuous , real - valued function and @xmath213 in particular , we have the following theorem : [ theoremsubgenmean ] let @xmath0 be defined as in theorem [ theoremsubmean ] . let @xmath5 be the distribution of root ( [ eqgeneralf ] ) , where @xmath212 is continuous . suppose further that for all @xmath214 that @xmath215 for any sequence @xmath147 such that @xmath216 under @xmath217 and @xmath218 , where @xmath219 . then @xmath220\\[-8pt ] & & \qquad\geq1 - \alpha_1 - \alpha_2\nonumber\end{aligned}\ ] ] for any @xmath87 and @xmath208 such that @xmath209 . suppose further that if @xmath219 for some @xmath221 satisfying @xmath222 for all @xmath185 , then @xmath223 is continuously distributed . then , ( [ eqgensubmeancover ] ) remains true if @xmath105 is replaced by @xmath108 , where @xmath74 is defined by ( [ equationfeassubdist ] ) with @xmath224 . moreover , the @xmath24 and @xmath225 may be replaced by @xmath26 and @xmath27 , respectively . in order to verify ( [ eqgensubmeanconv1 ] ) and ( [ eqgensubmeanconv2 ] ) in theorem [ theoremsubgenmean ] , it suffices to assume that @xmath223 is continuously distributed . under the assumptions of the theorem , however , @xmath223 need not be continuously distributed . in this case , ( [ eqgensubmeanconv1 ] ) and ( [ eqgensubmeanconv2 ] ) hold immediately for any @xmath226 at which @xmath227 is continuous , but require a further argument for @xmath226 at which @xmath228 is discontinuous . see , for example , the proof of theorem [ theorembootmomineq ] , which relies on theorem [ theorembootgenmean ] , where the same requirement appears . [ remarkonesided ] @xcite considers the following example . let @xmath189 be an i.i.d . sequence of random variables with distribution @xmath4 on @xmath229 . suppose it is known that @xmath230 for all @xmath4 and one wishes to construct a confidence interval for @xmath182 . a natural choice of root in this case is @xmath231 this root differs from the one considered in theorem [ theoremsubmean ] and the ones discussed in theorem [ theoremsubgenmean ] in the sense that under weak assumptions on @xmath0 , @xmath232 holds , but @xmath233 fails to hold . to see this , suppose ( [ equi ] ) holds with @xmath234 . note that @xmath235 where @xmath236 and @xmath237 . since @xmath238 for any @xmath4 , @xmath239 is bounded from above by @xmath240 it now follows from the uniform central limit theorem established by lemma 3.3.1 of @xcite and theorem 2.11 of @xcite that ( [ eqonesidedunif ] ) holds . it therefore follows from theorem [ theoremsubsample ] that ( [ eqtwosidedsub ] ) holds with @xmath20 and any @xmath125 . to see that ( [ eqothersideunif ] ) fails , suppose further that @xmath241 , where @xmath242 for some @xmath243 . for @xmath244 , @xmath245 the left - hand side of ( [ eqothersideunif ] ) is therefore greater than or equal to @xmath246 for any @xmath226 . in particular , if @xmath247 , then the second term is zero for large enough @xmath168 , and so the limiting value is @xmath248 . it therefore follows from theorem [ theoremconverse ] that ( [ eqtwosidedsub ] ) fails for @xmath249 and some @xmath250 . on the other hand , ( [ eqtwosidedsub ] ) holds with @xmath23 and any @xmath251 . to see this , consider any sequence @xmath252 and the event @xmath253 . for the root in this example , this event is scale invariant . so , in calculating the probability of this event , we may without loss of generality assume @xmath254 . since @xmath255 , we have for any @xmath256 that @xmath257 and similarly for @xmath258 . using the usual subsampling arguments , it is thus possible to show for @xmath259 that @xmath260 the desired conclusion therefore follows from slutsky s theorem . arguing as the the proof of corollary [ corrsubsample2 ] and remark [ remarkdec6 ] , it can be shown that the same results hold when @xmath105 is replaced by @xmath108 , where @xmath74 is defined as @xmath80 is defined but with @xmath182 replaced by @xmath195 . [ exmomineqsub ] the generality of theorem [ theoremsubsample ] illustrated in example [ remarkonesided ] is also useful when testing multisided hypotheses about the mean . to see this , let @xmath3 be an i.i.d . sequence of random variables with distribution @xmath4 on @xmath181 . define @xmath261 and @xmath262 . consider testing the null hypothesis that @xmath263 versus the alternative hypothesis that @xmath264 at level @xmath265 . such hypothesis testing problems have recently received considerable attention in the `` moment inequality '' literature in econometrics . see , for example , @xcite , @xcite , @xcite , @xcite , @xcite and romano and shaikh ( @xcite ) . theorem [ theoremsubsample ] may be used to construct tests that are uniformly consistent in level in the sense that ( [ equniftest ] ) holds under weak assumptions on @xmath0 . formally , we have the following theorem : [ theoremsubmomineq ] let @xmath0 be defined as in theorem [ theoremsubmean ] . let @xmath5 be the distribution of @xmath266 let @xmath67 be a sequence of positive integers tending to infinity , but satisfying @xmath68 and define @xmath267 by the right - hand side of ( [ equationsubdist ] ) with @xmath268 . then , the test defined by @xmath269 satisfies ( [ equniftest ] ) for any @xmath270 . the argument used to establish theorem [ theoremsubmomineq ] is essentially the same as the one presented in @xcite for @xmath271 though lemma s.6.1 in @xcite is needed for establishing ( [ eqonesidedunif ] ) here because of studentization . related results are obtained by @xcite . [ exmultsub ] we now illustrate the use of theorem [ theoremsubsample ] to construct tests of multiple hypotheses that behave well uniformly over a large class of distributions . let @xmath3 be an i.i.d . sequence of random variables with distribution @xmath4 on @xmath181 , and consider testing the family of null hypotheses @xmath272 versus the alternative hypotheses @xmath273 in a way that controls the familywise error rate at level @xmath274 in the sense that @xmath275 where @xmath276 for @xmath277 , define @xmath278 according to the right - hand side of ( [ equationsubdist ] ) with @xmath279 and consider the following stepwise multiple testing procedure : [ algstepsub ] step 1 : set @xmath280 . if @xmath281 then stop . otherwise , reject any @xmath282 with @xmath283 and continue to step @xmath284 with @xmath285 @xmath286 step @xmath287 : if @xmath288 then stop . otherwise , reject any @xmath282 with @xmath289 and continue to step @xmath290 with @xmath291 @xmath286 we have the following theorem : [ theoremsubmultiple ] let @xmath0 be defined as in theorem [ theoremsubmean ] . let @xmath292 be a sequence of positive integers tending to infinity , but satisfying @xmath68 . then , algorithm [ algstepsub ] satisfies @xmath293 for any @xmath270 . it is , of course , possible to extend the analysis in a straightforward way to two - sided testing . see also @xcite for related results about a multiple testing problem involving an infinite number of null hypotheses . [ exsubemp ] let @xmath3 be an i.i.d . sequence of random variables with distribution @xmath4 on @xmath229 . suppose one wishes to construct a confidence region for the cumulative distribution function associated with @xmath7 , that is , @xmath294\}$ ] . for this purpose a natural choice of root is @xmath295\bigr\ } - p\bigl\{(-\infty , t]\bigr\}\bigr\|.\ ] ] in this setting , we have the following theorem : [ theoremsubemp ] fix any @xmath296 , and let @xmath297\bigr\ } < 1 - \varepsilon\mbox { for some } t \in\mathbf r\bigr\}.\ ] ] let @xmath5 be the distribution of root ( [ eqks ] ) . then @xmath298\bigr\ } - p\bigl\{(-\infty , t ] \bigr\}\bigr\|\nonumber\\ & & \hspace{182pt}\qquad \leq l_n^{-1}(1-\alpha_2 , p ) \bigr\ } \\ & & \qquad= 1 - \alpha_1 - \alpha_2\nonumber\end{aligned}\ ] ] for any @xmath87 and @xmath208 such that @xmath209 . furthermore , ( [ eqsubemp ] ) remains true if @xmath105 is replaced by @xmath108 , where @xmath74 is defined by ( [ equationfeassubdist ] ) with @xmath210 . [ exsubustat ] let @xmath3 be an i.i.d . sequence of random variables with distribution @xmath4 on @xmath229 . suppose one wishes to construct a confidence region for @xmath299,\ ] ] where @xmath300 is a symmetric kernel of degree @xmath301 . the usual estimator of @xmath41 in this case is given by the @xmath2-statistic @xmath302 here , @xmath303 denotes summation over all @xmath304 subsets @xmath305 of @xmath306 . a natural choice of root is therefore given by @xmath307 in this setting , we have the following theorem : [ theoremsubustat ] let @xmath308 - \theta(p)\ ] ] and @xmath309.\ ] ] suppose @xmath0 satisfies the uniform integrability condition @xmath310 = 0\ ] ] and @xmath311}{\sigma^2 ( p ) } < \infty.\ ] ] let @xmath5 be the distribution of the root ( [ equstatroot ] ) . let @xmath67 be a sequence of positive integers tending to infinity , but satisfying @xmath68 , and define @xmath80 by ( [ equationsubdist ] ) . then @xmath312\\[-8pt ] & & \qquad = 1 - \alpha_1 - \alpha_2\nonumber\end{aligned}\ ] ] for any @xmath87 and @xmath208 such that @xmath209 . furthermore , ( [ eqsubustat ] ) remains true if @xmath105 is replaced by @xmath108 , where @xmath74 is defined by ( [ equationfeassubdist ] ) with @xmath210 . [ examplebootmean ] let @xmath202 be an i.i.d . sequence of random variables with distribution @xmath4 on @xmath181 . suppose one wishes to construct a rectangular confidence region for @xmath182 . as described in example [ examplesubmean ] , a natural choice of root in this case is given by ( [ eqmaxroot ] ) . in this setting , we have the following theorem , which is a bootstrap counterpart to theorem [ theoremsubmean ] : [ theorembootmean ] let @xmath0 be defined as in theorem [ theoremsubmean ] . let @xmath5 be the distribution of the root ( [ eqmaxroot ] ) . then @xmath313\\[-8pt ] & & \qquad= 1 - \alpha_1 - \alpha_2\nonumber\hspace{-35pt}\end{aligned}\ ] ] for any @xmath87 and @xmath208 such that @xmath209 . theorem [ theorembootmean ] generalizes in the same way that theorem [ theoremsubmean ] generalizes . in particular , we have the following result : [ theorembootgenmean ] let @xmath0 be defined as in theorem [ theoremsubmean ] . let @xmath5 be the distribution of the root ( [ eqgeneralf ] ) . suppose @xmath212 is continuous . suppose further that for all @xmath214 @xmath314 for any sequence @xmath147 such that @xmath216 under @xmath217 and @xmath218 , where @xmath219 . then @xmath315\\[-8pt ] & & \qquad \geq1 - \alpha_1 - \alpha_2\nonumber\end{aligned}\ ] ] for any @xmath87 and @xmath208 such that @xmath209 . let @xmath3 be an i.i.d . sequence of random variables with distribution @xmath4 on @xmath181 and define @xmath316 and @xmath317 as in example [ exmomineqsub ] . @xcite propose testing the null hypothesis that @xmath263 versus the alternative hypothesis that @xmath264 at level @xmath265 using an `` adjusted quasi - likelihood ratio '' statistic @xmath318 defined as follows : @xmath319 here , @xmath320 is understood to mean that the inequality holds component - wise , @xmath321 and @xmath322 where @xmath52 and @xmath323 is the @xmath324-dimensional identity matrix . @xcite propose a procedure for constructing critical values for @xmath318 that they term `` refined moment selection . '' for illustrative purposes , we instead consider in the following theorem a simpler construction . [ theorembootmomineq ] let @xmath0 be defined as in theorem [ theoremsubmean ] . let @xmath5 be the distribution of the root @xmath325 where @xmath326 is defined as in ( [ eqznp ] ) . then , the test defined by @xmath327 satisfies ( [ equniftest ] ) for any @xmath270 . theorem [ theorembootmomineq ] generalizes in a straightforward fashion to other choices of test statistics , including the one used in theorem [ theoremsubmomineq ] . on the other hand , even when the underlying choice of test statistic is the same , the first - order asymptotic properties of the tests in theorems [ theorembootmomineq ] and [ theoremsubmomineq ] will differ . for other ways of constructing critical values that are more similar to the construction given in @xcite , see @xcite . theorem [ theoremboot ] may be used in the same way that theorem [ theoremsubsample ] was used in example [ exmultsub ] to construct tests of multiple hypotheses that behave well uniformly over a large class of distributions . to see this , let @xmath3 be an i.i.d . sequence of random variables with distribution @xmath4 on @xmath328 , and again consider testing the family of null hypotheses ( [ eqfamily ] ) versus the alternative hypotheses ( [ eqaltfamily ] ) in a way that satisfies ( [ equniffwer ] ) for @xmath265 . for @xmath277 , let @xmath329 be the distribution of the root @xmath330 under @xmath7 , and consider the stepwise multiple testing procedure given by algorithm [ algstepsub ] with @xmath331 replaced by @xmath332 . we have the following theorem , which is a bootstrap counterpart to theorem [ theoremsubmultiple ] : [ theorembootmultiple ] let @xmath0 be defined as in theorem [ theoremsubmean ] . then algorithm [ algstepsub ] with @xmath331 replaced by @xmath332 satisfies ( [ eqfwercontrol ] ) for any @xmath270 . it is , of course , possible to extend the analysis in a straightforward way to two - sided testing . let @xmath3 be an i.i.d . sequence of random variables with distribution @xmath4 on @xmath229 . suppose one wishes to construct a confidence region for the cumulative distribution function associated with @xmath7 , that is , @xmath294\ } $ ] . as described in example [ exsubemp ] , a natural choice of root in this case is given by ( [ eqks ] ) . in this setting , we have the following theorem , which is a bootstrap counterpart to theorem [ theoremsubemp ] : [ theorembootemp ] fix any @xmath296 , and let @xmath0 be defined as in theorem [ theoremsubemp ] . let @xmath5 be the distribution of the root ( [ eqks ] ) . denote by @xmath79 the empirical distribution of @xmath164 . then @xmath333\bigr\ } - p\bigl\{(-\infty , t]\bigr\}\bigr\|\\ & & \qquad\hspace*{182pt } \leq j_n^{-1}(1-\alpha_2 , \hat p_n ) \bigr\}\\ & & \qquad = 1 - \alpha_1 - \alpha_2\end{aligned}\ ] ] for any @xmath87 and @xmath208 such that @xmath209 . some of the conclusions of theorem [ theorembootemp ] can be found in @xcite , though the method of proof given in @xcite is quite different . [ exbootustat ] let @xmath3 be an i.i.d . sequence of random variables with distribution @xmath4 on @xmath229 and let @xmath300 be a symmetric kernel of degree @xmath301 . suppose one wishes to construct a confidence region for @xmath334 given by ( [ eqthetaustat ] ) . as described in example [ exsubustat ] , a natural choice of root in this case is given by ( [ equstatroot ] ) . before proceeding , it is useful to introduce the following notation . for an arbitrary kernel @xmath335 , @xmath52 and @xmath336 , denote by @xmath337 the set of all distributions @xmath7 on @xmath229 such that @xmath338 \leq b.\ ] ] similarly , for an arbitrary kernel @xmath335 and @xmath169 , denote by @xmath339 the set of all distributions @xmath7 on @xmath229 such that @xmath340 where @xmath341 is defined as in ( [ eqsigmap ] ) finally , for an arbitrary kernel @xmath335 , @xmath52 and @xmath342 , let @xmath343 be the set of distributions @xmath7 on @xmath229 such that @xmath344 \leq b,\ ] ] whenever @xmath345 for all @xmath346 . using this notation , we have the following theorem : [ theorembootustat ] define the kernel @xmath347 of degree @xmath348 according to the rule @xmath349\\[-8pt ] & & { } - h(x_1,\ldots , x_m)h(x_{m+1},\ldots , x_{2m}).\nonumber\end{aligned}\ ] ] suppose @xmath350 for some @xmath169 and @xmath336 . let @xmath5 be the distribution of the root @xmath35 defined by ( [ equstatroot ] ) . then @xmath351 for any @xmath9 and @xmath31 such that @xmath352 . note that the kernel @xmath347 defined in ( [ eqhprime ] ) arises in the analysis of the estimated variance of the @xmath2-statistic . note further that the conditions on @xmath0 in theorem [ theorembootustat ] are stronger than the conditions on @xmath0 in theorem [ theoremsubustat ] . while it may be possible to weaken the restrictions on @xmath0 in theorem [ theorembootustat ] some , it is not possible to establish the conclusions of theorem [ theorembootustat ] under the conditions on @xmath0 in theorem [ theoremsubustat ] . indeed , as shown by @xcite , the bootstrap based on the root @xmath35 defined by ( [ equstatroot ] ) need not be even pointwise asymptotically valid under the conditions on @xmath0 in theorem [ theoremsubustat ] . to see ( i ) , first note that @xmath374 implies that @xmath375 for all @xmath376 . thus , @xmath377 , from which it follows that @xmath378 . similarly , to prove ( ii ) , first note that @xmath358 implies that @xmath379 for all @xmath380 , so @xmath381 . therefore , @xmath382 . to prove ( iii ) , note that because @xmath383 , it follows from ( i ) that @xmath384 . hence , @xmath385 . using the same reasoning , ( iv ) follows from ( ii ) and the assumption that @xmath358 . to see ( v ) , note that @xmath386 where the first inequality follows from the bonferroni inequality , and the second inequality follows from ( iii ) and ( iv ) . to prove ( vi ) , note that @xmath387 where the second inequality follows from ( i ) . a similar argument using ( ii ) establishes ( vii ) . finally , ( viii ) follows from ( vi ) and ( vii ) by an argument analogous to the one used to establish ( v ) . [ lemmavcreal ] let @xmath3 be an i.i.d . sequence of random variables with distribution @xmath7 . denote by @xmath5 the distribution of a real - valued root @xmath6 under @xmath7 . let @xmath388 , @xmath389 and define @xmath80 according to ( [ equationsubdist ] ) . then , for any @xmath52 , we have that @xmath390 let @xmath52 be given and define @xmath391 by @xmath392 denote by @xmath393 the symmetric group with @xmath168 elements . note that using this notation , we may rewrite @xmath394 as @xmath395 note further that @xmath396 which is a sum of @xmath397 identically distributed random variables . let @xmath52 be given . it follows that @xmath398 using markov s inequality , ( [ equationdumb ] ) can be bounded by @xmath399\nonumber\\[-8pt]\\[-8pt ] & & \qquad = \frac{1}{\varepsilon } \int_0 ^ 1 p \bigl\ { \sup_{x \in\mathbf{r } } \bigl\|s_n(x , p ; x_1,\ldots , x_n)\bigr\| > u \bigr\ } \,du.\nonumber\end{aligned}\ ] ] we may use the dvoretsky kiefer wolfowitz inequality to bound the right - hand side of ( [ equationintreal ] ) by @xmath400 < \frac{1}{\varepsilon}\sqrt{\frac{2 \pi}{k_n}},\ ] ] which establishes ( [ equationbound2 ] ) . [ lemmafinite ] let @xmath3 be an i.i.d . sequence of random variables with distribution @xmath4 . denote by @xmath5 the distribution of a real - valued root @xmath6 under @xmath7 . let @xmath389 and define @xmath80 according to ( [ equationsubdist ] ) . let @xmath401 then , for any @xmath52 and @xmath402 , we have that : let @xmath52 and @xmath402 be given . note that @xmath406 where the final inequality follows from lemma [ lemmavcreal ] . assertion ( i ) thus follows from the definition of @xmath407 and part ( vi ) of lemma [ lemmaquant ] . assertions ( ii ) and ( iii ) are established similarly . proof of theorem [ theoremsubsample ] to prove ( i ) , note that by part ( i ) of lemma [ lemmafinite ] , we have for any @xmath52 and @xmath402 that @xmath408 where @xmath409 by the assumption on @xmath410 , we have that @xmath411 for every @xmath412 . thus , there exists a sequence @xmath413 tending to @xmath414 so that @xmath415 . the desired claim now follows from applying part ( i ) of lemma [ lemmafinite ] to this sequence . assertions ( ii ) and ( iii ) follow in exactly the same way . we prove only ( i ) . similar arguments can be used to establish ( ii ) and ( iii ) . let @xmath20 , @xmath83 and @xmath416 be given . choose @xmath169 so that @xmath417 whenever @xmath418 for @xmath419 and @xmath32 . for @xmath168 sufficiently large , we have that @xmath420 for such @xmath168 , we therefore have that @xmath421 it follows from part ( vi ) of lemma [ lemmaquant ] that for such @xmath168 @xmath422 since the choice of @xmath423 was arbitrary , the desired result follows .	this paper provides conditions under which subsampling and the bootstrap can be used to construct estimators of the quantiles of the distribution of a root that behave well uniformly over a large class of distributions @xmath0 . these results are then applied ( i ) to construct confidence regions that behave well uniformly over @xmath1 in the sense that the coverage probability tends to at least the nominal level uniformly over @xmath0 and ( ii ) to construct tests that behave well uniformly over @xmath0 in the sense that the size tends to no greater than the nominal level uniformly over @xmath0 . without these stronger notions of convergence , the asymptotic approximations to the coverage probability or size may be poor , even in very large samples . specific applications include the multivariate mean , testing moment inequalities , multiple testing , the empirical process and @xmath2-statistics .
much recent progress in the general area of nonequilibrium steady states has come about by focusing on simple driven lattice - gas models which have the same symmetries as the more complex physical situations they represent . for instance , the asymmetric exclusion process a simple model of hard - core particles performing biased random walks is representative of a class of current - carrying states and growth processes involving a single scalar field @xcite . now , there are several examples of physical processes which involve two coupled scalar fields @xcite ; for such situations , the corresponding lattice model involves two sets of ising variables , with the time evolution of each species depending on the instantaneous configuration of the other . such systems can display interesting , new types of steady states , and in this paper we discuss their occurrence within a recently introduced lattice model of coupled fields . this model was initially proposed by lahiri and ramaswamy @xcite ( hence referred to as the lr model ) in connection with collective effects in sedimentation , the process by which heavier particles settle in a lighter fluid . interactions between sedimenting particles are mediated by the fluid @xcite , and give rise to a coupled dynamics of two fields , namely the concentration ( the particle number density ) and the tilt field ( the orientation of the principal axes of the particle distribution ) . a distinctive aspect of the inter - field coupling in this case is that within a hydrodynamic description , the coupling involves _ linear terms _ , in addition to more customary nonlinear coupling terms . this essential aspect is captured by the 1-d lr lattice model . linear couplings can also arise in other contexts , for instance in reduced models of magnetohydrodynamics @xcite , drifting flux lattices in superconductors @xcite , and the motion of stuck and flowing grains on a sandpile @xcite . the essential aspect of linear coupling is captured by the 1-d lr lattice model . in the sedimentation problem , analysis of the linearized hydrodynamic equations points to an instability in a certain regime , but leaves open the question of the ultimate steady state reached by the system . the lattice model allows this question to be answered ; the instability can be studied by varying the transition rates within the model , and the resulting steady state can be analysed and characterized @xcite . the instability in question is towards large - scale clustering , and the question arises to what extent the phenomenon resembles phase separation , familiar from equilibrium statistical mechanics . in fact , the nature of phase separation in nonequilibrium steady states is an area of great interest at present @xcite , and the study of the lr model proves quite illuminating in this context . the unstable phase turns out to be strongly phase separated , much as in the three - component model of evans et al @xcite . the phase separation is robust , and survives at all temperatures . the behaviour is quite different , however , at the threshold of the instability , which corresponds to the particularly simple situation in which one of the two fields evolves autonomously while the other is driven by the fluctuations of the first , as in the passive scalar problem . there is phase ordering in this case as well , but it is quite delicate . fluctuations of the order parameter are very strong , but do not destroy the ordered phase @xcite ; rather , they change its character in an essential way , leading to the breakdown of familiar characteristics such as the porod law for the decay of scaled correlations . finally , moving away from the threshold , even the disordered regime of the model has some surprises to offer . in this regime , the density is homogeneous on a macroscopic scale , but fluctuations are transported ballistically by two sets of kinematic waves @xcite . the waves are coupled nonlinearly , yet under special symmetry - determined conditions ( i.e. , on a special locus in parameter space ) the dissipation of each is governed by a distinct critical exponent @xcite . this is the first such instance to come to light in a system with nonvanishing nonlinear coupling . in the kinematic wave phase in general , the equations are a one - dimensional reduced model for the dynamics of small fluctuations of a flux lattice drifting through a type ii superconductor @xcite , and also serve as a generalised burgers equation for magnetohydrodynamics ( mhd ) @xcite . in the remainder of this section , we discuss the occurrence of the instability within a linearized hydrodynamic description , and then define the lr lattice model and briefly discuss its phase diagram . in the sections that follow , we discuss respectively strong phase separation ( in the unstable regime ) , fluctuation - dominated phase separation ( at the threshold ) , and kinematic waves and their decay ( in the stable regime ) . the hydrodynamic equations of a coupled - field system involve the density fluctuations @xmath0 of the two species in question . when dealing with conserved fields , the equations governing their time evolutions may be derived from the continuity equations @xmath1 and @xmath2 , where @xmath3 and @xmath4 are the currents of particles of species 1 and 2 respectively . we may proceed by writing each of the currents @xmath5 and @xmath6 in terms of a systematic part @xmath7 , a diffusive part @xmath8 , and noise @xmath9 . the coupling between the two species arises from the systematic parts , for each of @xmath10 and @xmath11 depend both on @xmath12 and @xmath13 . we obtain the requisite equations on expanding in powers of the density fluctuations . we find it convenient to write these equations in terms of the integrated density fields @xmath14 . then , to second order , we obtain & = & c_11 h_1x + c_12 h_2 x + d_1 ^2 h_1 x^2 + _ 1 ( h_1x)^2 + _ 1 ( h_2 x)^2 + _ 1 ( h_1x ) ( h_2 x ) + _ 1 ( x , t ) + [ 2 mm ] h_2 t & = & c_21 h_1x + c_22 h_2 x + d_2 ^2 h_2 x^2 + _ 2 ( h_1x)^2 + _ 2 ( h_2 x)^2 + _ 2 ( h_1x ) ( h_2 x ) + _ 2 ( x , t ) . + [ 2 mm ] [ coupled ] let us first examine the effect of keeping only the linear first derivative terms on the right hand side of the above equations . these terms , which are of primary interest to us here , were absent on symmetry grounds in coupled - field problems involving directed polymer motion and surface growth , considered respectively in references @xcite and @xcite . let the eigenvectors of the @xmath15 matrix @xmath16 be @xmath17 and @xmath18 , and let the corresponding eigenvalues be @xmath19 and @xmath20 . there are three cases to consider : 1 . _ stable : _ if @xmath19 and @xmath20 are real ( i.e. if @xmath21 ) , they represent the speeds of two waves . the two waves involve the eigenvectors @xmath17 and @xmath18 , each of which is composed of a linear combination of @xmath22 and @xmath23 . to proceed , rewrite eqs . [ coupled ] in terms of @xmath24 and @xmath25 . notice that in the rest frame of each wave , the other wave moves with a finite speed @xmath26 ; it is impossible to make both stationary by a galilean shift . the nonlinear terms describe the couplings between the two waves which govern their dissipation . threshold : _ the discriminant @xmath27 vanishes if one of the off - diagonal terms , say @xmath28 , is zero . at the linear level , this would imply that the field @xmath22 evolves autonomously , but ( since @xmath29 ) the time evolution of the ` passive scalar ' @xmath23 is influenced by @xmath22 . there is a tendency for particles of type 2 to be driven together . in the corresponding lattice model , this system shows phase separation accompanied , however , by large - scale fluctuations . unstable : _ there is an instability if @xmath19 and @xmath20 pick up an imaginary part ( i.e. if @xmath30 ) . the solutions of the linear equations then have an unbounded exponential growth of fluctuations . this instability signals the advent of a new state which is qualitatively different from the statistically homogeneous state assumed at the outset . by studying the corresponding lattice model in the unstable regime , we find that the system undergoes macroscopic phase separation of a robust sort . the lahiri ramaswamy model @xcite is the lattice counterpart of the continuum equations discussed above , and analysing its phases helps us understand the behaviour of the system in each of the cases ( a ) , ( b ) and ( c ) above . the lr model is defined in terms of two sets of variables @xmath31 and @xmath32 which reside on two interpenetrating sublattices ; the former occupy the integer sites and the latter the half - integer bond locations of a one - dimensional lattice with @xmath33 sites . each @xmath34 and @xmath35 is an ising variable taking on values @xmath36 . they represent discrete versions of the density and tilt fields in the sedimentation problem : if @xmath34 is @xmath37 , there is a particle ( @xmath38 ) at site @xmath39 , and if @xmath40 , there is no particle ( @xmath41 ) . the variable @xmath42 and @xmath43 , implies two values @xmath44 and @xmath45 of the local tilt respectively . a typical configuration of the full system is thus : @xmath46 . both sets of variables are conserved , i.e. @xmath47 and @xmath48 and the associated densities @xmath49 and @xmath50 are constant . the linear terms in the hydrodynamic equations lead us to consider a @xmath51-dependent local field which guides the @xmath52-current and _ vice versa_. thus , for instance , the kawasaki exchange dynamics of the adjacent spins @xmath34 and @xmath53 occurs at a rate which depends on @xmath54 . the moves and the corresponding rates are depicted below : the rates may be written succinctly as @xmath56 in a simple correspondence to a hydrodynamic model , discussed in section iv below , the coefficients in the continuum description ( eq . [ coupled ] ) depend on the rates @xmath57 , as also on the overall densities @xmath58 and @xmath59 of the two species . this correspondence allows us to demarcate each of the regimes ( a ) , ( b ) and ( c ) discussed above in terms of the parameters of the lattice model . throughout the rest of the paper , we will assume that @xmath60 . thus particles tend to move preferentially downhill , and holes uphill . holding the ratio @xmath61 fixed , it is then interesting to ask for the effect of varying @xmath62 . the result is summarized in the phase diagram of fig . 1 , for the half - filled case @xmath63 . we discuss the characteristics of each phase in brief . ( the light grey region in fig . 1 ) : in the corresponding continuum equation , the diagonal coefficients @xmath64 and @xmath65 vanish , while @xmath28 and @xmath66 both have the same sign . thus @xmath27 is negative , corresponding to the unstable case ( c ) . since @xmath67 , a local peak ( @xmath68 ) tends to become a valley ( @xmath69 ) if a particle resides on it , while a valley tends to convert into a hill if a hole sits on it . these moves act in concert with the motion of particles down slopes ( determined by @xmath60 ) , to promote segregation of both kinds of spins . in fact , using the condition of detailed balance , the steady state can be determined exactly along the the line @xmath70 shown in fig . the state shows phase segregation of a particularly strong sort , as discussed in section 2 . ( the thick central line in fig . 1 ) : this is the threshold case , with @xmath71 leading to vanishing @xmath27 . it describes a semi - autonomous problem in which the @xmath51 s evolve independently of the @xmath52 s ; they undergo a symmetric exclusion process . the @xmath52 s on the other hand obey the dynamics of eq . [ rates ] . the system shows a particularly delicate sort of phase separation , which is accompanied by large - scale fluctuations in the steady state @xcite . [ fig : phases ] * figure 1*:the phase diagram of the lr model at half - filling @xmath72 . the strongly phase separated phase ( sps ) is separated from the kinematic wave phase ( kw ) by the threshold , semi - autonomous line @xmath73 where there is fluctuation - dominated phase ordering ( fdpo ) . the steady state can be found exactly along the line @xmath74 ( using the condition of detailed balance ) and along @xmath75 ( using the condition of pairwise balance ) . 0.4 cm ( the dark grey region in fig . 1 ) : in this case @xmath28 and @xmath66 have opposite signs , so that @xmath27 is positive . the tendency of particles to slide downhill is countered by the tendency of particle - rich valleys to become hills . the resulting steady state has a uniform macroscopic density and slope , while density fluctuations are carried by two kinematic waves , each corresponding to a mode which involves both @xmath52 s and @xmath51 s . the steady state can be found exactly along the line @xmath75 shown in fig . 1 , by using the condition of pairwise balance @xcite . interestingly , the symmetries of the mode evolution equations can be tuned by varying the overall densities , and it is thereby possible to achieve a situation in which the dissipation of the two modes have different dynamical exponents in spite of being coupled at the nonlinear level @xcite . in the unstable phase of the lr model ( @xmath76 , the region marked sps in fig . 1 ) , there is a bootstrap effect at work : the tendency of @xmath52 particles to fall downhill into valleys works in tandem with the tendency of particle - rich regions of the @xmath51 landscape to become valleys themselves . this results in the formation of ever deeper valleys and large - scale clustering of particles in them , leading ultimately to a phase separated state . a similar sort of phase separation has been shown to occur in a three species abc model in which a cyclic rule determines preferential rates of exchanges of adjacent particles @xcite . for the case with an equal number of a , b and c particles , the steady state was found exactly , using the condition of detailed balance . in the lr model too , the exact steady state can be found @xcite through a detailed balance argument in the case @xmath77 , @xmath74 ( fig . 1 ) . this happens because although the problem is defined in terms of transition rates , it is possible to construct a hamiltonian with respect to which detailed balance is valid . this hamiltonian has the form = ^n_k=1 h_k \ { } _ k [ hamilt ] where @xmath78 is a height field ( the discrete analog of @xmath79 in eq . [ coupled ] ) defined by h_k \ { } = ^k_j=1 _ j-1/2 . [ height ] the steady state is thus described by an equilibrium boltzmann measure , and a configuration @xmath80 has weight @xmath81 where @xmath82 is the temperature . this can be seen by noting that the energy change on interchanging neighbouring @xmath52 spins is e_e ( _ i _ i+1 ) = _ i+12 ( _ i - _ i+1 ) , [ exchange ] while the change of energy under exchange of @xmath51 spins is given by a similar expression with @xmath52 s and @xmath51 s interchanged . thus the ratio of boltzmann weights of the configurations after and before an interchange of spins , say @xmath83 , is @xmath84 . this equals the ratio of forward to backward transition rates if the ratio @xmath85 is related to the rates @xmath86 and @xmath87 by = e^-2 /t [ balance ] the ratio @xmath88 of forward and backward transition rates involving the interchange of adjacent @xmath51 s is also related to the same weight factor . thus , provided @xmath74 holds , the steady state weight of a configuration is proportional to @xmath89 . notice that @xmath90 involves long - ranged couplings between spins @xmath34 and @xmath91 . the ground state exhibits complete phase separation , corresponding to all @xmath92 particles being at the bottom of the potential well formed by phase - separated @xmath51 spins . pictorially , the state is @xmath93 there are two interfaces across which there is a change of @xmath52 spin values , and two more involving the @xmath51 s . the spacing between the four interfaces is @xmath94 , and each is completely sharp in the limit @xmath95 . the effect of nonzero temperature is to smear out each of the four interfaces over a finite length . the interfacial profile can be calculated @xcite , but for our purpose here it suffices to note that the interface width is of the order of @xmath82 , and has basically no effect on the bulk of the spins . thus in an infinite sample , phase separation survives at all @xmath82 . this unusual behaviour can be traced to the fact that the energy grows super - extensively ( @xmath96 ( length)@xmath97 ) , and thus dominates over the entropy at all temperatures . this allows the strictures against phase separation in equilibrium one - dimensional systems @xcite to be avoided . when the filling is changed away from @xmath98 , the condition of detailed balance no longer holds , and there is no description in terms of an effective hamiltonian . nevertheless , kinetic arguments can be given to show that phase separation survives even when the filling is changed . the extreme robustness of the phenomenon leads us to call it strong phase separation . the approach to the steady state is exceptionally slow . it occurs through a coarsening process which involves thermal activation over large barriers , so we expect a coarsening length growing logarithmically in time . logarithmically slow coarsening was found in the abc model as well @xcite . in concluding this section , we mention a related model which has quite different behaviour . recall that in the lr model , the mobility of each spin species depends on the local density of the other species . what if the mobility depends not on density , but a higher derivative of the density ? a case in which the coupling is to the second derivative of the density can be analysed through detailed balance considerations , but this time , the hamiltonian has short - ranged interactions @xcite . there is thus no phase ordering in this model at nonzero @xmath82 , in contrast to the strong phase separation exhibited by the lr model in the unstable regime . in the marginal case @xmath99 ( the line fdpo in fig . 1 ) , the evolution of the @xmath51 s proceeds independently of the @xmath52 s . the fluctuating @xmath51 landscape provides a source of nonequilibrium noise on the @xmath52 s which tend to fall downwards along local slopes and cluster in the valleys . since the strong landscape fluctuations never cease , the clustering is never as complete as in the case of strong phase separation . rather , as we will see below , the character of long - range order is strongly modified by fluctuations , hence the appellation fluctuation - dominated phase ordering . starting from an initial random arrangement of the @xmath52 particles , the extent of clustering can be estimated quite simply @xcite : in any fluctuating surface characterized by a dynamical exponent @xmath100 , there are rearrangements of the profile over length scales @xmath101 in time @xmath102 . this sets the scale for the base lengths of new valleys which form in time @xmath102 . in the case at hand , where the scaling properties of the surface are those of the edwards - wilkinson model @xcite , we have @xmath103 . this valley base length should set the scale for spatial clustering of particles , so we expect the equal time correlation function to follow the scaling form _ o+r(t ) = c(r / l(t ) ) with l ~t^1/z . [ splt ] this is confirmed by numerical studies . this sort of scaling behaviour is a general characteristic of phase ordering dynamics , which describes the evolution from an initially disordered state to one which is phase separated @xcite . however , the scaling function here is quite different from that in usual cases . normally , with a conserved order parameter , @xmath104 drops linearly for small @xmath105 , i.e. @xmath106 as @xmath107 . by contrast , we find @xmath108 shows a cusp for small values of @xmath109 , i.e. c(y ) c_o ( 1-c_1 \|y\| ^ ) [ nonporod ] with @xmath110 @xcite . this is significant , since a linear dependence on @xmath109 ( which goes under the name of the porod law ) is a simple consequence of having phases separated by sharp interfaces @xcite . the breakdown of this law is the first pointer to the unusual character of the state that is developing . in steady state , @xmath108 again shows a similar scaling form , with the coarsening scale @xmath111 being replaced by the system size @xmath33 . the cusp for small argument remains unaltered . turning to one - point correlation functions , a suitable measure @xcite for our conserved spin system is the magnitude of the fourier components of the density profile @xmath112 where @xmath113 and @xmath114 runs over @xmath115 . a signature of an ordered state is that in the thermodynamic limit , the average values @xmath116 approach zero for all nonzero @xmath117 , but @xmath118 remains finite . we numerically monitored these averages @xmath119 over the ensemble of steady state configurations and found that @xmath120 in the thermodynamic limit for fixed @xmath121 , but @xmath122 approaches a nonzero value , where @xmath123 . @xmath122 provides a quantitative measure of phase separation . we find that @xmath124 indicating that the steady state is ordered , though not perfectly ( a perfectly ordered state has @xmath125 ) . as a function of time , @xmath126 fluctuates strongly , reflected in the fact that the distribution @xmath127 remains broad , with a root - mean - squared deviation that remains finite even as @xmath128 . one may ask whether phase ordering disappears when , in the course of its fluctuations , the value of @xmath126 falls close to zero . the answer is in the negative . we observe that a dip in @xmath126 is accompanied by a simultaneous rise in the value of either @xmath129 or @xmath130 . this implies that whenever the system loses a single large cluster ( making @xmath126 small ) either two or three such clusters appear in its place ( making the values of @xmath131 for some small @xmath114 go up ) . thus the system remains far from the disordered state , and always has a few large particle clusters which are of macroscopic size @xmath132 . a numerical study shows that the average size of the largest particle cluster @xmath133 . we observe further that the particle and hole cluster size distributions in the steady state decay as a power - law : prob(l ) ~l^- [ powerlaw ] with @xmath134 . this allows us to understand some features of the correlation functions discussed above . within the independent interval approximation @xcite which ignores correlations in the lengths of successive clusters , the two - point correlation function can be calculated and shown to have the scaling form of eq . [ splt ] @xcite . the cusp in the scaling function is reproduced as well , with the cusp exponent @xmath135 being related to @xmath136 by @xmath137 within this approximation . on a qualitative note , the power law cluster distribution makes the interfaces between phases broad and structured , so that the argument which leads to the porod law does not hold any longer . the phenomenon of fluctuation - dominated phase ordering , with the characteristics of broad distributions of order parameter fluctuations , power - law cluster distributions , and cusps in the scaled two - point correlation functions , can be shown analytically to hold for a related simple model of surface depth fluctuations . both steady state properties and time - dependent coarsening correlation functions can be calculated within this model @xcite , and the results bear out the picture of fdpo discussed above . in the stable phase of the lr model ( @xmath138 ) , particle motion and landscape fluctuations work oppositely , so that the resulting state is disordered . the density of both @xmath52 s and @xmath51 s is uniform , but the dynamics of density fluctuations is interesting , and this is the aspect we focus on . to begin with , we note that the steady state can be shown exactly to have a product measure form , provided that @xmath139 and @xmath140 ( the line marked pairwise balance in fig . 1 ) . to do this , choose new symbols to denote the values of @xmath52 and @xmath51 : use @xmath37 if a site or bond is occupied by a @xmath38 or @xmath44 and @xmath141 for @xmath41 or @xmath45 . then the elementary moves reduce to @xmath142 and @xmath143 ( with rate @xmath144 ) , and the reverse moves @xmath145 and @xmath146 ( with rate @xmath147 ) . now , in steady state the total probability flux into each configuration @xmath148 is equal to the total flux out of @xmath148 . a sufficient condition for this is the condition of pairwise balance @xcite : in - fluxes and out - fluxes must balance in pairs , i.e. for every out - flux @xmath149 , there is a unique in - flux @xmath150 such that @xmath151 where @xmath152 s are transition rates , and @xmath153 s are steady state probabilities . using intuition gained from particle hopping models , this condition can be shown to be valid @xcite if @xmath154 . thus every allowed configuration is equally likely in steady state , which implies that @xmath155 is the product of independent site occupation probabilities in the thermodynamic limit . an immediate consequence is that multi - point correlation functions in the steady state decouple into products of one - point functions , allowing us to find the exact expression for the current of each species : @xmath156 for the @xmath52-particle current , and a similar expression with 1 and 2 interchanged for the @xmath51-particle current . hydrodynamic equations for density fluctuations can be derived using these expressions for the current , on following the procedure of section 1.1 . further , by taking linear combinations of the fields , one can construct eigenmode fields @xmath157 which decouple at the linear level . these fields describe wave - like modes travelling with speeds @xmath158 . the time evolution of the field @xmath159 is governed by the equation @xmath160 + \eta_+(x , t ) \nonumber \\ \label{h+-}\end{aligned}\ ] ] and a similar equation holds for @xmath161 . the quadratic nonlinearities in this equation are of the same form as in eq . [ coupled ] . the coefficients of these nonlinear terms , as also the coefficient @xmath162 of the cubic nonlinearity in eq . [ h+- ] , depend on the densities @xmath58 and @xmath59 . the new noise terms @xmath163 are also delta - correlated . while the fields @xmath164 and @xmath161 are decoupled at the linear level , they are coupled through the nonlinear terms , so that each mode can influence the dissipation of the other . some of the coefficients in eq . [ h+- ] vanish for certain choices of densities @xmath165 and @xmath166 . then , special symmetries arise in the evolution equations and the dynamical exponents associated with the wave modes may change . four cases arise . \(a ) @xmath167 symmetry : invariance under up - down reflection ( @xmath168 ) @xmath169 and under inversion ( @xmath170 ) of space @xmath171 . \(b ) @xmath172 symmetry : invariance under @xmath169 and _ not _ under @xmath171 . \(c ) @xmath173 symmetry : invariance under @xmath171 and _ not _ under @xmath169 . \(d ) @xmath174 symmetry : invariance neither under @xmath171 , nor under @xmath169 . a term like @xmath175 obeys @xmath167 symmetry . terms like @xmath176 and @xmath177 obey @xmath172 symmetry . the @xmath173 symmetry is respected by the term @xmath178 , while a term like @xmath176 added to it breaks that and gives rise to @xmath179 symmetry . to illustrate the occurrence and effects of different types of symmetries , we consider three special pairs of densities ( @xmath58,@xmath59 ) . \(i ) for @xmath180 , eq . [ h+- ] and its partner for @xmath161 reduce to a pair of coupled equations , with linear first and second derivative terms and cubic gradient nonlinearities . quadratic nonlinearities are absent . the equations describe two kinematic waves moving with speed @xmath181 and @xmath182 where @xmath183 . the nonlinear couplings imply that each wave influences the evolution of the other . in order to study the dissipation of say the @xmath38mode , it is essential to move to the frame which co - moves with it , through the galilean shift @xmath184 , @xmath185 . evidently , in this frame , the @xmath41mode has a speed @xmath186 . the evolution equation is invariant under @xmath187 , @xmath188 but not @xmath171 , because of the linear @xmath189 and cubic nonlinear terms . the @xmath172 symmetry holds in the rest frame of @xmath164 mode . similarly the dissipation of the @xmath41mode can be studied by going to a frame which co - moves with it . it is easily seen that @xmath190 symmetry holds in this frame as well . each of the @xmath38 and @xmath41modes thus have the same symmetry as a single scalar field evolving through edwards - wilkinson dynamics with additional cubic nonlinearities . for the corresponding single - field system , we have @xmath191 ( though with multiplicative powers of logarithms ) @xcite , and we might expect the same behaviour here . \(ii ) for , on going to either of the frames in which the @xmath38 mode or the @xmath41 mode are at rest , we see that the @xmath179 symmetry applies for each of the fields ; the @xmath168 symmetry is broken by by quadratic nonlinear terms , and @xmath170 is broken because of linear first order and cubic terms . the most relevant terms at the linear fixed point are the quadratic nonlinear terms . thus we would expect these terms to govern the dissipation and give rise to the kpz value @xmath192 @xcite for both the modes . \(iii ) for @xmath193 , an interesting situation arises . in the co - moving frame of the @xmath41mode , the pair of equations have symmetry under @xmath188 and @xmath194 but not under @xmath171 . thus the @xmath161 field has @xmath172 symmetry , while the moving @xmath164 field has @xmath195 symmetry . the same symmetries hold in the rest frame of the @xmath38 wave . based on these observations , we expect @xmath103 for the @xmath41mode ( perhaps with multiplicative logarithmic corrections ) , and @xmath192 for the @xmath38 mode . thus based on simple considerations of symmetry , we would expect different types of behaviour in the cases i , ii and iii . these expectations are borne out by both numerical simulations and analytic calculations . the equal - space unequal - time autocorrelation function @xmath196 for the lattice model was investigated numerically for the three cases discussed above . our monte carlo results for @xmath197 are in accord with expectations based on symmetries of the continuum equations . in case i , we find @xmath198 with evidence for multiplicative logarithmic corrections for both modes . in case ii , we find @xmath199 for both modes , indicating kpz behaviour . finally , in case iii , we find that one mode has type i behaviour , while the other one has type ii behaviour , as expected on grounds of symmetry . an analytical treatment of the generalised langevin equations ( [ h+- ] ) is straightforward to carry out @xcite . the renormalised propagators and correlation functions for @xmath164 and @xmath161 are calculated in a 1-loop self - consistent scheme . in case iii in particular the character of the perturbation theory is controlled by the presence of kinematic waves . in contrast to what happens in @xcite , their effect can not be shifted away by galilean transformations , and we find as expected from the symmetry arguments above that @xmath200 and @xmath201 . the calculations for the other cases ( i and ii ) likewise give results consistent with the symmetry arguments above . we conclude by pointing out some connections to other models and discussing some open questions . as discussed in section 2 , the strongly phase separated phase of the lr model is closely related to the corresponding phase of the permutation - symmetric abc model @xcite especially at the symmetric point of each model where the detailed balance condition is valid . however , an analog of the threshold case of fluctuation - dominated phase ordering which occurs in the lr model has not yet been found in the abc or related models . in particular , the permutation non - symmetric three - component model of arndt et al @xcite shows a transition from a phase - separated phase to a disordered phase with a large correlation length @xcite , but the state at the transition point does not seem to show fdpo . the threshold case of the lr model is tantamount to the problem of hard - core particles sliding down a fluctuating surface . this leads to the question of whether fluctuation - dominated phase ordering survives if the surface in question evolves differently . a study of the problem with kpz dynamics for the surface shows that fdpo survives @xcite , though the value of the cusp exponent changes from that in the ew case . on the other hand , with the surface evolving according to das sarma - tamborenea dynamics @xcite , there is no cusp in @xmath202 ( eq . [ nonporod ] ) , and the porod law is recovered @xcite . another interesting direction to investigate is what happens if the falling particles are noninteracting . some results in this direction are available from a study of domain - wall dynamics in a model of growth of binary films @xcite . the continuum equations of motion in the wave phase are a one - dimensional reduced model of those governing the dynamics of small fluctuations of a flux lattice drifting through a clean type ii superconductor @xcite . these equations , or more properly the equations implied by ( [ coupled ] ) for @xmath203 , with some of the couplings set to zero , are burgers - like equations for magnetohydrodynamics ( mhd ) , as shown in @xcite . the dynamics in this kinematic - wave region of the phase diagram is largely governed by kpz - like critical exponents ( in particular @xmath192 ) . on special loci in the space of parameters , however , weak dynamic scaling prevails : if @xmath157 are the eigenmodes of the linearised versions of the equations , then the complete nonlinear theory on these loci is governed by @xmath204 for @xmath164 and @xmath161 respectively @xcite . finally , an important open question concerns the behaviour of the lr lattice model in higher dimensions . recently , it has been shown that a higher - dimensional version of the abc model continues to show strong phase separation @xcite . it would be interesting to investigate analogous questions for all phases of the lr model as well . so far , the only result along these lines is the demonstration of product measure along particular loci in the disordered phase in higher dimensions @xcite . the work summarised in this article began in collaboration with rangan lahiri whose untimely death in 1998 ended a scientific career of great promise . we dedicate this paper to his memory .	we discuss the properties of a one - dimensional lattice model of a driven system with two species of particles in which the mobility of one species depends on the density of the other . this model was introduced by lahiri and ramaswamy in the context of sedimenting colloidal crystals , and its continuum version was shown to exhibit an instability arising from linear gradient couplings . in this paper we review recent progress in understanding the full phase diagram of the model . there are three phases . in the first , the steady state can be determined exactly along a representative locus using the condition of detailed balance . the system shows phase separation of an exceptionally robust sort , termed strong phase separation , which survives at all temperature . the second phase arises in the threshold case where the first species evolves independently of the second , but the fluctuations of the first influence the evolution of the second , as in the passive scalar problem . the second species then shows phase separation of a delicate sort , in which long range order coexists with fluctuations which do not damp down in the large - size limit . this fluctuation - dominated phase ordering is associated with power law decays in cluster size distributions and a breakdown of the porod law . the third phase is one with a uniform overall density , and along a representative locus the steady state is shown to have product measure form . density fluctuations are transported by two kinematic waves , each involving both species and coupled at the nonlinear level . their dissipation properties are governed by the symmetries of these couplings , which depend on the overall densities . in the most interesting case , the dissipation of the two modes is characterized by different critical exponents , despite the nonlinear coupling .
we consider two - dimensional first passage percolation , where each pair of nearest neighbours of @xmath0 have an edge connecting them and each edge is equipped with a nonnegative random variable which is interpreted as the time it takes to traverse the edge ( see @xcite and @xcite for reviews ) . we specialize to the case where these passage times are i.i.d.exponentials ; this is the so called richardson model @xcite . by time scaling , we may assume without loss of genertality that the exponential distribution has mean one . for @xmath6 , let @xmath7 be the time taken to reach @xmath8 from @xmath9 , i.e. @xmath7 is the infimum over all paths from @xmath9 to @xmath8 of the sum of the passage times along the path . write @xmath10 for the origin and let , for @xmath11 , @xmath12 so that @xmath13 is the set of sites reached from the origin by time @xmath14 . it follows from the memoryless property of the exponential distribution that @xmath13 is a markov process , and the following interacting particle system formulation is natural . each site @xmath9 takes the value @xmath15 at time @xmath14 . let @xmath16 and write @xmath17 for the whole configuration of 0 s and 1 s at time @xmath14 . we may think of sites in state 0 as healthy and those in state 1 as infected . each site in state 1 remains in this state forever and tries to infect each of its four nearest neighbours at rate one , so that a site in state 0 flips to state 1 at rate equal to the number of nearest neighbours with state 1 . at time 0 , only the origin is infected . one of the most important results in first passage percolation is the existence of the so called time constant @xmath18 such that @xmath19 which is a consequence of kingman s subadditive ergodic theorem . this result in fact holds under much more general conditions than those considered in this paper . the precise value of @xmath18 is not known , although some bounds are available . we will need a lower bound on @xmath18 , and the best such bound we are aware of is @xmath20 as shown in @xcite . similar time constants exist for all directions . moreover , there is the following asymptotic shape result which , somewhat loosely speaking , states that analogues of ( [ kingman ] ) hold for all directions simultaneously : let @xmath21 ^ 2 , z\in b(t ) \mbox { such that } y+z = x\},\ ] ] i.e. @xmath22 is a `` fattened '' version of @xmath13 . then there exists a nonrandom compact convex set @xmath23 such that for all @xmath24 @xmath25 see e.g. @xcite . we will be interested in a different aspect of the evolution of infection , namely the * tree of infection * , to be denoted by @xmath26 . let @xmath27 be the graph with vertex set @xmath13 and edge set obtained as follows . for each @xmath28 , let @xmath29 be the edge connecting @xmath9 to the vertex @xmath8 from which @xmath9 got infected , and let @xmath30 be the edge set of @xmath27 . each vertex of @xmath27 gets a unique path to @xmath10 , so @xmath27 is indeed a tree . moreover , both the vertex set and the edge set of @xmath27 are increasing in @xmath14 , so the limiting object @xmath31 also exists . the tree structure of @xmath27 is inherited by @xmath26 . equivalently , we may define @xmath26 as the graph with vertex set @xmath0 and edge set given by @xmath32 our main interest is in the number of topological ends of @xmath26 , i.e. how many infinite self - avoiding paths starting at @xmath10 does @xmath26 contain ? let @xmath33 denote the number of such paths . by a standard compactness argument , @xmath34 . newman @xcite has shown that @xmath35 a.s . provided a certain hypothesis concerning uniformly bounded curvature of @xmath23 . ( see also @xcite for related results . ) the uniformly bounded curvature hypothesis is highly plausible , but has so far not been not proved . we shall prove the following result which is a small step towards the conjecture that @xmath35 a.s . * theorem 1.1 : * _ the number of topological ends @xmath33 of the tree of infection satisfies @xmath36 _ in order to prove this result , we shall first study a simple and natural model for competing spatial growth . we will now go on to describe this model , which is a variant of the richardson model and which we propose to call the * two - type richardson model. consider an interacting particle system on @xmath0 with state space @xmath2 , where 0 s may be thought of as healthy sites , and 1 s and 2 s as two different types of infection . the evolution is as follows . a site in state 1 stays in this state forever , and the same thing holds for a site in state 2 . both types of infected sites try to infect each of their nearest neighbours at rate one , so that a site in state 0 flips to state 1 at rate equal to the number of nearest neighbours in state 1 , and to state 2 at rate equal to the number of nearest neighbours in state 2 . we will write @xmath37 for the configuration at time @xmath14 ; @xmath37 will be an element of @xmath38 . note that if we disregard the type of infection ( i.e. if we watch this system evolve wearing a pair of glasses which prevents us from distinguishing between 1 s and 2 s ) , then the system behaves exactly as the ordinary ( one - type ) richardson model . this is an immediate consequence of the fact that the total flip rate of a site in state 0 equals its total number of infected nearest neighbours . we start at time 0 with an infection of type 1 at 0 * and an infection of type 2 at another site @xmath9 , all other sites being healthy . the model can then be described in first passage percolation terms : a site @xmath8 is infected at time @xmath39 , which we define as the infimum , over all paths starting at @xmath10 or @xmath9 and ending at @xmath8 , of the sum of the passage times along the path . since the distribution of the passage times of the edges is continuous , it is not hard to see that the infimum is in fact a.s . a minimum which is attained for a unique path . if this fastest path starts at @xmath10 , then @xmath8 gets infection of type 1 , otherwise it gets type 2 . we may think of the two - type richardson model as a crude model for two growing bacteria colonies ( or two political empires ) competing for space . it may happen that at some early stage , one of the types of infection completely surrounds the other type which then is prevented from growing indefinitely . write @xmath40 for the event that this does not happen , in which case both types of infection will grow indefinitely . the first question one would like to answer about the two - type richardson model is whether or not @xmath41 ( it is obvious that @xmath42 ) . the answer to this question is in fact independent of @xmath9 , as stated in the following proposition , where we write @xmath43 for the probability of @xmath40 with the described starting configuration . the proof will be given in section 2 . * proposition 1.2 : * _ for any @xmath44 , @xmath45 _ hence we may restrict attention to @xmath46 , i.e.to the two - type richardson model starting with a 1 at the origin , a 2 right next to the origin , and all other sites healthy . our main result on the two - type richardson model now says * theorem 1.3 : * @xmath47 so that with positive probability both bacteria colonies keep growing forever . inserting ( [ janson ] ) yields @xmath48 . let us now go back to the one - type richardson model and the question of the number @xmath33 of topological ends of the tree of infection . start with a single infected site at @xmath10 and suppose that the first site that @xmath10 infects is @xmath49 . it is then an immediate consequence of theorem 1.3 and the identification between the one - type richardson model and the two - type richardson model with types disregarded , that there is positive probability for the event that @xmath26 has two different self - avoiding paths to infinity : one which goes through @xmath49 and one which does not . hence * theorem 1.3 implies theorem 1.1. similarly , if it had turned out that infinite growth of both bacteria colonies had probability zero , then we would have been able to conclude that @xmath50 a.s . this remark is of course an empty statement , but we point it out anyway because it is feasible that a similar implication might be useful in some of the possible extensions of the model considered here ; see below . in order to prove theorem 1.3 , we will need the following proposition , which will be proved in section 2 . we think it is of some interest in its own , and it also seems related to questions concerning the roughness of the boundary of @xmath13 , studied e.g. in @xcite and @xcite . proposition 1.4 : * _ for any @xmath24 , there exist infinitely many @xmath51 in the right half - plane , such that @xmath52 > { \textstyle \frac{2\mu+1}{3 } } - \varepsilon,\ ] ] which is greater than @xmath53 when @xmath54 is small . in fact , the stronger result holds that for some @xmath55 @xmath56 > { \textstyle \frac{2\mu+1}{3}}-\varepsilon.\ ] ] _ inserting ( [ janson ] ) gives that the right hand sides in ( [ preprop ] ) and ( [ prop ] ) can be made greater that @xmath57 . intuitively , proposition 1.4 says that there are sites in @xmath0 arbitrarily far away from the origin which `` strongly feel '' from which direction the infection is coming . it seems reasonable to believe that the @xmath58 in ( [ prop ] ) in fact is a limit , and independent of @xmath59 . apart from proving the conjecture about @xmath26 a.s . having infinitely many topological ends , there are various other ways in which one might want to extend and improve the results of this paper . an obvious question is what happens when @xmath0 is replaced by @xmath60 for @xmath61 . another direction is to allow passage time distributions other than the exponential . the markovian behaviour of @xmath13 is then lost whence the interacting particle system formulation becomes less natural , but the questions about @xmath26 remain just as natural as for the exponential case . an inspection of the proof of proposition 1.4 shows that the result @xmath62 extends to the case where the passage times of the edges have a continuous distribution with a hazard rate which is bounded between two constants @xmath63 satisfying @xmath64 , ( here @xmath18 still denotes the time constant when the passage time distribution is a mean one exponential ) . one can also ask what happens if we extend the two - type richardson model to having three or more different types , in the obvious way . for this extension , showing that @xmath65 different types simultaneously can grow indefinitely from a finite starting configuration is equivalent to showing that @xmath66 . yet another direction of generalization would be to allow the two types to have different infection rates @xmath67 and @xmath68 , and for this extension we conjecture that @xmath69 whenever @xmath70 . since theorem 1.3 implies theorem 1.1 , it only remains to prove propositions 1.2 and 1.4 , and theorem 1.3 . the remainder of the paper is devoted to this task . the hard work is in the proof of proposition 1.4 . * proof of proposition 1.2 : * suppose @xmath71 , and let @xmath72 be the circle of radius @xmath73 ( here @xmath74 is the euclidean norm ) centered at @xmath10 . by conditioning on the first @xmath75 infections ( for some @xmath75 which we need not define explicitly here ) we can find @xmath76 and a configuration @xmath77 such that @xmath78 restricted to @xmath79 contains only 1 s and 2 s , and such that @xmath80 . note that @xmath81 and @xmath82 both must be connected sets ( in the graph - theoretic sense with edges between all nearest neighbours ) . let @xmath83 be the set the set of infected sites in @xmath78 ; by choice of @xmath84 also @xmath85 is connected . define @xmath86 i.e. @xmath87 is the set of infected sites with at least one healthy neighbour . it is easy to check , using the strong markov property , that @xmath88 does not depend on the values of @xmath78 on @xmath89 . by choice of @xmath84 , we may now construct a configuration @xmath90 such that * proof of proposition 1.4 : * our aim is to show that ( [ prop ] ) holds for fixed @xmath24 and some @xmath59 . we will consider the evolution of @xmath13 inside the half - strip @xmath102 for some large @xmath65 ( taking @xmath103 will suffice ) . writing @xmath104 for the number of infected sites in @xmath105 at time @xmath14 , we have , as an easy consequence of ( [ kingman ] ) , that @xmath106 for @xmath1 , write @xmath107 for the flip rate of @xmath9 at time @xmath14 , and write @xmath108 for @xmath109 . it follows from ( [ lim4 ] ) and an application of the strong law of large numbers that @xmath110 now write @xmath111 ( @xmath112 as in good ) for the number of infected sites in @xmath105 at time @xmath14 which became infected after their nearest neighbour to the left , and write @xmath113 ( @xmath114 for bad ) for the corresponding number of sites which became infected after their nearest neighbour to the right . by keeping track of the left - right nearest neighbour pairs , we have , as for ( [ lim4 ] ) , @xmath115 although note that @xmath116 holds only approximately because an individual site may sometimes be infected after neither or both of its left - right nearest neighbours . write @xmath117 for @xmath118 where this time the sum is taken over uninfected sites in @xmath105 whose nearest neighbour to the left is already infected at time @xmath14 , and define @xmath119 analogously . we have as in ( [ lim5 ] ) that @xmath120 and @xmath121 in order to show that ( [ prop ] ) holds for some @xmath122 , it suffices to prove that @xmath123 to this end , it is sufficient to show that @xmath124 so this is what we will proceed to prove . we now look at a configuration @xmath125 , which will serve as an example of what @xmath126 may look like at some reasonably large time point @xmath14 . since @xmath127 for @xmath128 and all sufficiently large @xmath14 a.s . , we assume that @xmath129 for @xmath128 . also @xmath130 for fixed @xmath14 and all but finitely many @xmath1 , so we furthermore assume that @xmath131 for all but finitely many @xmath132 . now pick @xmath133 and write @xmath134 for @xmath135 restricted to @xmath136 , i.e. @xmath134 is the configuration on two infinite adjacent horizontal rows . given @xmath134 , we may partition @xmath137 into 1-blocks , 0-blocks , down - blocks , up - blocks and excursion - blocks as in figure 1 . the blocks are determined ( uniquely , given @xmath134 ) as follows . a 1-block is a maximal sequence of @xmath138 $ ] s and a 0-block is a maximal sequence of @xmath139 $ ] s , while down - blocks , up - blocks and excursion - blocks are maximal sequences consisting of @xmath140 $ ] s and @xmath141 $ ] s . all blocks are finite except for the final 0-block . a 1-block is always followed either by an up - block or by an excursion - block . a finite 0-block is always followed by an up - block or an excursion - block . a down - block is always followed by a 0-block while an up - block always is followed by a 1-block . an excursion - block , finally , is always surrounded either by two 1-blocks or by two 0-blocks . down - blocks and up - blocks ( but no other blocks ) may be empty ; an empty down - block is inserted whenever a @xmath138 $ ] is followed by a @xmath139 $ ] , and , similarly , an empty up - block is inserted whenever a @xmath139 $ ] is followed by a @xmath138 $ ] . call 1-blocks and 0-blocks * pure * , and call other blocks * mixed. call a pair of nearest neighbours hot * if one of the sites has value 1 and the other has value 0 . each hot pair will be associated with a mixed block as follows . if a @xmath138 $ ] is followed by a @xmath139 $ ] , then the two hot pairs are associated with the corresponding empty down - block , and similarly when a @xmath139 $ ] is followed by a @xmath138 $ ] . any other hot pair intersects exactly one mixed block and is then associated with this block . we now consider rates of infection between sites in @xmath142 when @xmath143 . write @xmath144 for the weighted total infection rate for the hot pairs associated with the mixed block @xmath145 , with weight @xmath146 for horizontal infections and weight 1 for vertical infections . the reason for these weights is that each horizontal infection is accounted for in @xmath134 for two different values of @xmath147 ( this is not true for rows @xmath5 and @xmath65 , but that will be corrected for later ) . write @xmath148 ( resp.@xmath149 ) for the corresponding sum where only good ( resp . bad ) infections are counted . the following is easily checked to hold for @xmath144 , @xmath148 and @xmath149 . @xmath150 write @xmath151 ( resp . @xmath152 and @xmath153 ) for @xmath154 ( resp . @xmath148 and @xmath149 ) summed over all mixed blocks in @xmath134 . note that since @xmath155 starts with a 1-block and ends with a 0-block , we have that the number of down - blocks exceeds the number of up - blocks by exactly 1 . using this observation , and the above table , we may check ( e.g. via induction over the number of down - blocks ) that @xmath156 with equality if and only if the only mixed blocks in @xmath151 are empty down - blocks and nonempty up - blocks with @xmath157 for each up - block @xmath145 . applying ( [ lim8 ] ) to @xmath158 and summing over @xmath147 , we get @xmath159 for all sufficiently large @xmath14 a.s . , whence @xmath160 since the number of infections up to time @xmath14 in rows @xmath5 and @xmath65 is asymptotically @xmath161 , we have that @xmath162 and for the same reason @xmath163 hence , we may modify ( [ lim9 ] ) to get @xmath164 combining this with ( [ lim6 ] ) yields @xmath165 i.e. @xmath166 which in conjunction with ( [ lim5 ] ) implies @xmath167 taking @xmath103 , this implies ( [ lim7 ] ) , so the proof is complete . * proof of theorem 1.3 : * suppose for contradiction that @xmath168 for some @xmath24 . then , by symmetry , the probability that infection of type 2 eventually stops growing is greater than @xmath169 . this implies @xmath170 < { \textstyle 1-\frac{2 - 2\mu}{3}}-\varepsilon = { \textstyle\frac{2\mu+1}{3}}-\varepsilon,\ ] ] where @xmath59 can be chosen as in proposition 1.4 . by reflecting the realization of passage times in the line @xmath171 , we see that the pair @xmath172 has the same joint distribution as @xmath173 , so that @xmath174 < { \textstyle \frac{2\mu+1}{3 } } - \varepsilon\ ] ] contradicting ( [ prop ] ) . @xmath101	an interacting particle system modelling competing growth on the @xmath0 lattice is defined as follows . each @xmath1 is in one of the states @xmath2 . @xmath3 s and @xmath4 s remain in their states forever , while a @xmath5 flips to a @xmath3 ( resp . a @xmath4 ) at a rate equal to the number of its neighbours which are in state @xmath3 ( resp . @xmath4 ) . this is a generalization of the well known richardson model . @xmath3 s and @xmath4 s may be thought of as two types of infection , and @xmath5 s as uninfected sites . we prove that if we start with a single site in state @xmath3 and a single site in state @xmath4 , then there is positive probability for the event both types of infection reach infinitely many sites . this result implies that the spanning tree of time - minimizing paths from the origin in first passage percolation with exponential passage times has at least two topological ends with positive probability . * keywords : * first passage percolation , richardson s model , tree of infection , competing growth . * ams subject classification : * primary 60k35 , secondary 82b43 . 146 mm 210 mm
the @xmath2 nuclei at r - process path are assumed to play a significant role in providing nuclear abundances about @xmath3 @xcite . since nuclei contributing to this peak are extremely neutron - rich and are not accessible experimentally , there prevail conflicting view points @xcite on the strength of the shell effects near the neutron drip line . in this work , we examine how the shell effects evolve with isospin in the region of the astrophysically important magic number @xmath2 near the neutron drip line . using the experimental data on the waiting - point nucleus @xmath1zn ( @xmath4 ) , we explore the shell effects near the r - process path in the framework of the relativistic hartree - bogoliubov ( rhb ) theory with the vector self - coupling of @xmath0 meson . in this approach , the shell effects about the stability line have been described successfully @xcite . the rmf lagrangian describes the nucleons as dirac spinors moving in meson fields and is given by @xcite @xmath5 where @xmath6 is the bare nucleon mass and @xmath7 is its dirac spinor . in addition , we have the scalar meson ( @xmath8 ) , isoscalar vector mesons ( @xmath0 ) , isovector vector mesons ( @xmath9 ) and the photons @xmath10 , with the masses @xmath11 , @xmath12 and @xmath13 and the coupling constants @xmath14 , @xmath15 , @xmath16 , respectively . a nonlinear scalar potential @xmath17 for @xmath8-mesons is used for a realistic description of finite nuclei . recently , we have also added the non - linear vector self - coupling of @xmath0-meson @xcite which is represented by @xmath18 . for nuclei near a drip line , a coupling to the continuum is required . a self - consistent treatment of pairing is also necessary . we employ the framework of rhb theory which takes into account both these features . the corresponding relativistic dirac - hartree - bogoliubov ( rhb ) equations are obtained as @xmath19 where @xmath20 are quasiparticle energies and the coefficients @xmath21 and @xmath22 are four - dimensional dirac spinors . the pairing potential @xmath23 in eq . ( [ rhb ] ) is given by @xmath24 the rhb calculations are performed by expanding fermionic and bosonic wavefunctions in 20 oscillator shells for a spherical configuration . we have used the force nl - sh @xcite with the non - linear scalar self - coupling and the forces nl - sv1 and nl - sv2 with both the scalar and the vector self - couplings . the forces nl - sv1 and nl - sv2 have recently been developed in order to soften the high - density equation of state of nuclear matter and are shown to improve the ground - state properties of nuclei @xcite . in our previous work , it was shown that the existing nuclear forces based upon the nonlinear scalar self - coupling of @xmath8-meson exhibit shell effects which were stronger than suggested by the experimental data . this problem was solved by introducing the nonlinear vector self - coupling of @xmath0-meson in the rhb theory @xcite . having our working basis established , we explore the shell effects about @xmath1zn ( @xmath4 ) . @xmath1zn is a waiting point nucleus and here the r - process path comes closest to the @xmath25-stability line . we show in fig . 1 the results on @xmath26 values for the zn isotopes . the kink at @xmath4 ( fig 1.a ) shows that nl - sh ( scalar self - coupling only ) exhibits shell effects which are much stronger than the experimental data . in comparison , the shell gap with nl - sv2 ( both the scalar and vector self - couplings ) @xcite is reduced as compared to nl - sh . the shell gap is , however , still larger than the experimental one . in fig . 1(b ) we show the @xmath26 values obtained with the force nl - sv1 . the slope of the kink shows that nl - sv1 is able to reproduce well the empirical shell effects in the waiting point region . a comparison with the hfb+skp results shows that the shell effects with skp are strongly quenched . the shell effects were shown to be quench with skp already at the stability line @xcite . this behaviour is evidently due to the high effective mass @xmath27 . with the formalism established in @xcite and tested in fig . 1 , we extend it to explore the inaccessible region of the neutron drip - line about @xmath2 . in order to visualize the evolution of shell effects , we have chosen the isotopic chains of kr , sr , zr and mo ( @xmath28 ) . the results with rhb using various forces are shown in fig . 2 . for a given force , the shell gap at @xmath2 shows a steady decrease in moving from mo ( a ) to kr ( d ) . nuclei become increasingly unbound and a coupling to the continuum arises in going to larger neutron to proton ratios . the results with the non - linear scalar coupling ( nl - sh ) show a shell gap at @xmath2 , which is largest amongst all the forces . similarly strong shell effects with nl - sh were also shown for @xmath2 nuclei near the drip line in ref . @xcite within the bcs pairing . with the force nl - sv2 with the vector self - coupling of @xmath0-meson , the shell effects are milder as compared to nl - sh for all the chains . the results ( fig . 2 ) with our benchmark force nl - sv1 show that the shell gap at @xmath2 is reduced as compared to nl - sv2 . this is again similar to that observed for the zn isotopes . however , the shell effects with nl - sv1 are still stronger as compared to the hfb+skp results as shown in the figure . this is especially true for mo and zr isotopes at n=82 , where the r - process path is assumed to pass through . the quenching shown by skp in this region is as expected . this is consistent with a strong quenching exhibited by hfb+skp at the stability line @xcite and also near the waiting - point nucleus @xmath1zn as shown in fig . 1(b ) . in comparison , the rhb approach with nl - sv1 shows stronger shell effects about the stability line @xcite as well as in the waiting - point region at @xmath4 , which is consistent with the experimental data . the shell effects with nl - sv1 become successively weaker as one moves to nuclei with higher isospin such as @xmath29sr ( fig . 2(c ) ) and @xmath30kr ( fig . 2(d ) ) . it is worth making a special mention of the kr nucleus . all the forces ( except nl - sh ) show a complete washing out of the shell effects . this stems from the fact that for @xmath30kr the fermi energy is very close to the continuum . the shell gap ceases to exist and consequently binding energy of an extra neutron is nearly zero . this implies that the shell effects for @xmath30kr ( @xmath2 ) are washed out . such nuclei at the drip line ( @xmath31 ) are of little interest to the r - process as these nuclei can not bind an additional neutron . however , as the r - process path passes through @xmath32 mev , it is seen ( figs . 2.a and 2.b ) that nuclei which would contribute to the r - process show a persistence of stronger shell gaps in contrast to that seen for kr and sr . figures 2(a ) and 2(b ) show that with nl - sv1 the shell effects for mo ( z=42 ) and zr ( z=40 ) nuclei at @xmath2 are much stronger than with skp . for r - process nuclei with @xmath33 , the shell gaps are expected to be even larger . we have reproduced the shell effects about the waiting - point nucleus @xmath1zn with the vector self - coupling of @xmath0-meson in the rhb theory . with this basis , it is shown that the shell effects near the r - process path about @xmath2 remain strong vis - a - vis a quenching exhibited by hfb+skp . the quenching shown by skp is not consistent with the available data . it is noteworthy that on the basis of the results with skp , a quenching has been requested for an improved fit to the global r - process abundances @xcite . since skp is not able to reproduce the shell effects at the stability line and in the waiting - point region , an improved fit alone does not necessarily imply a shell quenching near the r - process path . thus , alternative mechanisms for nucleosynthesis need to be looked into . et al . _ , j. * 403 * , 216 ( 1993 ) . sharma et al.,phys . rev . lett.72 , 1431 ( 1994 ) . j. dobaczewski _ et al . _ , phys . rev . lett . * 72 * , 981 ( 1994 ) . sharma , a.r . farhan and s. mythili , phys . c * 61 * 054306 ( 2000 ) . b.d . serot and j.d . walecka , adv . * 16 * , 1 ( 1986 ) . bodmer , nucl . a * 526 * , 703 ( 1991 ) . sharma , m.a . nagarajan and p. ring , phys . b * 312 * , 377 ( 1993 ) . sharma , in preparation ( 2000 ) .	we have studied the evolution of the shell structure of nuclei near the neutron drip line in the relativistic hartree - bogoliubov ( rhb ) theory with the vector self - coupling of @xmath0 meson . the experimental data on the shell effects about the waiting - point nucleus @xmath1zn are reproduced successfully . it is shown that the shell effects at @xmath2 near the r - process path remain strong . a quenching exhibited by the hfb+skp approach is shown not to be compatible with the available data .
an @xmath1 _ doubly stochastic matrix _ refers to a matrix of non - negative entries whose columns and rows each sum to @xmath2 . the doubly stochastic matrices form a convex subset of all @xmath1 matrices in fact a convex polytope , whose extreme points are in bijective correspondence with the @xmath3 permutations on @xmath4-letters , according to a theorem of birkhoff @xcite and von neumann @xcite . for example , the @xmath5 doubly stochastic matrices , @xmath6 form a 4-dimensional polytope with 6 vertices . shortly after proving this characterization , birkhoff ( * ? ? ? * problem 111 ) initiated the search for a infinite - dimensional generalization , thus stimulating a line of research which remains fruitful even today . a _ doubly stochastic measure _ on the square refers to a non - negative borel probability measure on @xmath7 ^ 2 $ ] whose horizontal and vertical marginals both coincide with lebesgue measure @xmath8 on @xmath7 $ ] . the set of doubly stochastic measures forms a convex set we denote by @xmath9 ( which is weak-@xmath10 compact in the banach space dual to continuous functions @xmath11 ^ 2)$ ] normed by their suprema @xmath12 ) . a measure is said to be _ extremal _ in @xmath9 if it can not be decomposed as a convex combination @xmath13 with @xmath14 and @xmath15 , except trivially with @xmath16 . since the krein - milman theorem asserts that convex combinations of extreme points are dense ( in any compact convex subset of a topological vector space , figure [ fig.extreme ] ) , it is natural to want to characterize the extreme points of @xmath9 . another motivation for such a characterization is that every continuous linear functional on @xmath9 is minimized at an extreme point . whether or not this extremum is uniquely attained can be an interesting question : in figure [ fig.extreme ] the horizontal coordinate is minimized at a single point but maximized at two extreme points ( and along the segment joining them ) . motivated by applications like the optimization problem just mentioned , we prefer to formulate the question in slightly greater generality , by replacing the two copies of @xmath17,\lambda)$ ] with probability spaces @xmath18 and @xmath19 , where @xmath20 and @xmath21 are each subsets of a complete separable metric space , and @xmath22 and @xmath23 are borel probability measures on @xmath20 and @xmath21 respectively . this widens applicability of the answer to this question without increasing its difficulty . letting @xmath24 denote the borel probability measures on @xmath25 having @xmath22 and @xmath23 for marginals , we wish to characterize the extreme points of the convex set @xmath24 . ideally , as in the finite - dimensional case , this characterization would be given in terms of some geometrical property of the support of the measure @xmath0 in @xmath25 . indeed , if @xmath26 and @xmath27 are finite , our problem reduces to characterizing the extreme points of the convex set @xmath28 of @xmath29 matrices with prescribed column and row sums : @xmath30 a matrix @xmath31 is well - known to be extremal in @xmath28 if and only if it is _ acyclic _ , meaning for every sequence @xmath32 of non - zero entries occupying @xmath33 distinct columns and @xmath34 distinct rows , the product @xmath35 must vanish see figure [ fig.acyclic ] or denny @xcite , where the terminology _ aperiodic _ is used . similarly , a set @xmath36 is acyclic if for every @xmath33 distinct points @xmath37 and @xmath38 , at least one of the pairs @xmath39 lies outside of @xmath40 . a functional analytic characterization of extremality was supplied by douglas @xcite and by lindenstrauss @xcite : it asserts that @xmath0 is extremal in @xmath24 if and only if @xmath41 is dense in @xmath42 . although this result is a useful starting point , it is not quite the characterization we desire for applications , since it is not easily expressed in terms of the geometry of the support of @xmath0 . significant further progress was made by bene and tpn , who showed every extremal doubly stochastic measure vanishes outside some acyclic subset @xmath36 @xcite . hestir and williams refined this condition , showing that it becomes sufficient under an additional borel measurability hypothesis which , unfortunately , is not always satisfied @xcite . some of the subtleties of the problem were indicated already by losert s counterexamples losert82 . the difficulty of the problem resides partly in the fact that any geometrical characterization of optimality must be invariant under arbitrary measure - preserving transformations applied independently to the horizontal ( abscissa ) and vertical ( ordinate ) variables . in this manuscript we review this line of research , clarifying the nature of the gap separating necessity from sufficiency and pointing out that it can be narrowed slightly by replacing the borel @xmath43-algebra with suitably adapted measure - completions . we conclude by describing an application to the question of uniqueness in optimal transportation , which is one of the original and most important examples of an infinite - dimensional program @xcite , and appears naturally in applications @xcite @xcite . it arises when one wants to use a continuum of sources to supply a continuum of sinks ( modeled by @xmath22 and @xmath23 respectively ) as efficiently as possible . the question addressed is to identify cost functions @xmath44 on the product space @xmath25 whose minimum expected value against measures in @xmath24 is uniquely attained . when @xmath20 and @xmath21 are differentiable manifolds and @xmath45 , to guarantee uniqueness it turns out to be sufficient that @xmath46 imply @xmath47 has no critical points , except perhaps for a single global maximum and a single global minimum . this generalizes to some compact manifolds @xmath20 a criterion of gangbo @xcite , carlier @xcite , levin @xcite and ma , trudinger and wang @xcite , which asserts that the absence of critical points implies uniqueness ; ( their condition further implies that almost every source supplies a single sink , thus solving another transportation problem first posed by monge @xcite , which our condition does not do ) . when satisfied , our criterion implies that the manifold @xmath20 , if compact , has the topology of the sphere . uniqueness , however , remains an interesting open question for compact manifolds which are not topological spheres . this surprising application was first developed in an economic context by chiappori , mccann , and nesheim @xcite . before recalling the characterization of interest , let us develop a bit of notation in a simpler setting , and a key argument that we shall require . impatient or knowledgeable readers can skim the present section and proceed directly to the final sections below . let @xmath20 and @xmath21 be subsets of complete separable metric spaces , and fix a non - negative borel measure @xmath22 on @xmath20 . suppose @xmath48 is @xmath22-measurable , meaning @xmath49 is in the @xmath43-algebra completion of the borel subsets of @xmath20 with respect to the measure @xmath22 , whenever @xmath50 is relatively borel in @xmath21 . then a borel measure on @xmath21 is induced , denoted @xmath51 and called the _ push - forward _ of @xmath22 through @xmath52 , and given by @xmath53 : = \mu[f^{-1}(b)]\ ] ] for each borel @xmath54 . defining the projections @xmath55 and @xmath56 on @xmath25 , this notation permits the horizontal and vertical marginals of a measure @xmath57 on @xmath25 to be expressed as @xmath58 and @xmath59 respectively . the next lemma shows that any measure supported on a graph can be deduced from its horizontal marginal . it improves on lemma 2.4 of @xcite and various other antecedents , by using an argument from villani s theorem 5.28 @xcite to extract @xmath22-measurability of @xmath52 as a conclusion rather that a hypothesis . as work of , e.g. , hestir and williams @xcite implies , although measures on graphs are extremal in @xmath60 , the converse is far from being true ; this peculiarity is an inevitable consequence of the infinite divisibility of @xmath18 . [ pure implies unique ] let @xmath61 and @xmath62 be subsets of complete separable metric spaces , and @xmath63 a @xmath64-finite borel measure on the product space @xmath65 . denote the horizontal marginal of @xmath66 by @xmath67 . if @xmath66 vanishes outside the graph of @xmath68 , meaning @xmath69 has zero outer measure , then @xmath52 is @xmath70-measurable and @xmath71 . since outer - measure is subadditive , it costs no generality to assume the subsets @xmath61 and @xmath62 are in fact complete and separable , by extending @xmath0 in the obvious ( minimal ) way . any @xmath64-finite borel measure @xmath0 is regular and @xmath43-compact on a complete separable metric space ; e.g. p. 255 of @xcite or theorem i-55 of @xcite . since @xmath0 vanishes outside @xmath72 , there is an increasing sequence of compact sets @xmath73 whose union @xmath74 contains the full mass of @xmath0 . compactness of @xmath75 implies continuity of @xmath52 on the compact projection @xmath76 . thus the restriction @xmath77 of @xmath52 to @xmath78 is a borel map whose graph @xmath79 is a @xmath43-compact set of full measure for @xmath0 . we now verify that @xmath0 and @xmath80 assign the same mass to each borel rectangle @xmath81 . since @xmath82 we find @xmath83 proving @xmath84 . taking @xmath85 and @xmath86 shows @xmath87 is @xmath88-negligible . since @xmath89 differs from the borel map @xmath90 only on the @xmath88-negligible complement of the @xmath43-compact set @xmath91 , we conclude @xmath52 is @xmath88-measurable and @xmath92 as desired . the preceding lemma shows that any measure concentrated on a graph is uniquely determined by its marginals ; @xmath0 is therefore extremal in @xmath93 . as the results of the next section show , the converse is far from being true . in this section we adapt hestir and williams @xcite notion of a _ numbered limb system _ to @xmath25 . using the axiom of choice , hestir and williams deduced from the acyclicity condition of bene and tpn @xcite that each extremal doubly stochastic measure vanishes outside some numbered limb system . conversely , they showed that vanishing outside a number limb system is sufficient to guarantee extremality of a doubly stochastic measure , provided the graphs ( and antigraphs ) comprising the system are borel subsets of the square . our main theorem gives a new proof of this converse in the more general setting of subsets @xmath25 of complete separable metric spaces , and under a slightly weaker measurability hypothesis on the graphs and antigraphs . a simple example shows that some measurability hypothesis is nevertheless required . in the next section , we shall see how this converse is germane to the question of uniqueness in optimal transportation . given a map @xmath94 on @xmath95 , we denote its graph , domain , range , and the graph of its ( multivalued ) inverse by @xmath96more typically , we will be interested in the @xmath97 of a map @xmath98 . [ numbered limb system ] let @xmath61 and @xmath62 be borel subsets of complete separable metric spaces . a relation @xmath99 is a _ numbered limb system _ if there is a countable disjoint decomposition of @xmath100 and of @xmath101 with a sequence of maps @xmath102 and @xmath103 such that @xmath104 , with @xmath105 for each @xmath106 . the system has ( at most ) @xmath107 limbs if @xmath108 for all @xmath109 . notice the map @xmath110 is irrelevant to this definition though @xmath111 is not ; we may always take @xmath112 , but require @xmath113 . the point is the following theorem and its corollary , which extends and relaxes the result proved by hestir and williams for lebesgue measure @xmath114 on the interval @xmath115 $ ] . in it , @xmath24 denotes the set of non - negative borel measures on @xmath25 having @xmath116 and @xmath117 for marginals . as in the preceding lemma , we say @xmath0 _ vanishes _ outside of @xmath36 if @xmath0 assigns zero outer measure to the complement of @xmath40 in @xmath25 . [ hestirwilliams ] + let @xmath61 and @xmath62 be subsets of complete separable metric spaces , equipped with @xmath43-finite borel measures @xmath22 on @xmath20 and @xmath23 on @xmath21 . suppose there is a numbered limb system @xmath118 with the property that @xmath119 and @xmath120 are @xmath0-measurable subsets of @xmath25 for each @xmath121 and for every @xmath122 vanishing outside of @xmath40 . if the system has finitely many limbs or @xmath123<\infty$ ] , then at most one @xmath122 vanishes outside of @xmath40 . if such a measure exists , it is given by @xmath124 where @xmath125 here @xmath126 is measurable with respect to the @xmath127 completion of the borel @xmath43-algebra . if the system has @xmath128 limbs , @xmath129 for @xmath130 , and @xmath131 and @xmath132 can be computed recursively from the formulae above starting from @xmath133 . let @xmath134 be a numbered limb system whose complement has zero outer measure for some @xmath43-finite measure @xmath135 . this means that @xmath136 gives a disjoint decomposition of @xmath137 and of @xmath138 , and that @xmath139 for each @xmath140 . assume moreover , that @xmath141 and @xmath142 are @xmath0-measurable for each @xmath121 . we wish to show @xmath0 is uniquely determined by @xmath22 , @xmath23 and @xmath40 . the graphs @xmath143 are disjoint since their domains @xmath144 are disjoint , and the antigraphs @xmath145 are disjoint since their domains @xmath146 are . moreover , @xmath143 is disjoint from @xmath147 for all @xmath148 : @xmath149 prevents @xmath143 from intersecting @xmath150 unless @xmath151 since the domains @xmath152 are disjoint , and @xmath143 can not intersect @xmath153 since @xmath154 is disjoint from @xmath155 . let @xmath132 denote the restriction of @xmath0 to @xmath156 for @xmath34 even and to @xmath157 for @xmath34 odd . then @xmath158 by our measurability hypothesis , and @xmath132 restricts to a borel measure on @xmath159 if @xmath34 is even , and on @xmath160 if @xmath34 odd . defining the marginal projections @xmath161 and @xmath162 , setting @xmath163 if @xmath34 even and @xmath164 if @xmath34 odd yields ( [ alternating representation ] ) and the @xmath165-measurability of @xmath166 immediately from lemma [ pure implies unique ] . since @xmath167 vanishes outside @xmath168 , from @xmath169 we derive @xmath170 . for @xmath34 even , @xmath171 vanishes outside @xmath172 , while for @xmath34 odd , @xmath171 vanishes outside @xmath173 , which is disjoint from @xmath168 unless @xmath174 . thus @xmath175 . the formula ( [ alternating marginals ] ) for @xmath176 follows from similar considerations . it remains to show the representation ( [ alternating representation])([alternating marginals ] ) specifies @xmath177 uniquely for all @xmath140 , and hence determines @xmath178 uniquely . if the system has @xmath179 limbs , @xmath180 for @xmath109 and hence @xmath181 . we can compute @xmath165 and @xmath182 starting with @xmath133 , and then recursively from the formulae above for @xmath183 , so the formulae represent @xmath66 uniquely . if instead @xmath40 has countably many limbs , suppose there are two finite borel measures @xmath66 and @xmath184 vanishing outside of @xmath40 and having the same marginals @xmath185 and @xmath186 . for each @xmath187 , recall that @xmath188is measurable with respect to both @xmath0 and @xmath189 . given @xmath190 , take @xmath107 large enough so that both @xmath66 and @xmath184 assign mass less than @xmath191 to @xmath192 . set @xmath193 and @xmath194 and denote their marginals by @xmath195 and @xmath196 . observe that both @xmath197 and @xmath198 are concentrated on the same numbered limb system ; it has finitely many limbs , and the differences @xmath199 and @xmath200 between the marginals of @xmath201 and @xmath202 have total variation at most @xmath203 . since the @xmath204 are mutually singular , as are the @xmath205 , we find the sum of the total variations of @xmath206is bounded : @xmath207 . using ( [ alternating representation ] ) to derive @xmath208and summing on @xmath34 yields @xmath209 . since @xmath210 and @xmath211 as @xmath212 , we conclude @xmath213 to complete the uniqueness proof . as in hestir and williams @xcite , the uniqueness theorem above implies extremality as an immediate consequence . [ c : support characterization ] let @xmath20 and @xmath21 be subsets of complete separable metric spaces , equipped with @xmath43-finite borel measures @xmath22 on @xmath20 and @xmath23 on @xmath21 . suppose there is a numbered limb system @xmath118 with the property that @xmath119 and @xmath120 are @xmath0-measurable subsets of @xmath25 for each @xmath121 , for every @xmath122 vanishing outside of @xmath40 . if the system has finitely many limbs or @xmath123<\infty$ ] , then any measure @xmath122 vanishing outside of @xmath40 is extremal in the convex set @xmath24 . suppose a measure @xmath122 vanishes outside a numbered limb system @xmath40 satisfying the hypotheses of the corollary . if @xmath214 with @xmath215 and @xmath14 , then @xmath216 and @xmath217 , so both @xmath218 and @xmath219 vanish outside of @xmath40 . according to theorem [ hestirwilliams ] , they are uniquely determined by @xmath220 and their marginals , hence @xmath221 to establish the corollary . the following example confirms that a measurability gap still remains between the necessary and sufficient conditions for extremality . it is a close variation on the standard example of a non - lebesgue measurable set from real analysis . together with the lemma and theorem preceding , this example makes clear that measurability is required only to allow the graphs to be separated from each other and from the antigraphs in an additive way . + let @xmath8 denote lebesgue measure and define the maps @xmath222 and @xmath223 ( mod 1 ) on the unit interval @xmath224 $ ] . notice @xmath225 ^ 2 $ ] supports the doubly stochastic measure @xmath226 for @xmath227 and @xmath228 ; ( both measures are extremal in @xmath9 by corollary [ c : support characterization ] ) . irrationality of @xmath229 implies @xmath230 is an acyclic set , hence can be expressed as a numbered limb system according to hestir and williams @xcite . on the other hand , there are doubly stochastic measures such as @xmath231 which vanish outside of @xmath40 but which are manifestly not extremal . in this section we illustrate the significance of the foregoing results by applying them to the uniqueness question for optimal transportation on manifolds . given subsets @xmath20 and @xmath21 of complete separable metric spaces equipped with borel probability measures , representing the distributions @xmath22 of production on @xmath20 and @xmath23 of consumption on @xmath21 , the kantorovich - koopmans @xcite @xcite transportation problem is to find @xmath232 correlating production with consumption so as to minimize the expected transportation cost @xmath233 against some continuous function @xmath234 . hereafter we shall be solely concerned with the case in which @xmath20 is a differentiable manifold , @xmath22 is absolutely continuous with respect to coordinates on @xmath20 , and the cost function @xmath45 is differentiable with local control on the magnitude of its @xmath235-derivative @xmath236 uniformly in @xmath237 ; for convenience we also suppose @xmath21 to be a differentiable manifold and @xmath238 is bounded , though though this is not really necessary : substantially weaker assumptions also suffice @xcite . in this setting one immediately asks whether the infimum ( [ mkpa ] ) is uniquely attained . since attainment is evident , the question here is uniqueness . if @xmath238 satisfies a _ twist _ condition , meaning @xmath239 has no critical points for @xmath240 , then not only is the minimizing @xmath0 unique , but its mass concentrates entirely on the graph of a single map @xmath241 ( a numbered limb system with one limb ) , thus solving a form of the transportation problem posed earlier by monge @xcite @xcite . this was proved in comparable generality by gangbo @xcite , carlier @xcite , levin @xcite , and ma , trudinger and wang @xcite , building on the more specific examples of strictly convex cost functions @xmath242 in @xmath243 analyzed by caffarelli @xcite and gangbo and mccann @xcite , and in case @xmath244 by abdellaoui and heinich , brenier , cuesta - albertos , matran , and tuero - diaz , cullen and purser , knott and smith , and rschendorf and rachev ; see @xcite @xcite @xcite . adding further restrictions beyond this twist hypothesis allowed ma , trudinger , wang , and later loeper , to develop a regularity theory for the map @xmath241 , embracing delanoe , caffarelli and urbas results for the quadratic cost , gangbo and mccann s for its restriction to to convex surfaces , and wang s the reflector antenna design , which involves the restriction of @xmath245 to the sphere ; references may be found in @xcite @xcite . unfortunately , the twist hypothesis , also known as a generalized spence - mirrlees condition in the economic literature , can not be satisfied for smooth costs @xmath238 on compact manifolds @xmath25 , and apart from the result we are about to discuss there are no general theorems which guarantee uniqueness of minimizer to ( [ mkpa ] ) in this setting . with this in mind , let us state our main theorem , a version of which was established in a more complicated economic setting by chiappori , nesheim , and mccann @xcite . we expect the simpler formulation and argument given below to prove more interesting and accessible to a mathematical readership . + let @xmath20 and @xmath21 be complete separable manifolds equipped with borel probability measures @xmath22 on @xmath20 and @xmath23 on @xmath21 . let @xmath45 be a bounded cost function such that for each @xmath240 , the map @xmath246 has no critical points , save at most one global minimum and at most one global maximum . assume @xmath236 is locally bounded in @xmath235 , uniformly in @xmath21 . if @xmath22 is absolutely continuous with respect to coordinate measure on @xmath20 , then the minimum ( [ mkpa ] ) is uniquely attained ; moreover , the minimizer @xmath122 vanishes outside a numbered limb system having at most two limbs . here we give only the proof that there is a numbered limb system having at most two limbs , outside of which the mass of all minimizers @xmath0 vanishes . a detailed argument confirming the plausible fact that the graphs of these limbs are borel subsets of @xmath25 can be found in @xcite . uniqueness of @xmath0 then follows from theorem [ hestirwilliams ] . by linear programming duality due to kantorovich and koopmans in this context , it is well - known @xcite that there exist potentials @xmath247 and @xmath248 with @xmath249 such that @xmath250 from ( [ c - transform ] ) we see @xmath251 while ( [ duality ] ) implies any minimizer @xmath122 vanishes outside the zero set @xmath252 of the non - negative function appearing in ( [ zero set ] ) . it remains to show this set @xmath253 is contained in a numbered limb system consisting of at most two limbs ( apart from a @xmath254 negligible set ) . from ( [ c - transform ] ) , @xmath255 is locally lipschitz , since @xmath236 is controlled locally in @xmath235 , independently of @xmath256 . rademacher s theorem therefore combines with absolute continuity of @xmath22 to imply @xmath255 is differentiable @xmath22-almost everywhere ; we can safely ignore any points in @xmath20 where differentiability of @xmath255 fails , since they constitute a set of zero volume : @xmath257 = \mu[{\mathop{\rm dom}}dq]=1 $ ] . taking @xmath258 , suppose @xmath259 and @xmath260 both lie in @xmath253 , hence saturate the inequality ( [ zero set ] ) . then @xmath261 . in case the cost is twisted , meaning ( [ subtwist ] ) has no critical points , we conclude @xmath262 hence @xmath263 is contained in a graph . this completes the proofs by gangbo , carlier , and ma - trudinger - wang , of existence ( and uniqueness ) of a solution @xmath264 to monge s problem , pairing almost every @xmath265 with a single @xmath266 . notice uniqueness follows from lemma [ pure implies unique ] without further measurability assumptions . in the present setting , however , we only know that @xmath267 must be a global minimum or global maximum of the function ( [ subtwist ] ) . exchanging @xmath268 with @xmath269 if necessary yields @xmath270 for all @xmath271 , the second inequality being strict unless @xmath272 , in which case both inequalities are saturated . strictness of inequality ( [ keep away ] ) implies @xmath273 unless @xmath272 . in other words , @xmath274 lies on the antigraph of a function @xmath275 well - defined at @xmath269 . there may or may not be a point @xmath276 different from @xmath268 such that @xmath277 for all @xmath271 . if such a point @xmath278 exists , then @xmath279 as above . if no such @xmath278 exists , setting @xmath280 yields @xmath281 . since the range of @xmath282 is disjoint from the domain of @xmath283 , this completes the proof that up to @xmath0-negligible sets @xmath253 lies in a numbered limb system with at most two limbs , as desired . let us conclude by recalling an example of an extremal doubly stochastic measure which does not lie on the graph of a single map , drawn from work of gangbo and mccann @xcite and ahmad @xcite on optimal transportation , and developed in an economic context by chiappori , mccann , and nesheim @xcite . other examples may be found in the work of seethoff and shiflett @xcite , losert @xcite , hestir and williams @xcite , gangbo and mccann @xcite , uckelmann @xcite , mccann @xcite , and plakhov @xcite . imagine the periodic interval @xmath284 to parameterize a town built on the boundary of a circular lake , and let probability measures @xmath22 and @xmath23 represent the distribution of students and available places in schools , respectively . suppose the distribution of students is smooth and non - vanishing but peaks sharply at the northern end of the lake , and the distribution of schools is smooth and non - vanishing but peaks sharply at the southern end of the lake . if the cost of transporting a student residing at location @xmath285 $ ] to school at location @xmath286 $ ] is presumed to be given in terms of the angle commuted by @xmath287 , the most effective pairing of students with places in schools is given by the measure in @xmath24 which attains the minimum : @xmath288 according to results of gangbo and mccann @xcite , this minimizer is unique , and its support is contained in the union of the graphs of two maps @xmath289 . a schematic illustration is given in figure [ fig.extreme_doubly_stochastic ] , where the restriction of the support to the subsets marked by @xmath290 on the flat torus @xmath25 represent @xmath291 and @xmath292 respectively . the dotted lines mark @xmath293 . the necessary positivity of @xmath294>0 $ ] in this picture may be explained by observing that although it is cost - effective for all students to attend a school where they live , this is incompatible with the concentration of students at the north end of the lake , and of schools at the south end . once this imbalance is corrected by sending a sufficient number of northern students to southern schools by the map @xmath295 , the remaining students can be assigned to school near their home using the map @xmath296 . periodicity of graphs on the flat torus can be used to represent the support as a numbered limb system in more than one way ; see figure [ fig.example_numbered_limb ] , which exploits the fact that the support of @xmath0 in figure [ fig.extreme_doubly_stochastic ] intersects @xmath297 in a graph and @xmath298 in an anti - graph . chiappori , nesheim and mccann @xcite called the uniqueness hypothesis limiting the number of critical points to at most one maximum and and at most one minimum in ( [ subtwist ] ) the _ subtwist _ condition . although it is satisfied in the example above , it is an unfortunate fact that the subtwist condition can not be satisfied by any smooth function @xmath299 on a product of manifolds @xmath25 with more complicated morse structures than the sphere . it is an interesting open problem to find a criterion on a smooth cost @xmath299 on @xmath300 which guarantees uniqueness of the minimum ( [ mkp ] ) for all smooth densities @xmath22 and @xmath23 on the torus . although we expect such costs to be generic , not a single example of such a cost is known to us . hestir and williams criteria for extremality seems likely to remain relevant to such questions , and it is natural to conjecture that the complexity of the morse structure of the manifold @xmath20 plays a role in determining the required number of limbs in the system . v. bene and j. tpn . the support of extremal probability measures with given marginals . in m.l . puri , p. rvsz and w. wertz , editors , _ mathematical statistics and probability theory _ , volume a. theoretical aspects , pages 3341 . d. reidel publishing co. , dordrecht , 1987 . allocation maps with general cost functions . in _ partial differential equations and applications _ , p. marcellini et al , eds . new york : dekker , lecture notes in pure and appl * 177 * ( 1996 ) 2935 . l. kantorovich . on the translocation of masses ( 1942 ) 199201 . l. kantorovich . on a problem of monge ( in russian ) ( 1948 ) 225 - 226 . y .- h . kim and r.j . continuity , curvature , and the general covariance of optimal transportation . to appear in _ ( jems)_. l. uckelmann . optimal couplings between onedimensional distributions . in v. bene and j. tpn , editors , _ distributions with given marginals and moment problems _ , pages 261273 . kluwer academic publishers , dordrecht , 1997 . j. von neumann . a certain zero - sum two - person game equivalent to the optimal assignment problem . in _ contributions to the theory of games _ , volume 2 , pages 512 . princeton university press , princeton nj , 1953 .	this article connects the theory of extremal doubly stochastic measures to the geometry and topology of optimal transportation . we begin by reviewing an old question ( # 111 ) of birkhoff in probability and statistics @xcite , which is to give a necessary and sufficient condition on the support of a joint probability to guarantee extremality among all measures which share its marginals . following work of douglas , lindenstrauss , and bene and tpn , hestir and williams @xcite found a necessary condition which is nearly sufficient ; we relax their subtle measurability hypotheses separating necessity from sufficiency slightly , yet demonstrate by example that to be sufficient certainly requires some measurability . their condition amounts to the vanishing of @xmath0 outside a countable alternating sequence of graphs and antigraphs in which no two graphs ( or two antigraphs ) have domains that overlap , and where the domain of each graph / antigraph in the sequence contains the range of the succeeding antigraph ( respectively , graph ) . such sequences are called _ numbered limb systems_. surprisingly , this characterization can be used to resolve the uniqueness question for optimal transportation on manifolds with the topology of the sphere .
the present paper is one in a recent series , devoted to space - time physics , by the present author . in ref . @xcite the classical ` rockets - and - string ' @xcite and ` pole - and - barn ' @xcite paradoxes of special relativity were re - analysed taking into account the distinction between the real and apparent positions of uniformly moving objects . different results were obtained from those of the usual text - book interpretations of these experiments and a new causality - violating paradox was pointed out . this paradox , as well as the related ` backwards running clocks ' one of soni @xcite , was resolved in ref . @xcite where , in order to avoid these paradoxes as well as manifest breakdown of translational invariance , in some applications of the standard space - time lorentz transformation , the use of a ` local ' lorentz transformation . i.e. one where the transformed event in the moving frame lies at the coordinate origin in this frame , was advocated . when this is done the closely correlated ` relativity of simultaneity ' ( rs ) and ` length contraction ' ( lc ) effects of conventional special relativity theory do not occur . the connection between these effects is explained in ref . @xcite . ref . @xcite also contains a ` mini review ' of all experimental tests of special relativity where it is pointed out that , whereas time dilatation is well - confirmed experimentally , no experimental evidence exists for the rs and lc effects . in the later papers @xcite it is explained how the spurious rs and lc effects result from a misuse of the time symbols in the standard space - time lorentz transformation . refs . @xcite present the argument in a pedagogical manner , whereas ref . @xcite contains a concise summary of it . in the following section the necessary formulae for the analysis of the muon decay thought experiment essentially the prediction of a universal time dilatation effect are derived from first principles . here there is considerable overlap with work presented in refs . @xcite the analysis of the thought experiment presented in section 3 shows the absence of the spurious text - book rs effect : muons which decay simultaneously in a common proper frame , are observed to do so in all inertial frames . finally the results of the analysis presented in this paper are used to shed light on the physical basis of the differential aging effect in the travelling - twin thought experiment @xcite . the lorentz transformation ( lt ) relates observations ( @xmath0,@xmath1,@xmath2,@xmath3 ) of the coordinates of space - time events in one inertial frame s , to observations of the coordinates ( @xmath4,@xmath5,@xmath6,@xmath7 ) of the same events in another inertial frame s. as is conventional , the cartesian spatial coordinate axes of the two frames are parallel , and the origin of the frame s moves with constant speed , @xmath8 , along the @xmath0-axis . in any actual experiment , times are recorded by clocks , and positions specified by marks on fixed rulers ( or their equivalent ) . therefore , in order to relate the space - time coordinates appearing in the lt to actual physical measurements they must be identified with clock readings and length interval measurements @xcite . this can be done in two distinct ways depending on whether the experimenter observing the clocks and performing the length measurements is at rest in s or in s. in the former case ( only events with spatial coordinates along the @xmath0-axis are considered so that @xmath1 = @xmath5 = @xmath2 = @xmath6 = @xmath9 ) the appropriate lt , for a clock , c , situated at the origin of s , is : @xmath10 = 0 \\ t ' & = & \gamma_v[\tau-\frac{\beta_v x({{\rm c}}')}{c}]\end{aligned}\ ] ] and in the latter case , for a clock , c , situated at the origin of s , is : @xmath11 = 0 \\ t & = & \gamma_v[\tau'+\frac{\beta_v x'({{\rm c}})}{c } ] \end{aligned}\ ] ] in these equations @xmath12 , @xmath13 and @xmath14 is the speed of light in vacuum . the clocks in s and s are synchronised so that for ( 2.1 ) and ( 2.2 ) , @xmath15 when @xmath16 , and for ( 2.3 ) and ( 2.4 ) , @xmath17 when @xmath16 . in ( 2.1 ) and ( 2.2 ) the transformed events lie on the worldline of a clock , c , at rest in s , which is observed from s. the observed time in s registered by c ( which is in motion in this frame ) is @xmath7 , while @xmath18 is the time registered by the clock , c , identical to c , but at rest in s. in contrast , in ( 2.3 ) and ( 2.4 ) the transformed events lie on the worldline of c , which is observed from s. the time @xmath3 is that registered by c as observed from s and @xmath19 is the time registered by c as observed in its own proper frame . thus two distinct experiments are possible involving one stationary and one moving clock , depending on whether the experimenter performing the space and time measurements is in the rest frame of one , or the other , of the two clocks . to describe both of these experiments , four different time symbols , @xmath18 , @xmath19 , @xmath3 and @xmath7 , with different operational meanings , are required . from ( 2.1 ) , the equation of motion of c in s is : @xmath20 while from ( 2.3 ) the equation of motion of c in s is : @xmath21 using ( 2.5 ) to eliminate @xmath0 from ( 2.2 ) , and in view of the definition of @xmath22 : @xmath23 similarly , using ( 2.6 ) to eliminate @xmath4 from ( 2.4 ) gives : @xmath24 ( 2.7 ) and ( 2.8 ) are expressions of the relativistic time dilatation ( td ) effect in the two ` reciprocal ' experiments that may be performed using the clocks c and c. they show that , according , to the lt , ` moving clocks run slow ' in a universal manner ( no spatial coordinates appear in ( 2.7 ) and ( 2.8 ) ) . in fact : @xmath25 to discuss measurements of the spatial separations of moving clocks , at least two clocks , ( say , @xmath26 and @xmath27 , at rest in s ) must be considered . it is assumed that they lie along the @xmath4-axis separated by the distance @xmath28 , @xmath26 being at the origin of s and @xmath27 at @xmath29 . the space transformation equations analogous to ( 2.1 ) for @xmath26 and @xmath27 and are then : @xmath30 = 0 \\ x'({{\rm c}'_{b}})-l ' & = & \gamma_v[x({{\rm c}'_{b}})-l - v\tau ] = 0 \end{aligned}\ ] ] inspection of ( 2.11 ) shows that @xmath31 , a relation valid for all values of @xmath8 for the choice of coordinate systems in ( 2.10 ) and ( 2.11 ) . in particular , it is valid when @xmath32 , @xmath33 , and @xmath34 . then for @xmath35 : @xmath36 so that @xmath37 the spatial separation of the clocks is therefore a lorentz - invariant quantity . suppose now that two objects move with speeds @xmath38 , @xmath39 ( @xmath40 ) along the positive @xmath0-axis in s and that they are coincident with the origins of s and s at the time @xmath41 . at later times @xmath18 , @xmath7 , the separation of the objects in s is @xmath42 and in s is @xmath43 . where @xmath44 is the relative velocity of the objects in s. in view of ( 2.13 ) and the time dilatation formula ( 2.7 ) it follows that @xmath45 a particular case of ( 2.14 ) , to be used in the following section , is @xmath46 , @xmath47 giving @xmath48 where @xmath49 is the observed speed of the origin of s along the negative @xmath0-axis in s . the relation ( 2.14 ) is the transformation formula of the _ relative velocity _ of two objects between two inertial frames , to be contrasted with the relativistic parallel velocity addition formula : @xmath50 which relates kinematical configurations of a single moving object in the frames s and s. in the case @xmath51 , ( 2.16 ) gives @xmath52 and this equation relates the kinematical configuration in s in the primary experiment described by ( 2.1 ) and ( 2.2 ) to that in s in the ( physically independent ) reciprocal experiment described by ( 2.3 ) and ( 2.4 ) . in contrast ( 2.14 ) describes the observed _ relative velocity transformation _ within the primary experiment . for further discussion of this important point see refs . muon decays constitute an excellent laboratory for testing the predictions of special relativity . for example , the td effect of eqn(2.7 ) was experimentally confirmed at the per mille level of relative precision in the ultrarelativisic domain ( @xmath53 ) by observation of the decay of muons in the storage ring of the last cern muon @xmath54 experiment @xcite . in the present paper , it is shown that thought experiments involving muons provide a graphic illustration of the predicted space - time behaviour , in special relativity , of clocks in different inertial frames . unlike most other unstable particles , muons are particularly suitable for precise tests of the td effect because of the ease of their production from pion decay and their long mean lifetime of 2.2 @xmath55s . the former yields high events statistics and the latter the possibility of precise time interval measurements using accurate clocks in the laboratory frame @xcite . the thought experiment developed in the present paper is an elaboration of the well - known demonstration that the very presence of cosmic muons at the earth s surface is , by itself , sufficient to demonstrate the existence of the td effect @xcite . muons are produced predominantly by the weak decay of charged pions @xmath56 . the velocity of the muon , @xmath57 , depends upon that of the parent pion , @xmath58 , and the centre - of - mass decay angle , @xmath59 . if the pion has the same velocity , @xmath60 , as the muon in the pion rest frame , ( corresponding to a pion momentum of 49.5 mev / c ) and @xmath61 , the muon is produced at rest in the laboratory system . the maximum muon decay energy @xmath62 correponds to @xmath63 and is given , in terms of the parent pion energy @xmath64 , and the pion velocity @xmath65 , by the relation : @xmath66}{2 m_{\pi}^2}\ ] ] due to the thickness of the earth s atmosphere , the majority of interactions of primary cosmic protons , that produce the parent pions of cosmic muons , occur at high altitude , @xmath69 20 km above the earth s surface . a muon with speed close to that of light then takes at least @xmath69 700 @xmath70 to reach the surface of the earth . this may be compared with the muon mean lifetime of 2.2 @xmath70 . without the td effect , only a fraction @xmath71 \simeq 10^{-138}$ ] of the muons produced at altitude would reach the earth s surface . however a 10 gev muon , which has @xmath72 , has a 3.5 @xmath73 probability to reach the earth s surface , before decaying , when the td effect is taken into account . in the thought experiment considered here it is assumed that two muons @xmath74 and @xmath75 are produced simultaneously at the same point a ( see fig.1a ) by decay of pions from a primary cosmic ray interaction with the nucleus of a gas atom in the atmosphere . the muon @xmath74 is produced at rest in the atmosphere ( inertial frame s ) while @xmath75 ( with proper frame s ) is produced with velocity @xmath76 , so that @xmath77 . it happens that both muons decay after time @xmath78 in their proper frames . because of the td effect , the muon @xmath75 will then be seen by an observer at rest in the atmosphere to decay after time @xmath79 at a point b at a distance @xmath80 km from a. it is also supposed that at the same time , @xmath41 , that @xmath74 and @xmath75 are created , another muon , @xmath81 , ( also with proper decay lifetime @xmath78 ) is created at rest in the atmosphere at the point b , by decay of pion from another primary cosmic ray interaction . since @xmath74 and @xmath81 are at rest in the atmosphere and have no td effect , they will decay simultaneously at @xmath82 ( fig.1b ) in the frame s. at this instant the muon @xmath75 is still undecayed and is at the point m , midway between a and b , when @xmath75 decays ( fig.1c ) @xmath74 and @xmath81 no longer exist , however the centres of mass of their , by now distant , decay products @xmath83 , @xmath84 and @xmath85 , denoted as ( @xmath74 ) and ( @xmath81 ) , and indicated in figs . 1 - 3 as two concentric circles , still remain at the points a and b. the sequence of events that would be seen by an observer in the rest frame , s, of @xmath75 corresponding to those of fig.1 , is shown in fig.2 . according to the relative velocity transformation formula ( 2.15 ) , @xmath74 and @xmath81 are observed to move to the left with velocity @xmath86 . the configuration at @xmath87 is shown in fig.2a . according to the time dilatation relation ( 2.7 ) , appropriate to this case , @xmath74 and @xmath81 are seen to decay simultaneously at time @xmath88 ( fig.2b ) . as in fig.1 , it can be sees that @xmath75 is aligned with m , the mid point of the line segment ab , at the time of simultaneous decay of @xmath74 and as shown in fig.2c , @xmath75 decays at time @xmath89 , when it is aligned with b , as is also the case for an observer in the frame s , as shown in fig.1c . since @xmath74 and @xmath81 are seen to decay simultaneously by observers at rest in both s and s , there is no rs effect . the sequence of events that would be observed in the frame s , moving with velocity @xmath90 in the positive @xmath0-direction will now be considered . according to the relative velocity transformation formula ( 2.14 ) , @xmath75 is observed to move with speed @xmath91 in the positive @xmath0-direction in s since the _ relative velocity _ of s and s in the frame s is @xmath92 . also , according to ( 2.15 ) , @xmath74 and @xmath81 are observed to move with speed @xmath93 in the negative @xmath0-direction in s . the velocity of @xmath75 relative to @xmath74 and @xmath81 in the frame s is then @xmath94 . the sequence of events seem by an observer at rest in s is illustrated , for the case @xmath95 , in fig.3 . using the time dilatation realation ( 2.7 ) , with @xmath8 set equal to @xmath90 in order to relate times in s and s , @xmath74 and @xmath81 decay at time @xmath96 , when @xmath75 is aligned with m ( see fig.3b ) . note that the time of this event in s depends only on the value of @xmath90 , being independent of the value of @xmath8 . the muon @xmath75 decays at time @xmath97 when it is aligned with b ( see fig.3c ) . in all three frames , s , s and s , @xmath74 and @xmath81 are observed to decay simultaneously and earlier than @xmath75 . these decay times , in each frame , are presented in table 1 . the entries in this table satisfy the following condition : @xmath98 since the last member of this equation is independent of @xmath90 , it follows that ratio of the decay times given by the time dilatation relation ( 2.7 ) is the same for all inertial observers , and so is an invariant , fixed by the velocity of @xmath75 in the rest frame of @xmath74 where the time dilatation effect is defined i.e . observer at rest in s , observed muon at rest in s. different ( but reciprocal , i.e. related to those of eqn(3.2 ) by exchange of primed and unprimed quantities ) results would be obtained in the situation where the observer of the time dilatation effect is at rest in s , while the observed muon is a rest in s , so that the reciprocal time dilatation relation ( 2.8 ) is applicable . inspection of , and reflection on , figs.1 and 2 reveals the physical basis of the differential aging effect in the ` twin paradox ' introduced by langevin @xcite . the ` travelling twin ' can be indentified with @xmath75 , the ` stay at home ' one with either @xmath74 or @xmath81 since their associated ` clocks ' are synchronous . at the end of the outward journey , when @xmath75 arrives at b , the synchronous clocks at a and b were seen in s to run twice as fast as that of @xmath75 in fact @xmath74 and @xmath81 have already decayed when @xmath75 arrives at b. in fig.1 , the observer in s sees @xmath75 aging less rapidly than @xmath74 or @xmath81 . however , as required by the time dilatation relation ( 2.7 ) , an observer in s ( fig.2 ) sees @xmath74 or @xmath81 _ aging more rapidly _ , by a factor two , than @xmath75 . this is true even though these clocks are in motion relative to the observer on s. fig.2 also reveals that the physical basis of time dilatation , and differential aging , is not , as hitherto , to be found in ` length contraction ' in the frame s , but instead in the greater velocity of @xmath74 or @xmath81 relative to @xmath75 in the ` travelling frame ' s , than that of @xmath75 relative to @xmath74 or @xmath81 in the ` base frame ' s from which the time dilatation effect is observed . for futher discussion of base and travelling frames in relation to primary and reciprocal space - time experiments see refs . @xcite . the much - discussed ` twin paradox ' arises when it is attempted to describe the sequence of events shown in fig.2 by use of the time dilatation relation ( 2.8 ) of the reciprocal experiment which requires clocks in s to run slower ( not faster , as in fig.2 ) than those in s , when observed in the latter frame . this is a nonsensical interpretation since the time variables @xmath18 and @xmath7 appearing in the time dilatation relation ( 2.7 ) have a completely different operational meaning to those , @xmath19 and @xmath3 in the time dilatation relation of the reciprocal experiment . see ref . @xcite for a more detailed discussion of the standard and incorrect interpretation of the twin paradox based on the spurious lc effect , in contrast with its correct interpretation following from the relative velocity transformation formula ( 2.14 ) . 99 j.h.field , ` on the real and apparent positions of moving objects in special relativity : the rockets - and - string and pole - and - barn paradoxes revisited and a new paradox',arxiv pre - print : http://xxx.lanl.gov/abs/physics/0403094v3 . cited 30 nov 2007 . e.dewan and m.beran , am j. phys.27 517 ( 1959 ) . e.dewan , am j. phys . * 31 * 383 ( 1963 ) . v.s.soni , eur . * 23 * 225 ( 2002 ) . j.h.field , ` the local space - time lorentz transformation : a new formulation of special relativity compatible with translational invariance ' , arxiv pre - print : http://xxx.lanl.gov/abs/ physics/0501043v3 . cited 30 nov 2007 . j.h.field , ` spatially - separated synchronised clocks in the same inertial frame : time dilatation , but no relativity of simultaneity or length contraction ' . arxiv pre - print : http://xxx.lanl.gov/abs/0802.3298v2 . cited 4 mar 2008 . j.h.field , clock rates , clock settings and the physics of the space - time lorentz transformation , arxiv pre - print : http://xxx.lanl.gov/abs/physics/0606101v4 . cited 4 dec 2007 . j.h.field , translational invariance and the space - time lorentz transformation with arbitary spatial coordinates , http://xxx.lanl.gov/abs/physics/0703185v2 . cited 15 feb 2008 . p.langevin , scientia , * 10 * 31 ( 1911 ) . j.h.field , ` the physics of space and time i : the description of rulers and clocks in uniform translational motion by galilean or lorentz transformations ' , arxiv pre - print : http://xxx.lanl.gov/abs/physics/0612039v3 . cited 28 mar 2008 . j.h.field , ` the physics of space and time iii : classification of space - time experiments and the twin paradox ' . arxiv pre - print : http://xxx.lanl.gov/abs/0806.3671v1 . cited 23 jun 2008 . j.h.field , primary and reciprocal space - time experiments , relativistic reciprocity relations and einstein s train - embankment thought experiment . arxiv pre - print : http://xxx.lanl.gov/abs/0807.0158v1 . cited 1 jul 2008 . et al _ nature * 268 * 301 ( 1979 ) . r.p.feynman , r.leighton and m.sands , the feynman lectures in physics , volume i ( addison - wesley , reading massachusetts , 1966 ) section 15 - 3 . e.f.taylor and j.a.wheeler , ` spacetime physics ' , w.h.freeman and company , san franciso 1966 , section 42 , p89 . p.a.tipler and r.a.lewellyn , ` modern physics ' w.h.freeman and company , new york , ch 1 , p40 . a.walker , cern courier , * 46 * number 4 may 2006 , p23 .	observation of the decay of muons produced in the earth s atmosphere by cosmic ray interactions provides a graphic illustration of the counter - intuitive space - time predictions of special relativity theory . muons at rest in the atmosphere , decaying simultaneously , are subject to a universal time - dilatation effect when viewed from a moving frame and so are also observed to decay simultaneously in all such frames . the analysis of this example reveals the underlying physics of the differential aging effect in langevin s travelling - twin thought experiment . 24.5 cm -5pt -5pt -50pt addtoresetequationsection * j.h.field * dpartement de physique nuclaire et corpusculaire universit de genve . 24 , quai ernest - ansermet ch-1211 genve 4 . e - mail : [email protected]
in this paper , the symmetric obrechkoff methods for solving special class of initial value problems associated with second order ordinary differential equations of the type @xmath0 in which the first order derivatives do not occur explicitly , are discussed.the numerical integration methods for ( [ dif ] ) can be divided into two distinct classes : 1 . problems for which the solution period is known ( even approximately ) in advance . problems for which the period is not known . for several decades , there has been strong interest in searching for better numerical methods to integrate first - order and second - order initial value problems , because these problems are usually encountered in celestial mechanics , quantum mechanical scattering theory , theoretical physics and chemistry , and electronics . generally , the solution of @xmath1 is periodic , so it is expected that the result produced by some numerical methods preserves the analogical periodicity of the analytic solution [ 9 - 22 ] . computational methods involving a parameter proposed by gautschi @xcite , jain et al . @xcite , sommeijer and et al @xcite and steifel and bettis @xcite yield numerical solution of problems of class ( 1 ) . chawla and et al @xcite , ananthakrishnaiah @xcite , shokri and et al . @xcite , dahlquist @xcite , franco @xcite , lambert and watson [ 9 ] , tsitouras and simos @xcite , simos and et al . @xcite , hairer @xcite , wang et al . @xcite , saldanha and achar @xcite , and daele and vanden berghe @xcite have developed methods to solve problems of class ( 2 ) . consider obrechkoff method of the form @xmath2 for the numerical integration of the problem ( [ dif ] ) . the method ( [ ob ] ) is symmetric when @xmath3 , @xmath4 , @xmath5 , and it is of order @xmath6 if the truncation error associated with the linear difference operator is given as @xmath7 where @xmath8 is a constant dependent on @xmath9 . when the method ( [ ob ] ) is applied to the test problem , we get the characteristic equation as @xmath10 where @xmath11 and @xmath12 the method ( [ ob ] ) is said to have interval of periodicity @xmath13 if for all @xmath14 the roots of eq . ( [ ro ] ) are complex and at least two of them lie on the unit circle and the others lie inside the unit circle . the method ( [ ob ] ) is said to be p - stable if its interval of periodicity is @xmath15 . for any symmetric multistep methods , the phase - lag ( frequency distortion ) of order @xmath16 is given by @xmath17 where @xmath18 is the phase lag constant and @xmath16 is the phase - lag order . the characteristic equation of the method ( [ ob ] ) is given by @xmath19 where @xmath20 @xmath21 contains polynomial functions together with trigonometric polynomials @xmath22 the resulting methods are then based on a hybrid set of polynomials and trigonometric functions . if @xmath23 is limited to @xmath24 , we called method with zero phase - lag . we present here the trigonometric versions of the set . in case @xmath25 is purely imaginary one obtains the hyperbolic description of this set . this set is characterized by two integer parameters @xmath26 and @xmath23 . the set in which there is no polynomial part is identified by @xmath27 while the set in which there is no trigonometric polynomial component is identified by @xmath27 . for each problem one has @xmath28 , where @xmath29 is the maximum exponent present in the full polynomial basis for the same problem . from the form and without loss of generality we assume @xmath30 , @xmath31 , @xmath32 and we can write @xmath33,\ ] ] when @xmath34 we get @xmath35\nonumber\\ & + & h^4\left[\beta_{20}(y_{n+1}^{(4)}+y_{n-1}^{(4)})+\beta _ { 21}y_n^{(4)}\right]\nonumber\\ & + & h^6\left[\beta_{30}(y_{n+1}^{(6)}+y_{n-1}^{(6)})+\beta _ { 31}y_n^{(6)}\right].\end{aligned}\ ] ] @xmath36 for method ( [ ob3 ] ) is 11 so that if @xmath37 , @xmath38 we obtain classic method and the coefficients of this method are @xmath39 where its phase - lag is given by @xmath40 and its local truncation error is given by @xmath41 if @xmath42 , @xmath43 we obtain the method with zero phase - lag ( pl ) , and the coefficients of this case are given in @xcite . if @xmath44 , @xmath45 , so we called pl@xmath46 , we have @xmath47 @xmath48 where @xmath49 and @xmath50 for small values of @xmath51 the above formulae are subject to heavy cancelations . in this case the following taylor series expansion must be used : @xmath52 @xmath53 @xmath54 @xmath55 @xmath56 @xmath57 the phase - lag and the local truncation error for the pl@xmath46 method are given by @xmath58 hence @xmath59 and @xmath60 where @xmath61 , @xmath25 is the frequency and @xmath9 is the step length . as @xmath62 , the lte of the method with derived coefficients tends to @xmath63 . which agrees with the lte of the three methods due to wang @xcite , simos @xcite and daele @xcite , achar @xcite , as @xmath64 . the behavior of the coefficients of the pl@xmath46 method are shown in figures 2.1 , to 2.6 . if @xmath65 , @xmath66 , so we called pl@xmath67 , we have @xmath68 @xmath69 where @xmath70 and @xmath71 for small values of @xmath51 the above formulae are subject to heavy cancelations . in this case the following taylor series expansion must be used : + @xmath72 @xmath73 @xmath74 @xmath75 @xmath76 @xmath77 the phase - lag and the local truncation error for the pl@xmath67 method are given by @xmath78 and @xmath79 where @xmath61 , @xmath25 is the frequency and @xmath9 is the step length . the behavior of the coefficients of the pl@xmath67 method are shown in figures 4 , 5 , 6 . the characteristic equation @xmath80 has complex roots of unit magnitude when @xmath81 , i.e. when @xmath82 . substituting for @xmath83 and @xmath84 for these the two - step methods , the interval of periodicity of the classical obrechkoff method , pl@xmath46 and pl@xmath67 methods when @xmath62 are obtained [ 0 , 25.2004 ] , [ 0,408.04 ] and [ 0,1428.84 ] respectively in this section , we present some numerical results obtained by our new two - step trigonometrically - fitted obrechkoff methods and compare them with those from other multistep methods as + achar : the 12th order obrechkoff method of achar @xcite . + daele : the 12th order obrechkoff method of van daele @xcite . + neta : the p - stable 8th - order super - implicit method of neta @xcite . + simos : the 12th order obrechkoff method of simos @xcite . + wang : the 12th order obrechkoff method of wang @xcite . + we consider the nonlinear _ undamped duffing equation _ @xmath85 where @xmath86 , @xmath87 and @xmath88 $ ] . we use the following exact solution for from @xcite , @xmath89 where @xmath90 in order to integrate this equation by a obrechkoff method , one needs the values of @xmath91 , which occur in calculating @xmath92 . these higher order derivatives can all be expressed in terms of @xmath93 and @xmath94 through , i.e. @xmath95 the absolute errors at @xmath96 , for the new method , in comparison with methods of simos , daele , achar , wang and the new method are given in table 3.1 and the cpu times are listed in table 3.2 . also the absolute errors at @xmath97 , with @xmath98 , for the new method pl@xmath67 , in comparison with methods neta and the new method are given in table 3.3 . @xmath99 @xmath100 @xmath101 consider the initial value problem @xmath102 with the exact solution @xmath103 . this equation has been solved numerically for @xmath104 using exact starting values . in the numerical experiment , we take the step lengths @xmath105 , @xmath106 , @xmath107 , @xmath108 , @xmath109 and @xmath110 . in table 3.4 , we present the absolute errors at the end - point and the cpu times are listed in table 3.5 . @xmath111 @xmath112 consider the initial value problem @xmath113,\ ] ] with the exact solution the theoretical solution of this problem is @xmath114 the absolute errors at @xmath115 for the new method , in comparison with methods of wang , simos , daele and achar are given in the table 3.6 . the relative cpu times of computation of the new method in comparison with the other four referred methods are given in table 3.7 . @xmath116 @xmath117 in this paper , we have presented the new trigonometrically - fitted two - step symmetric obrechkoff methods of order 12 . the details of the procedure adapted for the applications have been given in section 2 . with trigonometric fitting , we have improved the local truncation error , phase - lag error , interval of periodicity and cpu time for the classes of two - step obrechkoff methods . the numerical results obtained by the new method for some problems show its superiority in efficiency , accuracy and stability . the authors wish to thank the professor theodore e. simos and the anonymous referees for their careful reading of the manuscript and their fruitful comments and suggestions . chawla , m. m. , rao , p. s. , a numerov - type method with minimal phase - lag for the integration of second order periodic initial value problems . ii : explicit method , j. comput . appl . ( 1986 ) , 329 - 337 . neta , b. , and fukushima , t. , obrechkoff versus super - implicit methods for the integration of keplerian orbits , proc . aiaa / aas astrodynamics specialist conference , denver , co , august 14 - 17 , 2000 , paper number aiaa 2000 - 4029 . neta , b. , and fukushima , t. , obrechkoff versus super - implicit methods for the solution of first and second order initial value problems , j. comput . applic . , special issue on numerical methods in physics , chemistry and engineering , t. e. simos and g. abdelas ( guest editors ) , 45 , ( 2003 ) , 383 - 390 . psihoyios , g. , simos , t. e. , effective numerical approximation of schrodinger type equations through multiderivative exponentially - fitted schemes , appl . comput . math . 1 ( 1 ) ( 2004 ) , 205 - 215 . sakas , d. p. , simos , t. e. , multiderivative methods of eighth algebraic order with minimal phase - lag for the numerical solution of the radial schrodinger equation , j. comput appl . ( 2005 ) , 161 - 172 . saldanha , g. , achar , s. d. , symmetric multistep obrechkoff methods with zero phase - lag for periodic initial value problems of second order differential equations , appl . comput . 218 ( 2011 ) , 2237 - 2248 . shokri , a. , rahimi ardabili , m. y. , shahmorad , s. , and hojjati , g. , a new two - step p - stable hybrid obrechkoff method for the numerical integration of second - order ivps . , j. comput . math . , 235 ( 2011 ) , 1706 - 1712 . wang , z. , zhao , d. , dai , y. and song , x. , a new high efficient and high accurate obrechkoff four - step method for the periodic non - linear undamped duffings equation , comput . phys . commun . 165 ( 2005 ) , 110 - 126 .	in this paper , we present the two - step trigonometrically fitted symmetric obrechkoff methods with algebraic order of twelve . the method is based on the symmetric two - step obrechkoff method , with 12 algebraic order , high phase - lag order and is constructed to solve ivps with periodic solutions such as orbital problems . we compare the new method to some recently constructed optimized methods from the literature . the numerical results obtained by the new method for some problems show its superiority in efficiency , accuracy and stability .
-0.6 cm -.2 cm -space ( c ) cross cuts with a time step of 0.5 ps . the lut colour scale in the case of the amplitude chart in ( a ) is relative to the initial top amplitude , and represents an enhancement factor of 4 at around 10 ps after the pulse arrival , which corresponds to a factor of 16 in density . ] fwhm gaussian spot and linear polarization . each row is relative to a different initial density . the first column represents the time - integrated images of the bare emission in space directly aquired on the camera . the second and third columns represent the density and phase profiles , respectively , along a central diameter and versus time , retrieved by means of the ultrafast imaging . the powers are increasing from top to bottom rows and are relative to initial total populations of @xmath33@xmath34 = @xmath35 , @xmath36 , @xmath37 , @xmath38 , @xmath39 and @xmath40 polaritons . the corresponding initial top densities are : 345 @xmath8 , 790 @xmath8 , 1200@xmath8 , 1900@xmath8 , 3450@xmath8 and 7200@xmath41 , respectively . the maxima on the colour bars of the density charts are expressed in terms of the initial top densities and represent the achieved enhancement factors . from the blueshift associated to @xmath42 , we evaluated a nonlinearity @xmath43 to be renormalized to @xmath44 for a single qw . ] , @xmath45 and @xmath42 as a function of time . the blue curve plotted on the left gives the total number of polaritons with time while the black curve plotted on the right gives the corresponding density in the centre as a ratio with respect to the one at the time of the first peak . ] -2 cm @xmath21 emission of the microcavity polaritons is obtained after offresonant cw excitation at low power . the separation of the upb and lpb branches is 5.4 mev@xmath463 nm . ] and ( b ) @xmath47 in - plane wavevectors @xmath48 . the cuts of the spatial polariton density as a function of time are taken along the propagation direction . ] the model commonly used to describe the dynamics of polaritons is based on the mean - field approach . the time evolution of the photonic ( excitonic ) wavefunctions @xmath49 ( @xmath50 ) is given by coupled gross - pitaevskii equations @xcite : @xmath51where @xmath52 ( @xmath53 ) is the cavity photon ( exciton ) effective mass , @xmath54 is the exciton - exciton interaction constant , @xmath55 is the rabi frequency determining the exciton - photon coupling , @xmath56 is the pumping field describing the pulsed excitation of the cavity with a gaussian spot of diameter @xmath57 and pulse - duration of @xmath58 and @xmath59 is a parameter determining the density of photo - created carriers . while this model has been successful to describe much of the fluid dynamics of polaritons , including ballistic motion @xcite , superfluidity @xcite , solitons @xcite , etc . @xcite , in presence of repulsive interactions , @xmath60 , it can not account for real - space localization . many additional ingredients to these equations can produce qualitatively the most important feature of our experiment , the formation of a high density peak in the center . however , these extended models , based on different physical assumptions , have some implications , present other qualitative features and/or require certain ranges or parameters that rule them out as an explanation of the phenomenon . some hypotheses that work reasonably well demand assumptions that are difficult to justify , such as explicit attractive interactions , @xmath61 . we found that one model only is sufficiently consistent with all the observations and provides a reasonably close agreement with the data to be retained as a possible mechanism for our experiment . it is presented last in a series of alternative descriptions , below : the loss of strong coupling in section [ sec : viemay8085844cest2015 ] , a localisation due to the exciton reservoir or dark excitons in section [ sec : viemay8121915cest2015 ] , polaritons with attractive interactions in section [ sec : viemay8123718cest2015]all of these models failing in some fatal way to account for the experiment and , ultimately , the collective polaron effect , in section [ sec : viemay8130427cest2015 ] , which , on the contrary , reproduces adequately the findings , based on likely assumptions and parameters corresponding to our sample . the measured emission of total emitted photons from a single experiment suggests that the exciton density is close to the saturation value of @xmath62 . at such a high density , the reduction of the rabi splitting should occur due to phase space filling @xcite and lead to a modification of the polariton dispersion . at the first order of perturbation in the exciton density , the saturation is momentum - independent , or local in space ( precisely , the range of the interaction is of the order of the exciton radius ) @xcite . the linear regime rabi coupling @xmath63 is then renormalized to : @xmath64 where @xmath65 is the coefficient dependent on the quantum well geometry , @xmath54 the polariton - polariton interaction strength and @xmath66 the exciton wavefunction ( as already defined before ) . this leads to effective attractive interactions but only for the up branch , while the interactions remain repulsive for the lp branch @xcite . this is because the up branch is redshifted whilst the rabi coupling decreases , while the lp branch is blueshifted , just as it would be from repulsive interactions . a numerical simulation for this mechanism is shown in fig . [ fig : mikhail ] . this model correctly reproduces the experimental rabi bending and allows to describe a central density peak with both a femtosecond pulse and a picosecond pulse tuned to the up branch . critically , however , the ps pulse excitation tuned to the lp branch leads to strong defocusing in this model , in contradiction with the experimental data . , @xmath67 mev , @xmath68 , @xmath69 ps , @xmath70 ps , @xmath71 , @xmath72 mev @xmath73m@xmath74 , @xmath75 , @xmath76 mev @xmath73m@xmath74 , @xmath77 m , @xmath78 fs , @xmath79 , @xmath80 , @xmath77 m , @xmath78 fs ( fs - pulse case ) or @xmath81 ps ( ps - pulse case ) . , title="fig:",width=264 ] , @xmath67 mev , @xmath68 , @xmath69 ps , @xmath70 ps , @xmath71 , @xmath72 mev @xmath73m@xmath74 , @xmath75 , @xmath76 mev @xmath73m@xmath74 , @xmath77 m , @xmath78 fs , @xmath79 , @xmath80 , @xmath77 m , @xmath78 fs ( fs - pulse case ) or @xmath81 ps ( ps - pulse case ) . , title="fig:",width=264 ] , @xmath67 mev , @xmath68 , @xmath69 ps , @xmath70 ps , @xmath71 , @xmath72 mev @xmath73m@xmath74 , @xmath75 , @xmath76 mev @xmath73m@xmath74 , @xmath77 m , @xmath78 fs , @xmath79 , @xmath80 , @xmath77 m , @xmath78 fs ( fs - pulse case ) or @xmath81 ps ( ps - pulse case ) . , title="fig:",width=264 ] , @xmath67 mev , @xmath68 , @xmath69 ps , @xmath70 ps , @xmath71 , @xmath72 mev @xmath73m@xmath74 , @xmath75 , @xmath76 mev @xmath73m@xmath74 , @xmath77 m , @xmath78 fs , @xmath79 , @xmath80 , @xmath77 m , @xmath78 fs ( fs - pulse case ) or @xmath81 ps ( ps - pulse case ) . , title="fig:",width=264 ] to overcome the defocusing in the case of lp branch excitation , we can consider a momentum - dependent loss of the rabi coupling . assuming a stronger reduction of @xmath82 in the vicinity of @xmath20 , where lies the condensate whose density is responsible for the phase - space filling , allows to create a negative effective mass for the lower polariton . indeed , the lp polariton branch blueshifts more at @xmath20 , creating a negative curvature of the dispersion around the ground state , allowing for self - focusing even with repulsive interactions . technically , this can be achieved by filtering eq . ( [ eq : viemay8103312cest2015 ] ) in @xmath21 space . in the simplest formulation , this is done by transforming the excitonic wavefunction @xmath83 to @xmath21-space , applying a gaussian filter @xmath84 , transforming it back , and subtracting from the @xmath63 term after multiplication by @xmath85 . this nonlocal and nonlinear reduction of the rabi coupling is equivalent to introducing interactions that are nonlocal in real space . the range of the nonlocal effective interaction is given by @xmath86 . figure [ fig : negative ] shows an example of self - focusing in this model in the case of picosecond lp branch excitation . in the case of femtosecond excitation , the results are similar to the ones obtained with the previous model . ( for ps pulse excitation at the lp branch ) with @xmath87 and @xmath88m@xmath89 . , width=302 ] while this model provides a good agreement with experiments performed at different settings , its justification is not straightforward . while the @xmath21-dependent reduction of rabi coupling can be predicted by the simple model of phase space filling or exchange effects @xcite , it occurs at @xmath21 vectors comparable to the inverse exciton bohr radius @xmath90 . the dependence of the @xmath82 on momentum for @xmath21 comparable to @xmath91m@xmath89 requires an existence of correlations over distances much larger than the exciton radius , for which we see no rationale in the mean - field framework . an explanation along these lines is thus thwarted in a straightforward model and would require a drastic reconsideration of the physics of polaritons based on the mean - field approximation and perturbation theory in the exciton basis . in the close vicinity of the mott transition , the description in terms of the dilute exciton gas may become questionable , and collective effects of strongly - correlated quantum gases could be required instead . indeed , bcs - related physics is expected to occur close to the phase space filling threshold @xcite . such an interpretation can not be excluded and would open a new page of the field , but it is , however , much beyond the scope of this text . the exciton reservoir is another possible candidate to explain the redistribution of the polariton condensate . a conceivable scenario involves the creation of a ring of reservoir excitons around the center of the pumping spot , which subsequently creates a backflow of polariton waves , interfering constructively in the center and giving rise to the strong intensity peak . such a model consists of the photon , exciton , and reservoir fields , similar to ref . @xcite , but treating coherent photons and excitons with separate fields : @xmath92 is the energy relaxation constant , @xmath93 is the pumping field and with an initial reservoir density @xmath94 . in this simple model , the backscattering from the condensate to the reservoir is not taken into account , but the initial reservoir density is assumed to be excited by the incoming pulse . , @xmath95 , @xmath96 , @xmath97 mev , @xmath98 ps , @xmath70 ps , @xmath99 mev @xmath73m@xmath74 , @xmath100 , @xmath76 mev @xmath73m@xmath74 , @xmath77 m , @xmath78 fs , @xmath79 . , title="fig:",width=302 ] , @xmath95 , @xmath96 , @xmath97 mev , @xmath98 ps , @xmath70 ps , @xmath99 mev @xmath73m@xmath74 , @xmath100 , @xmath76 mev @xmath73m@xmath74 , @xmath77 m , @xmath78 fs , @xmath79 . , title="fig:",width=302 ] a numerical simulation of this model is shown in fig . [ fig : reservoir ] . the depletion in the center of the reservoir is created due to the stimulated scattering to the condensate , which is faster in the center , where the condensate density is larger . the bright density peak is then created because the polariton waves emitted form the remaining ring - shaped reservoir `` focus '' in the center . while the main feature of the experiment is thus reproduced , there are several problems with quantitative features : ( a ) the central peak appears much later than is observed in the experiment and ( b ) the bending of the rabi oscillations is initially in the incorrect direction ( backwards in the center ) . there are even more serious conceptual difficulties : ( c ) in such a model , the peak can not move `` ballistically '' , as is observed in the experiment if the excitation pulse is injected at an angle , since the reservoir would stay essentially in the position where it is created . ( d ) the peak persists when resonantly exciting the lower branch , which is expected to produce little or no reservoir excitons . ( e ) finally , the effect does not show any strong polarization dependence . in the regime of quasiresonant pulsed excitation we expect the reservoir to be polarised as the typical exciton spin - relaxation time in our system is supposed to be long compared to the characteristic time scale of the rabi - oscillations and collapse dynamics variations on this theme are possible . for instance , one can assume the intermission of dark excitons . since the coupling is incoherent and assuming the exciton mass as infinite , which allows to treat dark excitons as a density field , their effect can be modeled by the following equations : @xmath101 a numerical simulation of this model is shown in fig . [ fig : alexey ] . the scattering to polaritons creates a hole in the gaussian distribution of excitons . the flow towards the hole can create a localization , however , the effect is not as clear as in the experiment : the localization is rather broad and has a low intensity . , except @xmath102 , @xmath69 ps , @xmath103 @xmath73m@xmath74 ps@xmath89 , @xmath104 @xmath73m@xmath105 ps@xmath89 , @xmath72 mev @xmath73m@xmath74 , @xmath106.,width=302 ] the interaction constant in the gross - pitaevskii equations , and variants , that describe the polariton dynamics is positive , as indeed same - spin excitons are largely believed to be repulsive , but several effects could potentially give rise to attractive interactions , as we discuss shortly . since we actually observe a real - space collapse , it is instructive to simply assume attractive interactions between polaritons . in this case self - focusing is indeed possible for a positive effective mass . the model is identical to the one above , eqs . ( [ eq : gpsystem ] ) , but with @xmath61 . we do not include the reservoir excitons that we have seen can be ruled out on several grounds . a result of numerical simulation is shown in fig . [ fig : attractive ] . as expected , the high intensity peak appears due to self - focusing . while some features of the experiment are well reproduced in this model , the existence of attractive polariton interactions is highly questionable . potentially the strongest effect in this sense is the indirect exchange through the biexciton state , but this occurs only for opposite spins , while we did not observe any strong spin dependence of the focusing effect in the experiment . the other possible sources of attractive interaction , such as the indirect exchange @xcite , or van der waals interactions appear to be negligible for the low - momentum excitons created by resonant pumping @xcite . it was also recently observed that the sign of the interactions may change to negative at large polariton momenta @xcite , and other effects could similarly give rise to polariton attraction @xcite , but the proposed configurations are difficult to fit in with our experiment . the change of sign of the polariton interactions with no further justification is thus not tenable and would be in conflict with a large literature that successfully describes various experimental regimes on a physically well motivated assumption of repulsive interactions . , except @xmath107 , @xmath108 mev , @xmath109 mev @xmath73m@xmath74 , @xmath110 , @xmath111 mev @xmath73m@xmath74 . , width=302 ] we now conclude this series of possible models to reproduce the real - space collapse of the polariton condensate with the one model that we found could stand as an explanation for the observed effect . when the exciton density is large enough , as in the case of pulsed excitation , the macroscopic occupation of the ground state with excitons may lead to enhancement of the exciton - phonon interaction that produces a cooling of the excitonic reservoir against its blueshift due to their attraction @xcite . the corresponding released energy is transfered to the crystal lattice via acoustic phonons . another plausible mechanism of heating is through the polariton - polariton auger process followed by emission of the cascade of acoustic phonons as discussed by klembt _ et al . _ @xcite . in any case , a local heating may occur which leads to a band - gap narrowing , a variation of the exciton - photon detuning and a redshift of the lower polariton branch . the latter effect may lead to self - focusing of the polariton condensate . such an effect can be referred to as a `` collective polaron effect '' . formally , it can be described by the modulation of the exciton energy through a retarded function : @xmath112 where @xmath113 characterises the magnitude of band - gap renormalization and @xmath114 is the characteristic heat relaxation time in the crystal lattice . both @xmath115 and @xmath114 are fitting parameters . the fourth power of the exciton wavefunction @xmath116 under the integral accounts for the quadratic dependence of the polariton auger process on the concentration of polaritons . the dynamics of the system is governed by the initial pumping strength . the rabi oscillations generally have an anharmonic behavior due to the changes of the bare exciton energy and the interplay between the blueshift @xmath117 and the redshift @xmath118 . the effective frequency of these oscillations is then determined by @xmath119 @xcite . in addition to the change in detuning , at larger population densities close to the mott transition , the coupling @xmath120 itself would be reduced , resulting from the change of the exciton oscillator strength due to the exciton phase space filling @xcite . , @xmath121 mev , @xmath122 , @xmath123 ps , @xmath124 ps , @xmath125 mev m@xmath74 , @xmath126 , @xmath127 cm@xmath89 . the pumping field is described by a fs - pulse with @xmath129 m , @xmath130 fs , @xmath131 , and a variable amplitude @xmath132 ( see text ) . ] the result of the numerical simulation of eqs . ( [ eq : gpsystem ] ) with an eq . ( [ eq : ex ] ) dependence is shown in fig . [ fig : heating ] . at the initial stage , the rabi oscillations between photon and exciton states are seen . the dependence of the oscillations frequency on the local detuning results in appearance of the radial waves spreading from the center of the spot . the modified density - dependent value of the rabi frequency allows to reproduce some retardation of their motion in the spot center . when the density decreases due to the photonic decay , the direction of the wave - spreading changes . in addition , the heating reservoir forms a potential well , in which the polaritons are trapped through their excitonic fraction , that results in a stabilization of both excitonic and photonic components of the polariton flow . further dynamics exibits the decay of both components of the polariton many - body wavefunction . one can see the typical behavior of the sharply localised peak appearance extracted at different pumping power , which is modeled by setting increasing amplitude of @xmath133 in eq . . the results are shown in fig . [ fig : densitypowerdependence ] . the photonic phase component is plotted in fig . [ fig : phasepowerdependence ] . the number of polaritons increases proportionally from @xmath33 to @xmath134 . other parameters are the same as in the caption of fig . [ fig : heating ] . ] . ] in the present model as well , we assume that the incoherent reservoir is either completely empty or does not play any role . this assumption is certainly valid in the case of selective picosecond excitation of the lower polariton ( lp ) branch . we note also that the weak - coupling saturable optical nonlinearity could arise from instantaneous effects or sample heating @xcite . the instantaneous part , however , provides only defocusing nonlinearity for excitation below the band edge , as follows from the nonlinear kramers - kronig relations @xcite . heating generally leads to redshift which corresponds to focusing nonlinearity but has a longer response time . in conclusions , while there is no compelling argument that impose the collactive polaron effect as the mechanism causing the collapse of the polariton condensate , our simulations show that it is a reasonable candidate , when other more straightforward scenarios fail to provide even a qualitative agreement or demand assumptions too strong to be justified . it is not excluded that more exotic many - body physics is at play , such as a variation of the bcs mechanism , holding the bosonic condensate together in a way similar as the superconducting phase molds the fermion liquid , or a dynamical casimir effect @xcite pulling non - zero - momentum particles out of the suddenly hit polariton vacuum . we feel that it is left to theorists and further investigations to elucidate which exact physics is causing the peculiar phenomenology that we report . -.1 cm movies s1-s2 . the movies show the complete imaging of the fluid dynamics over a timespan of 40 ps with a timestep of 50 fs , as 3d and 2d maps , respectively . this gives a direct view of the ultrafast superfluid dynamics , combining rabi oscillations and the central collapse along with the other phenomena described in the text and in fig . . movie s3 . same as the previous movies but now showing the corresponding radial cross section of the polariton density @xmath135 . as an interesting detail , it is possible to observe how the initial expansion in the first few ps is happening during the half - periods of the rabi cycles where there is a low photonic density ( i.e. , a high excitonic one ) . [ link to supplementary movies ]	* polaritons in microcavities are versatile quasi-2d bosonic particles with a high degree of coherence and strong nonlinearities , thanks to their hybrid light - matter character @xcite . in their condensed form @xcite , they display striking quantum hydrodynamic features analogous to atomic bose - einstein condensates , such as long - range order coherence , superfluidity @xcite and quantized vorticity @xcite . their variegated dispersive and dissipative properties @xcite , however , set significant differences from their atomic counterpart @xcite . in this work , we report the unique phenomenology that is observed when a pulse of light impacts the polariton vacuum : the condensate that is instantaneously formed does not splash in real space but instead coheres into an enigmatic structure , featuring concentric rings and , most notably , a sharp and bright peak at the center . using a state - of - the - art ultrafast imaging with 50 fs time steps , we are able to track the dynamics of the polariton mean - field wavefunction in both real and reciprocal space . the observation of the real - space collapse of the condensate into an extremely localized resolution limited peak is at odd with the repulsive interactions of polaritons and their positive effective mass . an unconventional mechanism is therefore at play to account for our observations . our modeling suggests that self - trapping due to a local heating of the crystal lattice that can be described as a collective polaron formed by a polariton condensate could be involved . these observations hint at the fascinating fluid dynamics of polaritons in conditions of extreme intensities and ultrafast times . * microcavity polaritons have been praised for their fast response times @xcite and ease of manipulation as well as detection , inherited by the photonic component , while keeping a strong nonlinear character @xcite , conferred by the excitons @xcite . this makes them increasingly strong contenders in the field of interacting quantum fluids , where they have demonstrated the prevailing phases of strongly correlated systems @xcite , including bose einstein condensation @xcite , superfluidity @xcite , e.g. , scatterless flow , quantized vortices @xcite , together with rich spinorial patterns @xcite and nonlinear interference effects . polaritons have also demonstrated their suitability to investigate another mainstream concept fueled by dispersive and dissipative nonlinearities : shock waves and solitons , respectively characterized by step disturbances moving in the medium with sound velocity and by self - localization in space or shape preservation in time . beyond polariton fluids @xcite , these have also drawn much attention in nonlinear media @xcite , atomic bose - einstein condensates ( becs ) @xcite or microcavities in general @xcite . for instance , the response of a nonlinear medium or atomic becs to an impinging blast resulting in the irradiation of shock waves @xcite , or the appearance of solitonic states @xcite have been recently reported . time - resolved exploration of such effects in strongly correlated gases remains largely unexplored and with the ultrafast imaging techniques now available , one is at the dawn of a new era for the investigation of quantum fluid dynamics . while they have reproduced most of the known phenomenology of quantum gases , polaritons also come with peculiarities of their own , such as their dispersion relation or their short lifetime , making them intrinsically out - of - equilibrium . here , we report what appears to be a unique phenomenology of these systems , observed after the sudden coherent generation of a polariton condensate . the fluid undergoes a space redistribution leading to a central localization of a great number of polaritons , despite their repulsive interactions . the peak that is formed reaches a localization ( @xmath0 m , resolution limited ) at least ten times sharper than the initial gaussian spot injected by the laser ( 18.5 @xmath1 m ) . it also gathers a large number of particles , with a local enhancement up to 10 times the original density of polaritons . this striking dynamics takes place in a few ps . all these features can be tuned continuously with the excitation power . another interesting feature of the dynamics is the generation of a shock wave at early times and concentric rings at later times . similar rings and shock waves have been observed in nonlinear defocusing optical media or repulsive atomic becs @xcite , but the presence of a central localised and enhanced peak has never been reported so far in any system , to the best of our knowledge . the sample we used consists of a 2@xmath2 microcavity ( mc ) containing a triple quantum well ( qw ) positioned in the maxima of the electromagnetic field @xcite . an excellent quality factor ( photonic @xmath3 ) is given by two high reflectance multilayers mirrors ( dbr , distributed bragg reflectors ) embedding the mc . the strong coupling between the qw excitons and mc photons manifests itself as an anticrossing of their original modes splitting into two new normal modes , known as polaritonic upper ( upb ) and lower ( lpb ) branches ( see supplementary figure s5 ) . all the experiments shown here are performed at zero detuning between the mc and the qw exciton , both resonant at @xmath4 nm , and in a region of the sample clean from defects in order to avoid any effect due to spatial inhomogeneities . the device is kept at a temperature of 10 k. we implemented an unltrafast imaging technique based on the off - axis digital holography @xcite to study the dynamics of the polariton flow with spatial and temporal steps of 0.16 @xmath5 and 50 fs respectively . a 130 fs or 3.5 ps laser pulse , on resonance with the bottom of the lpb and circularly or linearly polarized , is directed onto the sample at normal incidence and the evolution of the polariton state in time is recorded by making interfere the sample emission with a delayed reference beam into a ccd camera . digital elaboration in the reciprocal space of the interferograms allows to retrieve the complex wavefunction of the photonic emission , which is coherent with the polaritonic wavefunction , and thus allows us to image both the amplitude and phase of the light - matter fluid . for energy resolved images a standard 550 cm long spectrometer is used before the ccd and the spectra are time integrated . additional details on the sample and the technique can be found in ref . @xcite and the supplementary material therein . m area as 3d view and * b , f , j * ( second row ) amplitude maps as 2d view . the three columns represent time frames at @xmath6 ps ( a - d ) , 2.8 ps ( e - h ) and 10.4 ps ( i - l ) . these time frames correspond respectively to the pulse arrival , the ignition of the dynamical peak and its long - lived state sitting at the center of a ring structure ( see also video s1s3 ) . * c , g , k * ( third row ) phase maps and * d , h , l * ( fourth row ) unwrapped phase profiles along the radius . the phase gradient subtends the superflow and here exhibits a reversal of the phase curvature , leading to the development of an opposite flow , toward the center . the total number of particles intially excited in the whole area is @xmath7 polaritons as evaluated by an independent measure of the bare emission . ] the dynamics of the polariton fluid , generated instantaneously by a resonant laser pulse hitting the sample , is shown in fig . 1 . initially , the polariton distribution is simply a footprint of the incoming laser spot , i.e. , a gaussian of 18.5 @xmath1 m ( fwhm ) . panels ( a - c ) shown the sample emission at the pulse arrival , in density ( a ) , amplitude ( b ) and phase ( c ) . the unwrapped phase along a central diameter is shown in panel ( d ) , with the shallow positive curvature associated to a weak outward gradient . the central density of polaritons is initially 550 @xmath8 . for the first few ps , the fluid only slightly decreases its total number of particles due to radiative losses however , at @xmath9 ps , the maximum density suffers a sudden depletion , down to only a quarter of the initial density . this is shown in panels ( e , f ) for the intensity and ( g , h ) for the phase . this unanticipated jolt marks the beginning of the redistribution of the fluid . note that the phase space profile starts to manifest a negative curvature with a reversal point of the phase gradient ( @xmath10 ) at @xmath11 m . such an inversion in the phase gradient is an indication of a change in the fluid direction , from waves expanding outwards , to contracting towards the centre . this behaviour is then followed by the appearance of the bright and sharp central peak , contoured by concentric rings ( i , j ) . this central peak collects 6% of the total particles still present in the fluid , and is surprisingly brighter than the initial spot was in the same area ( here by a factor of two , up to ten times in other realizations , see supplementary fig . s3 ) despite the total population having largely decreased ( to less than @xmath12 after 10 ps ) . its localization goes below the resolution of our experimental setup and is thus less than 2 @xmath1 m in width . the peak is furthermore extremely robust as it occurs over the whole sample area , at different mc / qw detunings and even sustains motion , as it actually propagates if imparted with an initial momentum ( see supplementary material ) . it is observed in the time - integrated camera images of the direct emission before any subsequent digital elaboration , which excludes any artifact of the technique ( see also fig . s3 in the supplementary material ) . the full dynamics of the polariton fluid and the connection between the density accumulation and the radial flows are further studied in fig . 2 . here are shown the amplitude and phase profiles versus time [ panels ( a ) and ( b ) ] . the amplitude chart reveals the bright peak as a central horizontal line which reaches its maximum intensity at a time of @xmath1311 ps . the vertical stripes observed in the first 5 - 6 ps are rabi oscillations between the excitonic and photonic fields . the period of @xmath14 fs corresponds to the energy separation between the upb and lpb of 5.4 mev , while their fast decay is due to the rapid scattering of the upb polaritons . the rabi oscillations are triggered by the femtosecond pulse that excites simultaneously both polariton branches ( 9 nm energy width , see also supplementary fig . s1 and ref . @xcite ) . they do not , however , play an important role in the observed phenomenology , since the dynamical localization also happens under the excitation of the lpb alone by a picosecond laser pulse ( 0.35 nm width , see supplementary fig . s2 ) . this confirms the robustness of this peak . since it has an homogeneous phase , as can be seen from the phase graph in fig . 2(b ) , it is a standing wave . we also note that a neat time - space cone marks a boundary between two regions : an expanding internal domain with almost horizontal black and white ( b / w ) bands ( strong inward @xmath15 ) and an external domain with almost vertical b / w bands ( null or weak @xmath15 ) . this is a clear evidence that a circular front of phase disturbance ( marked also by a low density ) is expanding ( @xmath16m / ps ) , leaving after its passage a fragmentation into multiple rings , as illustrated in the still image of panel ( d ) . if an attractive term is at play here , it thus seems to be in the course of expanding its range of action . to further characterize the flow dynamics and its role in the formation of the central peak , we extend our study to the reciprocal space , reporting the 1d cross section of the @xmath17 plane in panel ( c ) of figure 2 . the initial @xmath18 width of the polariton population created by the laser pulse is very small ( @xmath19 fwhm ) and concentrated around @xmath20 . after a couple of picoseconds , the fluid suddenly ejects a disk in @xmath21-space which stabilizes within 10 ps into a ring at finite momenta around @xmath22 . these are clearly associated to the inner growing flow . looking at the onset of the rabi oscillations allows us to make a correspondence between real and reciprocal space , indicating that the central bright peak in real space is associated to the finite momenta travelling waves in reciprocal space . the energy of these waves is shown in the integrated dispersion of fig . 2(f ) @xcite . note that , as oberved in the dispersion plotted in fig . 2(f ) , the ring in @xmath21 space gradually expands from @xmath20 at early times ( maximum blushift of the polariton pupulation ) to the value of @xmath23 when the dispersion is redshifted to its original bare energy . the space - energy profile in figure 2(e ) explains in part the inner flow . here it is clear that the central part of the spot is redshifted with respect to the sides , hinting at the presence of an effective attractive potential responsible for the central peak of high polariton density . although it is not clear at the outset what the origin of such a potential is , it is consistent with an attractive term ( nonlinearity inversion ) or with nonlocal interactions , i.e. , @xmath21-dependent blueshift @xcite . in our case , we may reasonably infer that the inward coherent waves generated by such a potential interfere in the center with opposite @xmath24 vectors , explaining the central accumulation enhanced by a polar scaling of @xmath25 which is typical of an interfering ring wave ( see figure s7 in the sm ) . ps is shown together with the associated amplitude profile . the dip in the intensity with the @xmath26-jump in phase is a signature of radial interference and of a possible dark ring soliton , surrounding the bright peak . * c * , radial phase profile at later time , taken each 5 ps over a 040 ps timespan . the phase slope increases with time up to @xmath27 . * d * , amplitude and phase profiles at 16 ps showing that a nonlinear interference is reshaping the fluid in a series of concentric rings . ] to verify the nature of the interference , we report the phase and amplitude profiles at different time stills . in fig . 3(a ) , the phase profile during the first 5 ps is shown . soon after the phase reversal , a sudden , almost instantaneous switch happens at @xmath28 ps . the rigid switch is a @xmath29-jump , corresponding to the dark notch in the density , as seen in panel ( b ) . this is a typical sign of an interference between counterpropagating waves . in the following dynamics , as already said , the reversed gradient tends to grow and expand all over the originally excited area , as depicted in panel ( c ) with profiles at 5 ps intervals . other @xmath29-jumps in the phase can be observed at a later time , corresponding to dark rings in the density , see fig . 3(d ) . this is an evidence that interference phenomena of coherent waves are acting in reshaping the fluid density as a series of concentric rings . it also suggests a possible link to without being a `` per - se '' proof of ring dark solitons ( rds ) @xcite . these specific solutions to nonlinear schrdinger equations ( nlse ) under repulsive interactions are predicted to become stable in the case of 2d fluids such as polaritons @xcite . it appears that a rds indeed holds for the first dark ring around the rising bright peak , given its stability for several tens of ps . the nonlinear nature of the effect is demonstrated in fig . 4 . at low density , fig . 4(a , b , c ) , the condensate behaves as expected from any fluid freely released , with a small diffusion and remaining homogeneous in both density and phase , as well as , in our case , rabi oscillations at earlier times . it is also clear that the expansion speed imparted by the small initial @xmath30 is negligible . at 5-times higher excitation power [ fig . 4(d , e , f ) ] , there is a density redistribution of polaritons to form the central localization peak , surrounded by ring structures and out - radiating shock waves , however , yet without a strong enhancement and a moderate phase reversal . in fig.4(g , h , i ) , at 18 times the initial pumping power , the structure gets fully formed , with a central peak gathering over twice the population locally present at the initial time and with a much steeper phase reversal , giving rise to the striking structure in fig . 4h . here again we emphasize that the central peak is resolution limited and is likely sharper than is resolved in our experiment . additional examples are provided in the supplementary fig . s3 for an extended set of excitation densities . we now discuss which physical mechanism could be responsible for such a remarkable phenomenology . the strong - coupling regime of light and matter at the core of the polariton physics can lead to distinctive dispersive and dissipative nonlinearities @xcite . for instance , polaritons support dissipative solitons @xcite , have demonstrated bistability domains with a switching on / off of both bright and dark solitons @xcite as well as moving bright solitons along a steadily pumped background @xcite ( this last based on the negative curvature of the polariton dispersion above the inflexion point ) . all these features are accountable by one of the several models used to describe polariton fluids . while the positive nonlinearities intrinsic to polariton interactions , due to the excitonic repulsions , are supposedly able to force the expansion and reshaping of a polariton fluid and to sustain dark solitons @xcite , possibly shedding light to some aspect of our experiment , there is no documented mechanism to explain the most striking feature : the real - space collapse in the center of the spot . real space localization could in principle appear under negative , i.e. , attractive , nonlinearities @xcite . as we review below all the obvious candidates to account for the observed phenomenology , we can rule out this and other tempting explanations . our analysis will show that the most likely origin of the nonlinear activation of a central bright spot is the self - trapping of a polariton condensate by a type of collective polaron effect . a first possibility to explain the self - localization is for the polariton population at early times to undergo a transition to the weak coupling regime ( screening and reduction of coupling ) . although this can be partially happening at the largest power ( given the excited density there is @xmath31 per qw and approaches values of the so - called mott density , in the order of @xmath32 ) , it does not explain the effect since ( i ) the blueshift of the lpb is continuous and does not reach the photonic mode ; ( ii ) increasing power / density does not increase the size of the bright peak , as expected if merely enlarging the size of an above - threshold region ; ( iii ) the total intensity decays with the lp polariton lifetime of 10 ps ( see supplementary fig . s1 ) and ( iv ) transition to weak coupling for excitation below the band edge would manifest as a higher blueshift as follows from the nonlinear kramers - kronig relations @xcite , while we observe a lesser blueshift in the centre with respect to the side bright ring , as shown in fig . 2(e ) . an exciton reservoir , separate from the polariton condensate , is known to play a significant role in many polariton experiments . we can exclude it in our case for the following reasons : ( i ) the effect persists when resonantly exciting the lower polariton branch , a configuration that does not populate the reservoir ( see the supplementary fig . s2 on the picosecond experiment ) , ( ii ) a moving bright peak is observed when exciting with nonzero initial in - plane wavevector @xmath21 , ruling out a reservoir that would need to be dragged by the structure as it propagates , which is impossible given the heavy mass of the reservoir excitons ( see the supplementary fig . s6 on the moving peak ) , ( iii ) the effect does not show any strong polarization dependence . these considerations thus exclude the case of a dissipative bright soliton @xcite predicted under cw pumping where the localization is sculpted by an interplay between source and decay regions and the compensation of their steady flows . the appealing recourse to attractive interactions can not be sustained either . while polariton attractions are possible due to the various superexchange processes through dark excitons or bi - exciton states , or due to van der waals forces , none of these mechanisms can account for the experiment in a careful analysis . the former mechanism should be strongly polarisation dependent @xcite , which is not our case , and the latter are too small with respect to the repulsive coulomb or exchange terms @xcite . similarly , nonlocality ( @xmath21-dependence , which could also lead to a negative dispersion in the centre of the spot @xcite ) of one or more of these terms and even retardation effects are negligible as well @xcite and fail to produce the real - space collapse in numerical simulations . intriguingly , the dynamical casimir effect @xcite recently proposed @xcite in a configuration very similar to our experiments , was predicted to generate finite momentum excitations from the vacuum adding up to the suddenly excited fluid at @xmath20 . however , this effect , was considered only for the homogeneous 1d polariton at zero temperature , and the model is not ready to be compared at its stage of development with the configuration of our experiment . the failures of these analyses point at an unconventional mechanism ruling the high - density , ultrafast dynamics of polaritons . given that the object appears to be self - sustained , it is important to elucidate its nature , as it may have important applications , especially as the control of ultra - sharp localized light peaks is clearly of technological interest , for instance for high - resolution displays or memory units . one of the unconventional scenarios that we found to be fairly consistent with most the hypotheses and observations of our experiment involves a sort of collective polaron effect . the recent work by klembt _ et al . _ @xcite shows that the resonant pumping of exciton polaritons into a microcavity may result both in cooling or heating of the crystal lattice depending on the initial lattice temperature and the optical pump power . in our experiments , realised at a cryogenic temperature and sufficiently high pump power , one should expect such a local heating of the crystal lattice due to the polariton auger process , followed by the emission of a cascade of acoustic phonons . the probability of this process is quadratic in the polariton density . the heating results in the local band - gap renormalisation which is responsible for the red - shift of the exciton energy . the heating by 20 - 30 degrees results in a redshift of the exciton energy of 1 - 2 mev , which is sufficient for trapping the polariton condensate . in this way , a trap in real space is formed under the pump spot . it becomes deeper as more polaritons are getting trapped , thus providing a positive feedback that stabilizes the self - trapping process and explain the robustness of the effect . we have modeled this mechanism through a generalised gross - pitaevskii equation , described in the supplementary material . figure 5 shows a result of the numerical simulation for the wavefunction of the polariton condensate ruled by this process . the upper panel shows the real - space dynamics and next panels the energies of the low and upper polariton branches , respectively . beyond the dynamics of the rabi oscillations , in particular their bending , the model also reproduces the self - localization in good qualitative agreement with the experimental data . in conclusions , we have observed the dynamical appearance of a bright and sharp peak sitting at the center of a series of concentric rings in a polariton fluid generated by the sudden excitation from a resonant laser pulse at @xmath20 . the peak that appears at high pumping is robust to other variations in the experimental parameters ( detuning , momentum , etc . ) , is resolution limited and gathers up to ten times the population initially present in its area . this striking structure can not be explained by any of the conventional mechanisms such as loss of strong - coupling nor by the common models of polariton dynamics , including gross - pitaevskii type of equations with or without reservoirs and/or attractive interactions . we have provided a possible intepretation in terms of the collective polaron effect , resulting in a self - trapping of the polariton condensate . our results show that much is left to explore in the high - density and ultrafast dynamics of polaritons , with a striking and unique phenomenology that could open new areas of research and applications . we acknowledge r. houdr for the growth of the microcavity sample and the project erc polaflow for financial support . this work has been partially funded by the quandyde project of the anr france and by the clermont4 network program . mm acknowledges support from the national science center grant dec-2011/01/d / st3/00482 . 46 url # 1`#1`urlprefix[2]#2 [ 2][]#2 _ et al . _ _ _ * * , ( ) . http://dx.doi.org/10.1038/nature07640 . _ et al . _ _ _ * * , ( ) . http://dx.doi.org/10.1038/nature05131 . _ et al . _ . _ _ * * , ( ) . http://dx.doi.org/10.1038/nphys1364 . _ et al . _ . _ _ * * , ( ) . http://dx.doi.org/10.1038/nphys1959 . _ et al . _ . _ _ * * , ( ) . http://link.aps.org/doi/10.1103/physrevb.87.205307 . _ et al . _ . _ _ * * , ( ) . http://link.aps.org/doi/10.1103/physrevb.82.075301 . & . _ _ * * , ( ) . http://link.aps.org/doi/10.1103/revmodphys.85.299 . _ et al . _ . _ _ * * , ( ) . http://dx.doi.org/10.1038/nphoton.2011.267 . _ et al . _ . _ _ * * , ( ) . doi:10.1103/physrevlett.113.226401 . _ et al . _ _ _ * * , ( ) . http://www.ncbi.nlm.nih.gov/pubmed/23653190 . _ et al . _ . _ _ * * , ( ) . http://www.nature.com/doifinder/10.1038/nphys1668 . , , , & . _ _ * * ( ) . http://www.nature.com/ncomms/2013/131009/ncomms3590/abs/ncomms3590.html?message-global=remove . , , , & . _ _ * * , ( ) . http://link.aps.org/doi/10.1103/physrevb.86.020509 . , & . _ _ * * , ( ) . http://link.aps.org/doi/10.1103/physrevb.83.144513 . _ et al . _ . _ _ * * , ( ) . http://link.aps.org/doi/10.1103/physrevlett.112.046403 . _ et al . _ _ _ * * , ( ) . http://dx.doi.org/10.1038/ncomms2760 . & _ _ * * , ( ) . http://www.opticsexpress.org/abstract.cfm?uri=oe-18-20-21219 . , , , & . _ _ * * , ( ) . http://link.aps.org/doi/10.1103/physrevlett.109.243902 . , & . _ _ * * , ( ) . http://dx.doi.org/10.1038/nphys486 . , & . _ _ * * , ( ) . http://link.aps.org/doi/10.1103/physrevlett.101.170404 . , , & . _ _ * * , ( ) . http://www.sciencemag.org/content/293/5530/663.abstract . _ et al . _ . _ _ * * , ( ) . http://link.aps.org/doi/10.1103/physrevlett.92.230401 . , & . _ _ * * , ( ) . http://link.aps.org/doi/10.1103/physreva.79.023603 . , , & . _ _ * * , ( ) . http://link.aps.org/doi/10.1103/physrevlett.106.150401 . , & . _ _ * * , ( ) . http://link.aps.org/doi/10.1103/physreva.69.063605 . _ et al . _ _ _ * * , ( ) . http://dx.doi.org/10.1038/ncomms2893 . _ et al . _ . _ _ * * , ( ) . http://link.aps.org/doi/10.1103/physrevlett.94.080404 . _ et al . _ _ _ * * , ( ) . http://dx.doi.org/10.1038/nature01049 . _ et al . _ . _ _ * * , ( ) . http://dx.doi.org/10.1038/nphoton.2012.50 . , , & . _ _ * * , ( ) . http://ol.osa.org/abstract.cfm?uri=ol-32-20-2930 . _ et al . _ . _ _ * * , ( ) . http://www.opticsexpress.org/abstract.cfm?uri=oe-20-15-16366 . & _ _ ( , ) . & . _ _ * * , ( ) . http://dx.doi.org/10.1002/pssb.200560960 . & . _ _ * * , ( ) . http://journals.aps.org/pre/abstract/10.1103/physreve.50.r40 . _ et al . _ . _ _ * * , ( ) . http://stacks.iop.org/0953-8984/26/i=15/a=155801 . , , , & . _ _ * * , ( ) . http://dx.doi.org/10.1038/nphys2999 . & . _ _ * * , ( ) . http://stacks.iop.org/0034-4885/70/i=2/a=r03 . & . _ _ * * , ( ) . http://link.aps.org/doi/10.1103/physrevlett.79.4870 . , , , & . _ _ * * , ( ) . http://link.aps.org/doi/10.1103/physreva.86.013636 . , & . _ _ * * , ( ) . http://ol.osa.org/abstract.cfm?uri=ol-33-4-321 . _ _ * * , ( ) . , & . in & ( eds . ) _ _ , vol . of _ _ , ( , ) . http://dx.doi.org/10.1007/10928028_6 . , , & . _ _ * * , ( ) . http://dx.doi.org/10.1007/s10773-013-1834-6 . & . _ _ * * , ( ) . http://link.aps.org/doi/10.1103/physrevlett.112.036406 . _ et al . _ . _ _ * * , ( ) . http://link.aps.org/doi/10.1103/physrevlett.114.186403 . * supplementary material * in this supplementary material , we describe the supplementary movies , some supporting experimental figures and discuss several theoretical models used in attempt to describe the collapse in real space of the polariton condensate .
four - fermion theories @xcite find applications in several areas of physics , from condensed matter systems ( for example in models for polymers , high temperature superconductors , etc ) to high energy physics , most notably as effective models for qcd , like the nambu jona - lasinio ( njl ) model @xcite and its variants , including the gross neveu ( gn ) model @xcite . these theories are typically employed in the study of chiral transition , either for the discrete form of the symmetry , @xmath5 , or in the continuous form , @xmath6 . the general interest is to understand how the chiral transition pattern is affected by external control parameters such as the temperature ( @xmath7 ) and the chemical potential ( @xmath8 ) . it is also well known that , apart from these control parameters , the presence of an external magnetic field may impact significantly on the phase transition patterns . so far we have a reasonable understanding of how chiral symmetry is affected by @xmath7 , @xmath8 as well as by the presence of a magnetic field @xmath9 . mostly , this understanding is acquired at the mean - field level @xcite for both the gn and the njl models . recently , the understanding of how an external @xmath9 field affects the symmetry aspects of these four - fermion theories turned out to be a question of general interest for the following reasons . first , the interaction of fermions with an external @xmath9 field is expected to be associated with phenomena such as the metal - to - insulating phase transition in semiconductors @xcite , quantum hall effects @xcite and the transport properties in superconductors @xcite , just to mention a few phenomena in the context of condensed matter physics , while in the high energy physics domain the effects of a magnetic background is important to the physics of compact stellar objects @xcite , heavy - ion collisions at the relativistic heavy ion collider and at the lhc @xcite and in the physics of the early universe @xcite , which are situations where high intensity fields , @xmath10 , are expected to be present or produced ( for a recent review , see e.g. ref . @xcite and references therein ) . of topical interest is to understand how a magnetic field will affect phase transitions , since they can induce dynamical symmetry breaking , or magnetic catalysis @xcite . within fermionic systems , magnetic catalysis refers to the generation of a mass gap for the fermions at any finite interaction strength , leading explicitly to chiral symmetry breaking . in this work , we apply the optimized perturbation theory ( opt ) @xcite ( see ref . @xcite for early works on this subject ) to the gn model in @xmath11 dimensions and investigate how dynamical chiral symmetry breaking ( csb ) is affected by the presence of a magnetic field , comparing our results with those obtained in the large-@xmath0 ( ln ) approximation , which is equivalent to the well known mean field approximation ( mfa ) . the opt has already established itself as a powerful method in dealing with critical theories . for example , in the bose einstein condensation case this method and its different variations have provided some of the most precise analytical results for the shift in the critical temperature for weakly interacting homogeneous bose gases @xcite . other applications to condensed matter situations include a precise evaluation of the critical density for polyacetylene @xcite . also , when extended by hard - thermal loops , the method was successful in predicting qcd thermodynamical properties at the three - loop level @xcite . improved by the renormalization group , and inspired by similar properties @xcite in the gross neveu model , a variation of the opt has been recently used in the evaluation of @xmath12 @xcite and @xmath13 @xcite , where the stability and convergence at higher orders of this renormalization group opt form was demonstrated . for the present application it is worth mentioning that the opt was instrumental in the determination of the phase diagram of the massless gn model in 2 + 1 dimensions at finite @xmath7 and @xmath8 in the absence of magnetic fields @xcite . in this case , the ln approximation predicts that the whole @xmath14 plane is dominated by a second - order phase transition , except at @xmath15 , where a first - order phase transition is predicted to occur . but , monte carlo numerical simulations @xcite have indicated that a first - order transition line should appear at the low-@xmath7 and high-@xmath8 region , terminating in a tricritical point at intermediate values of @xmath7 and @xmath8 . however , no precise location for this tricritical point was possible to be given . this situation has been changed when the complete phase diagram for the model was studied in the context of the opt method and the precise location of the tricritical point determined for any value of @xmath0 @xcite . the two - flavor njl model in @xmath16 dimensions with physical quark masses has also been treated with the opt at finite @xmath7 and @xmath8 in the absence of magnetic fields @xcite . the main outcome was that the @xmath17 corrections brought in by this approximation generate the effects of a repulsive vector channel ( absent in the original lagrangian ) , which weakens the first - order transition line and locates the critical end point at temperature values that are smaller than the ones predicted by the ln approximation . a detailed discussion about the physical nature of the opt @xmath17 corrections in the simplified abelian njl model context was recently carried out in ref . @xcite . in the massless gn model , the ln approximation predicts that the critical temperature , which signals that chiral symmetry has been restored through a second - order phase transition , increases with @xmath9 at vanishing fermionic densities . however , although the functional renormalization group technique @xcite has been recently applied to analyze magnetic catalysis at zero temperatures and densities , we are not aware of evaluations that go beyond mean field at finite @xmath7 , @xmath8 and @xmath9 in the context of the gn model in @xmath11 dimensions . in the present work we show that at @xmath18 the results obtained with the ln method and the opt agree from the qualitative point of view . namely , magnetic catalysis still takes place and the critical temperature rises with @xmath9 . however , at the other extreme of the phase diagram , when @xmath15 and @xmath19 , we find that the opt and the ln predict different qualitative and quantitative behaviors as far as the coexistence chemical potential @xmath20 is concerned . the differences are more pronounced for negative couplings ( @xmath21 ) , where the opt reproduces the phenomenon of inverse magnetic catalysis ( imc ) @xcite , which predicts the decrease of @xmath20 with increasing @xmath9 , even at large magnitude of magnetic fields . other qualitative differences happen in the region spanned by intermediate to high chemical potentials , where the opt adds terms of the form @xmath22 to the free energy , while only scalar condensates ( @xmath23 ) are considered within the ln approximation . then , depending on the sign of the four - fermion coupling @xmath2 and for sufficiently large values of the magnetic field , we observe the weakening ( @xmath24 ) or the enhancement ( @xmath4 ) of the entire csb region in the magnetized fermionic system . the opt also predicts that the value of @xmath20 ( at @xmath15 ) can be smaller than the tricritical point value ( @xmath25 ) , producing an important change in the shape of the phase diagram as compared to the one generated by taking @xmath26 . finally , we also discuss the possibility that the order parameter value suffers more than one discontinuity as @xmath8 increases when @xmath15 and @xmath9 is small . being observed by both approximations when @xmath3 , this feature is not an artifact of the ln approximation and can be easily explained by a close examination of the filling of the landau levels . this paper is organized as follows . in the next section we briefly review the planar gn four - fermion model within the opt formalism . in sec . [ sec3 ] we obtain the effective potential ( or free energy density ) , at first nontrivial order , which is adequate to treat a hot , dense and magnetized planar four - fermion system . next , in sec . [ sec4 ] we discuss how finite @xmath0 effects affect thermodynamical quantities such as the order parameter , the critical quantities and the overall shape of the phase diagram when a magnetic field is present . the results obtained from the opt are contrasted with those produced by the ln approximation in all the cases we have analyzed . finally , in sec . [ sec5 ] , we give our concluding remarks . two appendices are also included to show and clarify some technical aspects . in the presence of an external electromagnetic potential @xmath27 , the gn model with fermions with @xmath0 flavors , @xmath28 ( @xmath29 ) , is described by the lagrangian density @xcite @xmath30 note that a summation over fermionic species is implicit in the above equation with , e.g. , @xmath31 . when @xmath32 , which is the case considered by us here , the theory is invariant under the discrete chiral symmetry ( cs ) transformation @xmath33 with the gamma matrices being @xmath34 matrices and we follow the representation given , e.g. , in ref . @xcite for fermions in 2 + 1 dimensions dirac matrices ( see e.g. ref . @xcite for details ) , as we implicitly consider in this work . ] . a constant magnetic field @xmath9 along the @xmath35 direction , perpendicular to the plane of the system defined by eq . ( [ gn ] ) , can be considered by choosing a gauge where the external electromagnetic potential is given , for example , by @xmath36 . the ln limit ( or mfa ) of the model eq . ( [ gn ] ) is defined by considering the four - fermion interaction as @xmath37 and taking @xmath38 , while keeping @xmath2 fixed ( see , e.g. , ref . @xcite ) . in the following we will study the model of eq . ( [ gn ] ) beyond the simplest mfa / ln approximation by employing the opt method . within the opt framework one makes use of a linear interpolation on the original model in terms of a fictitious parameter , @xmath39 ( used only for bookkeeping purposes ) , which allows for further expansions @xcite . then , following e.g. refs . @xcite , the interpolated gn four - fermion theory can be expressed as @xmath40 note that at @xmath41 we have a theory of free fermions , while at @xmath42 we recover the original theory . we can now rewrite the four - fermion interaction in eq . ( [ gnlde ] ) by introducing an auxiliary scalar field @xmath43 in the usual way @xcite , such that eq . ( [ gnlde ] ) becomes : @xmath44 where @xmath43 and the chiral operator are related , from the saddle - point solution for @xmath43 , by @xmath45 . renormalization issues do not arise at the level of the approximation considered in the present application , but the interested reader can find a comprehensive discussion in the context of the opt method in ref . @xcite , for example . any quantity computed from the above interpolated lagrangian density ( [ gndelta ] ) , at some finite order in @xmath39 , is dependent on the arbitrary mass parameter @xmath46 , which also serves as an infrared regulator . then , after the formal mathematical manipulations associated with the evaluation of the relevant green functions , one must fix the arbitrary @xmath46 in a judicious way . here , as in most of the previous works on the opt method ( see , e.g. , refs . @xcite ) , @xmath46 is fixed by using the principle of minimal sensitivity ( pms ) . within the pms procedure one requires that a physical quantity @xmath47 , that is calculated perturbatively to some @xmath48-th order in @xmath39 , be evaluated at the point where it is less sensitive to this mass parameter . this criterion then translates into the variational relation @xcite @xmath49 the optimum value @xmath50 that satisfies eq . ( [ pms ] ) must be a function of the original parameters , including the couplings , thus generating in that sense nonperturbative dependences in the coupling and other parameters of the model . following , e.g. , ref . @xcite , the effective potential ( or free - energy density ) for a constant background scalar field , @xmath51 , at first order in the opt approximation is given by @xmath52 where @xmath53 brings the first @xmath17 corrections to the effective potential . this contribution is explicitly given by @xmath54\ ; , \label{vn1}\ ] ] where @xmath55 is the @xmath56 contribution to the fermion self - energy , @xmath57 the traces in eqs . ( [ general ] ) and ( [ vn1 ] ) are taken over dirac s matrices only ( a factor of @xmath58 , corresponding to a closed fermionic loop , has already been taken into account @xcite in the above expressions ) . after taking the traces over the dirac s matrices and rearranging the terms , eq . ( [ general ] ) can be written as @xmath59 ^ 2 + \delta \frac{2 \lambda}{n } \left [ i \int_p \frac { p_0}{p^2 - \eta^2 + i \epsilon } \right ] ^2 \;. \label{vlde1ddim}\end{aligned}\ ] ] in the above expressions we are using the notation for the momentum integrations in @xmath60 dimensions , at finite temperature and chemical potential and in the presence of a constant magnetic field . expressed in terms of sums over discrete matsubara s frequencies and landau levels ( lls ) the integral measure is given by @xmath61 where @xmath62 , with @xmath63 , are the matsubara frequencies for fermions , with @xmath7 the temperature . the sum over @xmath64 are over the landau levels ( ll ) , with a density of states @xmath65 . the time and space components of the momentum are @xmath66 , where @xmath8 is the chemical potential , while @xmath67 gives the ( square of the ) landau energy levels , with the factor @xmath68 accounting for the degeneracy of the @xmath69 landau levels @xcite . then , using eq.([intp ] ) and performing the sums over the matsubara frequencies , the explicit expression for the effective potential ( [ vlde1ddim ] ) can be obtained . it can be written in the form @xmath70\ ; , \label{veffhtmu}\end{aligned}\ ] ] where @xmath71 , @xmath72 , are given by ( see also appendix [ definitions ] ) @xmath73 + \ln [ 1 + e^{-(e_j-\mu)/t } ] \right \}\ ; , \label{int1}\ ] ] @xmath74 } + \frac{1}{e_j [ 1 + e^{(e_j-\mu)/t } ] } \right \}\ ; , \label{int2}\ ] ] @xmath75\,\ , , \label{int3}\ ] ] where @xmath76 , @xmath77 and @xmath78 is the hurwitz zeta function @xcite , @xmath79 it can be easily shown from eq . ( [ veffhtmu ] ) that in the large-@xmath0 limit we reobtain the standard expression for the effective potential for this model , as found , e.g. , in the seminal papers @xcite . it is also useful to realize that eqs . ( [ int2 ] ) and ( [ int3 ] ) can be both expressed in terms of eq . ( [ int1 ] ) : @xmath80 finally , let us analyze the physical meaning of the opt @xmath17 corrections displayed by eq . ( [ veffhtmu ] ) . by recalling that at one - loop order one can write the fermion number density as @xmath81 and the scalar condensate as @xmath82 then , it is easy to see that eq . ( [ vn1 ] ) becomes @xmath83 \,,\ ] ] where we have used @xmath84 . therefore , contrary to the ln approximation , the opt brings in a @xmath17 suppressed term that only contributes at finite densities . thus , one may expect some important differences to arise as this term becomes more important at increasing @xmath8 values @xcite . as an aside , concerning the values of the coupling @xmath2 leading to chiral symmetry breaking , note that we can define a renormalization condition for the coupling as @xmath85 , with @xmath86 . in terms of a cutoff regularization , in @xmath87 dimensions @xmath88 defines ( for vanishing magnetic field ) a critical value for which chiral symmetry breaking can happen @xcite , such that for @xmath89 ( i.e. , corresponding to @xmath90 ) the model can be in the broken phase of the discrete chiral symmetry , while for @xmath91 ( i.e. , corresponding to @xmath92 ) , there is no chiral symmetry breaking . some authors prefer to work directly in terms of the bare coupling @xmath2 ( like in ref . @xcite ) , while others prefer to work in terms of the redefined coupling @xmath93 ( like , for example , in refs . the latter is necessarily the case when working directly in terms of dimensional regularization , as we consider here , since then the above integral vanishes by definition , and @xmath94 . note also that within dimensional regularization there are no additional divergences in the opt case ( see , e.g. , ref . @xcite for more details ) . the optimization of the effective potential , eq . ( [ veffhtmu ] ) , is easily implemented by applying the pms condition , eq . ( [ pms ] ) , to @xmath95 . let us initially apply the pms to the most general order-@xmath39 effective potential , which is given by eq . ( [ vlde1ddim ] ) . this exercise will help the reader to visualize the way the opt - pms resums the perturbative series . setting @xmath42 and applying the pms to eq . ( [ vlde1ddim ] ) , we obtain that @xmath96 \left ( 1 + \eta \frac{d}{d\eta } \right ) \left [ i \int_p \frac { 1}{p^2 - \eta^2 + i \epsilon } \right . \right ] \nonumber \\ & & + \left . \frac{\lambda}{n } \left ( i \int_p \frac { p_0}{p^2 - \eta^2 + i \epsilon } \right ) \frac{d}{d \eta } \left ( i \int_p \frac { p_0}{p^2 - \eta^2 + i \epsilon } \right ) \right \}\bigr\|_{\eta = { \bar \eta } } = 0\,\ , . \label{pmsselfconsistent}\end{aligned}\ ] ] from the result given by eq . ( [ int3 ] ) , the last term of eq . ( [ pmsselfconsistent ] ) only survives when @xmath19 . in the case @xmath18 , eq . ( [ pmsselfconsistent ] ) factorizes in a nice way , which allows us to understand the way the opt - pms procedure resums the series producing nonperturbative results . then , when @xmath18 , and using eq . ( [ sigma1 ] ) , the opt - pms eq . ( [ pmsselfconsistent ] ) factorizes to @xmath97 \left ( 1 + { \bar \eta } \frac{\partial}{\partial{\bar \eta } } \right ) \left [ i \int_p \frac { 1}{p^2- { \bar \eta}^2 + i \epsilon } \right ] = 0\ ; , \label{selfconsistent}\ ] ] leading to the self - consistent relation limit . ] @xmath98 which is valid for any temperature and number of space - time dimensions provided that @xmath18 . in this way the opt fermionic loops get contributions containing @xmath51 as well as a rainbow ( exchange ) type of self - energy terms , given by eq . ( [ sigma1 ] ) . note that when @xmath99 , @xmath100 and the large-@xmath0 result is exactly reproduced @xcite . when @xmath101 , we can consider eq . ( [ pmsselfconsistent ] ) in order to get the general result in terms of the @xmath102 ( @xmath72 ) terms defined in the previous section . alternatively , using eqs . ( [ alti2 ] ) and ( [ alti3 ] ) we obtain that @xmath103 is given by the solution of @xmath104\bigr\|_{\eta = { \bar \eta}}\;. \label{etabar}\ ] ] again , we see from eq . ( [ etabar ] ) that as @xmath38 , the large-@xmath0 result @xmath105 is recovered . the pms equation ( [ etabar ] ) is to be solved together with the one defining the vacuum expectation value for the background field @xmath51 , @xmath106 we now have all the necessary tools to investigate all the possible cases when considering a finite constant magnetic field applied to the system . as a ballpark estimate we shall consider magnetic fields ranging from @xmath107 to @xmath108 , since the gap energy is about @xmath109 within this model . this choice is reasonable since , keeping in mind possible applications to condensed matter systems for example , typically , the gap energy lies within the range @xmath110 and if , for example , one considers the lower gap value @xmath111 then @xmath112 teslas , which is a realistic value within current planar condensed matter systems . finally , we set @xmath113 , since this is the relevant value as far as planar condensed matter systems ( like high - temperature superconductor films or graphene ) are concerned . let us preliminarily examine the case of zero temperature and zero chemical potential , @xmath116 . an important effect here is that of magnetic catalysis @xcite , which we next investigate in order to analyze how this phenomenon is affected by the nonperturbative inclusion of finite @xmath0 corrections through the opt . from eq . ( [ veffhtmu ] ) for the effective potential and upon using the results ( [ i1tmu0 ] ) , ( [ i2tmu0 ] ) , and ( [ i3tmu0 ] ) obtained in the appendix , we have that @xmath117 \nonumber \\ & + & \delta \frac { \lambda \eta^2}{32\pi^2 n}\left [ \frac{eb } { \eta } - ( 2eb)^{1/2 } \zeta\left ( \frac{1}{2},\frac{\eta^2}{2eb } \right ) \right ] ^2\;. \label{veffh}\end{aligned}\ ] ] then , from eq . ( [ etabar ] ) , we obtain the self - consistent relation to be evaluated for @xmath103 : @xmath118 \ , . \label{etabarh}\ ] ] by extremizing the effective potential ( [ veffh ] ) with respect to @xmath51 , setting @xmath42 and using the optimal @xmath119 , one finds that the order parameter satisfies @xmath120 \ , . \label{barsigmah}\ ] ] then , comparing the above equation with eq . ( [ etabarh ] ) , one is lead to the relation @xmath121 where @xmath122 by using eq . ( [ etaf ] ) in eq . ( [ barsigmah ] ) , we obtain @xmath123 chiral symmetry breaking can now be investigated by looking at the case @xmath4 , in which csb does not occur when @xmath124 , and also the case @xmath3 , where dynamical csb occurs even at @xmath124 . all quantities will be expressed in terms of the scale @xmath125 , which is the value of the chiral condensate @xmath126 in the ln case . our numerical results are compared to the ones given by the ln approximation in fig . [ sigmah ] for @xmath113 . we note that when either @xmath127 or @xmath3 , the magnetic catalysis is enhanced by the finite @xmath0 contributions that the opt - pms method brings , as fig [ sigmah ] shows . as in the previous subsection , we start with eq . ( [ veffhtmu ] ) for the free energy density and use the results ( [ i1mu0 ] ) , ( [ i2mu0 ] ) and ( [ i3mu0 ] ) to write @xmath129 \nonumber \\ & + & 4 \delta\ , \eta ( \eta - \sigma_c ) \left [ - \frac{eb}{8 \pi \eta } + \frac{(2eb)^{1/2}}{8\pi } \zeta\left ( \frac{1}{2},\frac{\eta^2}{2eb } \right ) - \frac{eb}{4\pi}\sum_{j=0}^{\infty}\alpha_j \frac{1 } { e_j \left(1 + e^{e_j / t } \right ) } \right ] \nonumber \\ & + & 2 \delta \ , \frac { \lambda}{n } \eta^2 \left [ - \frac{eb}{8 \pi \eta } + \frac{(2eb)^{1/2}}{8\pi } \zeta\left ( \frac{1}{2},\frac{\eta^2}{2eb } \right ) - \frac{eb}{4\pi}\sum_{j=0}^{\infty}\alpha_j \frac{1 } { e_j [ 1 + e^{e_j / t } ] } \right]^2 \ ; , \label{veffhtmu0}\end{aligned}\ ] ] the expression for the order parameter @xmath130 , equivalent to eq . ( [ barsigmahsol ] ) in the case of @xmath15 , now reads @xmath131}\;,\ ] ] where @xmath132 and we have used eq . ( [ etaf ] ) , which still holds at @xmath128 and @xmath115 . the thermal behavior for the order parameter , @xmath133 , is shown in fig . [ sigt ] for @xmath107 and for @xmath134 . in both cases the transition is of the second kind and , as expected , in the later case the symmetry restoration happens at a higher @xmath135 . figure [ tch ] shows how the critical temperature increases with @xmath9 , which is expected since @xmath126 increases with @xmath9 and @xmath136 . in this model even a strong magnetic field ( e.g. , @xmath137 ) is not able to change the character of the phase transition . it is interesting to note that for @xmath127 the opt predicts that the order parameter assumes higher values than the ones predicted by the ln approximation as @xmath9 increases as the left panel of fig . [ sigmah ] suggests . however , despite the fact that chiral symmetry seems to be more severely broken within the opt framework , the left panel of fig . [ tch ] shows that this symmetry will be restored at smaller critical temperatures than those predicted by the ln approximation . the next case we analyze is when @xmath138 and @xmath139 , which is relevant , for example , when analyzing charge asymmetries @xcite . the opt first - order result for the effective potential in this case becomes @xmath140\ ; , \label{veffht0mu}\end{aligned}\ ] ] where the expressions for @xmath141 , @xmath72 , are again as those given in the appendix [ definitions ] by eqs . ( [ i1t0 ] ) , ( [ i2t0 ] ) and ( [ i3t0 ] ) , respectively . then , using eq . ( [ barsigma ] ) one can write the order parameter as @xmath142\ ; , \label{barsigmat0mu}\ ] ] while the pms equation for @xmath143 now takes the more general form of eq . ( [ etabar ] ) , which after some algebra can be expressed for @xmath15 as @xmath144\ ; , \label{baretat0mu}\ ] ] where the relevant expressions for @xmath102 at @xmath15 are given in appendix [ definitions ] in eqs . ( [ i2t0])-([i3t0 ] ) and the opt corrections for @xmath143 to the simpler relation in eq . ( [ etaf ] ) , @xmath145 , are explicit from the second term . in eq . ( [ baretat0mu ] ) we have used @xmath146 which is independent of @xmath147 of @xmath148 before taking the @xmath149 limit ; otherwise , the derivative of expression ( [ i3t0 ] ) is ill - defined . the result ( [ di3eta ] ) is consistent with the @xmath150 limit @xcite for @xmath143 of expression ( [ baretat0mu ] ) ] . note , however , that the second term in ( [ baretat0mu ] ) being suppressed by @xmath151 gives a reasonably small correction , moreover only nonvanishing for @xmath152 due to the step function in @xmath153 eq . ( [ i3t0 ] ) . a legitimate approximation can thus be to use the simpler relation eq . ( [ etaf ] ) _ within _ this correction , instead of the implicit exact @xmath46 relation in ( [ baretat0mu ] ) , since the difference is of higher @xmath2 order , neglected anyway at the first opt @xmath39-order here considered . it is worth remarking that this opt correction term , when nonvanishing , may be positive or negative depending on the sign of @xmath2 and depending on the sign of @xmath154 ( while @xmath155 for any @xmath9 values ) . thus , it may enhance @xmath46 with respect to the ln result @xmath156 ( partly compensating the reduction from @xmath157 ) . using the @xmath15 analytical expression of @xmath158 and @xmath153 in eqs . ( [ i2t0 ] ) and ( [ i3t0 ] ) and some properties of the riemann hurwitz zeta functions , it is not difficult to recover the @xmath150 limit of ( [ baretat0mu ] ) , having relatively simple expressions : @xmath159 which is consistent with the direct @xmath124 calculation @xcite . + here , as we shall see , chiral symmetry is restored through a first - order phase transition as in the case of the absence of the external magnetic field @xcite . therefore , we must determine the coexistence chemical potential value , @xmath20 , at which the discontinuous chiral symmetry transition occurs . in this case , @xmath20 is obtained by solving ( see also discussion in the next subsection ) @xmath160 results for @xmath20 as a function of the magnetic field , for both cases of @xmath21 and @xmath161 , in the ln and opt cases , are shown in fig . [ figmuc ] . for @xmath162 and @xmath163 , one observes the typical de haas van alphen oscillations ( see appendix [ oscillations ] ) due to the filling of the landau levels . in fig . [ figmuclowh ] we show the critical chemical potential @xmath164 for @xmath21 for the region of low magnetic fields . this figure shows more clearly the typical oscillations at low magnetic fields , which are reminiscent of the de haas van alphen magnetic oscillations of the magnetization . these oscillations stop after @xmath165 reaches a value such that only the lowest landau level ( lll ) has to be considered ( here this happens at @xmath166 ) . after this point one sees a remarkable difference between the opt and the ln approximation results . the latter predicts that @xmath20 decreases with @xmath165 toward a minimum and then observes a sharp increase for @xmath167 . this ln result is in complete agreement with a mfa application to the three - flavor njl model performed in ref . @xcite ( see ref . @xcite for a detailed discussion on the first - order coexistence region ) . in contrast the decrease of @xmath20 with @xmath9 in the opt case is a manifestation of the inverse magnetic catalysis ( imc ) effect , which was explained e.g. in ref . @xcite ( the imc effect was first observed in the njl model in ref . @xcite at @xmath138 and in ref . @xcite for the full @xmath168 case ) . then , fig . [ figmuc ] shows that the opt results is more in line with this phenomenon , since only a smooth very moderate rise of @xmath20 is observed to occur between @xmath169 and @xmath170 , before it drops again at higher fields . this quantitative difference can be traced to the opt @xmath171 type of corrections , which are non - negligible in this region of high charge asymmetry . in fact the behavior can be essentially understood from a simple analytical approximation . first note that both terms of eq . ( [ 1stlinet0 ] ) considerably simplify . because of @xmath172 the right - hand side can be written as @xmath173 also , calculating @xmath95 from eq . ( [ veffht0mu ] ) at its minimum , using the relation eq . ( [ barsigma ] ) between @xmath174 and @xmath46 ( which is valid at the minimum of the potential for any values of the other parameters ) , the left - hand - side of eq . ( [ 1stlinet0 ] ) also simplifies to @xmath175 moreover , the last term in eq . ( [ vsigt0 ] ) vanishes whenever @xmath176 , which is the case in most of the parameter space considered , i.e. @xmath20 satisfying eq . ( [ 1stlinet0 ] ) will be such that @xmath177 . the ln case can be easily recovered from the above expressions by simply neglecting the opt correction @xmath178 term , and taking @xmath179 , @xmath180 in the remaining terms . for @xmath181 only the lowest landau level contributes to the relevant integrals , such that in this range @xmath181 eq . ( [ 1stlinet0 ] ) gives a relatively simple analytic ( implicit ) expression for @xmath20 in the opt case : @xmath182 + \frac{\lambda}{16\pi n\,e b}\left(e b+\mu^2_c \right)^2\ ; , \label{mucopt}\ ] ] while the corresponding expression in the ln case reads @xmath183\;. \label{mucln}\ ] ] the exact @xmath184 dependence in eqs . ( [ mucopt ] ) and ( [ mucln ] ) is rather involved , since @xmath51 depends nontrivially on @xmath184 , eq . ( [ barsigmat0mu ] ) . however , for a qualitative but essentially rather accurate understanding of the behavior in fig . [ figmuc ] , it is sufficient to know that @xmath185 is a moderately increasing function of @xmath184 . then the last term in eq . ( [ mucln ] ) involving the @xmath186-functions is monotonically increasing with @xmath184 ( first rapidly for moderate @xmath184 and then for large @xmath184 with a decreasing slope ) , so that together with the first terms it implies that @xmath187 gets a minimum at a moderate @xmath184 value , and then has a steeper rise . now , if there would only be the moderate difference @xmath188 from the ln to the opt case , the opt results would be qualitatively similar to the ln ones . in contrast , due to the last correction term in eq . ( [ mucopt ] ) , the behavior of @xmath189 is drastically different , since for @xmath3 the last terms goes for large @xmath184 as @xmath190 ( the @xmath191 in the last term being rapidly negligible in the relevant range @xmath192 ) , which thus prevents @xmath193 to increase fast , producing almost a plateau , before this term starts to drive @xmath20 to decrease for even larger values of @xmath184 . clearly the opposite behavior happens for @xmath127 , as seen in fig . [ figmuc ] ( left panel ) . of course , for extremely large @xmath184 values the opt correction term will become an unreasonably large perturbative correction and not very trustable , since higher @xmath2-order corrections are not considered at the opt first order . at low magnetic fields ( @xmath194 ) and when @xmath163 , a structure of intermediate phase transitions , where the vacuum expectation value of the chiral condensate can jump discontinuously from a value @xmath195 to another value @xmath196 , with @xmath197 is possible , as shown in fig . [ multiplept ] , where we show the normalized effective potential for both the ln and opt cases , @xmath198 , where @xmath199 note that the multiple transitions can happen in the ln case and also in the opt case . therefore , this is not an artifact of the mfa . the final transition is the actual chiral phase transition , where the system jumps from @xmath200 to @xmath201 . in the ln case , there is an intermediate ( nonchiral ) transition at a value of critical chemical potential given by @xmath202 when @xmath203 , while in the opt case ( for @xmath113 ) , this first transition happens at @xmath204 . the actual chiral phase transition happens at a larger value of chemical potential , given by @xmath205 in the ln case and by @xmath206 in the opt case . in fig . [ multisigma ] we show how the chiral order parameter changes with the chemical potential , also evidencing the intermediate transitions . an analogous structure of multiple phase transitions was first identified in ref . @xcite , for which besides including the perpendicular magnetic field component , it was also considered the inclusion of a parallel component for the magnetic field , which produces an enhancement of the zeeman energy term and an effective spin polarization of the system . we see here that , even in the absence of a parallel component of the magnetic field , we can also find a similar structure . it is quite surprising that no such structure has been reported before in the earlier literature of the gn model in a magnetic field . finally , it is interesting to note that within the opt the range of @xmath8 values for which the global minimum happens at @xmath207 is about one - third of the interval predicted by the ln approximation . in order to better understand these transitions we offer fig . [ dhva ] , which shows the many discontinuities associated with the de haas van alphen oscillations , which are produced when more landau levels are filled as @xmath8 increases . the numbers represent the highest landau level which for large @xmath0 , at @xmath15 , is given by @xmath208 or the nearest integer . the first transition occurs from @xmath209 to @xmath209 and is produced when @xmath210 , turning on all the heaviside step functions that appear in the free energy at @xmath15 . the subsequent transitions occur when a different @xmath211 is reached . for @xmath212 the only transition is from @xmath209 to @xmath209 , since this magnetic field ( which happened to be of the order of the gap value ) is high enough so that only the lll is always occupied . for @xmath213 there is a transition from @xmath209 to @xmath209 associated with @xmath214 and then a second associated with @xmath215 . exactly this type of behavior is observed in fig . [ multisigma ] for the case of @xmath216 , which can now be better understood . then , for @xmath217 we observe the type of behavior shown in fig . [ dhva ] with the transition that restores chiral symmetry happening at @xmath218 . for @xmath219 there are even more transitions since more levels can be filled . for example , at @xmath220 chiral symmetry is completely restored ( @xmath221 ) when @xmath222 . thus , in summary , for all @xmath165 values a first transition occurs due to a nonvanishing value of @xmath223 . if @xmath224 this is the only transition and only the lll is always filled , producing a smooth behavior for @xmath126 and @xmath20 . in the range @xmath225 there is also a first discontinuity in the value of @xmath126 due to a nonvanishing value of @xmath223 , but then , since @xmath165 is small , there can be subsequent discontinuities in @xmath126 due to the jumps among the integer values of @xmath211 , which accounts for the oscillations and discontinuities we have observed at @xmath3 for small @xmath165 . when @xmath127 and @xmath165 , the small chiral symmetry is restored only due to @xmath226 and the filling of higher landau levels only occurs after @xmath221 by increasing @xmath8 . appendix [ oscillations ] shows how these oscillations can be further understood by means of poisson s summation formula . finally , in the case of finite temperature and chemical potential , we have the effective potential as given by eq . ( [ veffhtmu ] ) . in this case , we search for points in the phase diagram in the plane @xmath227 , corresponding to either a first - order or a second - order phase transition . recall that in a first - order transition the effective potential develops different minima , @xmath228 , where one of them is a local minimum associated with metastability , while the other is a global minimum . these minima can get degenerate for some values of the parameters . for a given value of the magnetic field the first - order transition points in the @xmath227 plane can be determined from the condition of degeneracy of the minima of the effective potential , @xmath229 one of the minima is , in general , the trivial solution , @xmath201 , which then facilitates the determination of the first - order transition points . however , as noticed in the previous subsection , at low magnetic fields @xmath194 , other minima can emerge and , thus , at low magnetic fields the determination of the transition points must be done with care . in a first - order phase transition we then have that the minima @xmath230 change discontinuously at the transition point . on the other hand , the second - order phase transition critical points are found when the nontrivial minimum @xmath230 changes continuously and vanishes at the transition point . the point where the second - order transition line meets the first - order one , defines a tricritical point . the second - order and tricritical points are mostly easily found by using a landau expansion for the effective potential , which is valid for small values of the order parameter . this is the case close to a second - order or tricritical point . the landau s expansion ( for small @xmath51 ) for @xmath95 can be expressed in the general form @xmath231 where @xmath232 is a constant ( independent of the order parameter ) energy term . note that only even powers of @xmath51 are allowed due to the original chiral symmetry of the model . the coefficients @xmath233 , and @xmath234 appearing in eq . ( [ landau ] ) can be obtained , respectively , by a second , fourth , and sixth derivative of the effective potential expansion around @xmath172 . higher - order terms in the expansion ( [ landau ] ) can be verified to be much smaller than the first - order terms and can then be consistently neglected . in particular , note that a tricritical point can emerge whenever we have three phases coexisting simultaneously . from eq . ( [ landau ] ) , a second - order phase transition follows when the coefficient of the quadratic term vanishes ( @xmath235 ) and @xmath236 . a first - order transition happens for the case of @xmath237 . the tricritical point is found when both the quadratic and quartic coefficients in eq . ( [ landau ] ) vanish , @xmath238 ( with @xmath239 ) . thus , eq . ( [ landau ] ) offers a simple and immediate way for analyzing the phase structure of our model . for instance , to obtain @xmath135 at @xmath18 one only needs to consider eq . ( [ landau ] ) to order @xmath240 with @xmath241 to assure that the potential is bounded from below . then , the solution of @xmath242 sets the critical temperature . however , in order to use landau s expansion we must have @xmath95 in terms of @xmath51 , @xmath8 and @xmath7 only ( apart from @xmath0 and the scale @xmath109 , of course ) . in principle , this can be done by using the pms relation , eq . ( [ pmsselfconsistent ] ) . even though at finite @xmath0 , @xmath50 depends on @xmath51 in a highly nonlinear way , eq . ( [ pmsselfconsistent ] ) can be easily solved numerically by iteration in a very efficient way ( see , e.g. , ref . for example , at the first iteration , the use of the approximate pms solution obtained by using the large-@xmath0 solution @xmath156 within higher - order @xmath243 terms , @xmath244\bigr\|_{\eta = \sigma_c}\ ; , \label{pmsiteration}\ ] ] is already able to produce results for the tricritical points within a less than @xmath245 difference with respect to a full numerical calculation . furthermore , if the second term inside the square brackets in eq . ( [ pmsiteration ] ) involving the variation of the @xmath246 term with respect to the chemical potential is much smaller than the first term , it can be neglected and this can make the pms calculation procedure much simpler . in all cases we have checked the applicability of the use of this simplified form compared with the complete expression ( [ pmsiteration ] ) and used it whenever possible to simplify the numerical calculations . following this procedure , we obtain the tricritical point as a function of the magnetic field . in fig . [ mutricfig ] we give the results for the chemical potential as a function of the magnetic field at the tricritical point , for the cases of negative and positive couplings , for both the ln and opt ( at @xmath113 ) cases . interestingly enough , the ln results for @xmath3 ( right panel ) display exactly the same qualitative behavior found in ref . @xcite , where the mfa was applied to the three - flavor njl model in 3 + 1 dimensions . note again from this figure the effect of the imc , similar to the one seen in fig . [ figmuc ] . the figure shows that @xmath247 in the opt case only decreases with @xmath9 . again , as in the @xmath15 case discussed in sec . [ t0section ] , this result could be a sign of the importance brought in by the opt @xmath248 type of corrections , which start to play an important role in this region of intermediate to large charge asymmetries . for @xmath249 the right panel of fig . [ mutricfig ] shows that both the opt and the ln approximation predict that @xmath250 always increase with @xmath9 and that the opt predicted values are always higher than the ln ones . in fig . [ ttricfig ] we give the results for the temperature at the tricritical point as a function of the magnetic field , for the cases of negative ( right panel ) and positive couplings ( left panel ) , for both the ln and opt ( at @xmath113 ) cases . let us start by discussing the case of negative coupling at vanishing magnetic field , where the figure shows that the ln predicts @xmath251 , while the opt predicts @xmath252 . as we have already discussed ( see also refs . @xcite ) this ln result for @xmath124 can be shown to be wrong due to universality arguments , while the opt predicted values for finite @xmath0 are within the range estimated ( but not pinpointed ) by monte carlo simulations @xcite . despite these important quantitative differences , both approximations show that @xmath253 increases with the magnetic field . next , in fig . [ phasediagram ] , we show the complete phase diagram for the ln and opt cases for representative values of the magnetic field . note that in the ln case for @xmath124 and negative coupling , there is only a second - order transition line in the @xmath254 phase diagram , with the exception of the pair @xmath15 and @xmath255 , which correspond to a first - order transition point . as shown in ref . @xcite , it is only by including beyond mean - field effects that a first - order transition line ( along with the tricritical point ) emerges , in agreement with the expectations based on the results for the gn model in 1 + 1 dimensions and also for the njl model in 3 + 1 dimensions . note , however , that for nonvanishing magnetic fields , a tricritical point is produced even in the ln case . for positive couplings , recall that a chiral phase transition is only possible for nonvanishing magnetic fields @xcite . figure [ phasediagram ] shows that for both a negative or a positive coupling the presence of a magnetic field always increases the size of the first - order transitions . in the opt case this increase is even more pronounced since the term @xmath256 enhances this type of transitions when @xmath3 and the @xmath17 correction acts as an attractive vector term @xcite . in both the opt and ln cases and for any sign for the coupling , the csb region tends to get larger as @xmath9 grows . note that there is magnetic catalysis in @xmath135 and inverse catalysis in @xmath20 for @xmath24 , but only catalysis for @xmath4 in the opt case . for the ln case and @xmath3 , the inverse magnetic catalysis only happens untill some value of @xmath9 and then there is only catalysis beyond that value of magnetic field . so for @xmath257 or so , the csb region in the ln case will become always larger than in the opt when @xmath24 , while it is the opposite when @xmath127 , where @xmath1 acts as a repulsive vector term , which competes with the effect of the magnetic field by enhancing more the csb region in the opt case than in the ln case . in this work we have analyzed the effects of a nonvanishing constant magnetic field ( applied perpendicularly to the plane of the system ) on the phase structure of the massless discrete ( 2 + 1)-dimensional gn model , including contributions which go beyond the ln ( or mean field ) approximation . here we have used the opt method , which has already been successfully used before to study the properties of this model in the absence of a magnetic field . both the cases of positive and negative four - fermion coupling have been studied . we have produced some novel results concerning the phase structure of the model in the presence of a magnetic field . for negative couplings , we have shown that at low magnetic fields ( @xmath258 ) a rich structure of phases can emerge . in this case , it is possible to have intermediate transitions to nonvanishing values of the chiral vacuum expectation value . these transitions happen in the ln case , something that has not been previously noted in the literature and remains also when including corrections beyond the ln approximation , as we have shown by using the opt method . therefore , these intermediate transitions are not an artifact of the ln approximation . we have also traced the origin of these intermediate transitions as being a consequence of the magnetic de haas van alphen oscillations that arise at low values of the magnetic field , @xmath259 and for negative values of the coupling constant . as for the effect of the magnetic field on the phase structure of the model , we have shown two distinct effects depending on the sign of the coupling constant . for either positive or negative couplings , we still have an enhancement of the chiral - broken - symmetry region , as expected in general from the magnetic catalysis effect . however , when the coupling is negative , beyond some value of the magnetic field the chiral broken region is always smaller in the opt case than in the ln , while for positive coupling the reverse is observed , with the opt always producing a larger broken symmetry region . the tricritical points tend to be also enhanced in general by the presence of the magnetic field , with the results obtained in the opt larger than in the ln case . the exception is the value of the chemical potential at the tricritical point , @xmath260 when the coupling is negative . when @xmath3 , in the ln case @xmath260 tends to be strongly suppressed by the magnetic field initially , until for @xmath261 it turns again to be enhanced by the magnetic field . this decrease of the critical chemical potential with the magnetic field , the inverse magnetic catalysis effect , has some similarity with the phenomenon seen in the njl model and discussed in details recently in ref . but in our case it originates from the opt @xmath171 corrections beyond ln , as explained in sec . [ t0section ] . in the opt case @xmath260 continues to decrease for very large magnetic fields . thus , the inverse magnetic catalysis remains unsuppressed even for large values of the magnetic field in the opt context , which is opposite to what is seen in the ln case . note also that inverse magnetic catalysis is seen to operate only on the critical chemical potential values , while the critical temperature still shows only the standard magnetic catalysis , always increasing with @xmath9 . this same trend also applies to the coexistence chemical potential when @xmath15 , where , for @xmath24 , the opt shows an inverse magnetic catalysis effect even for large magnetic fields , while in the ln case , the chemical potential only decreases for relatively small values of the magnetic field and grows for @xmath261 . at @xmath18 , the critical temperature is seen only to increase with the magnetic field in both the opt and ln cases , independently of the sign of the coupling constant . the stability of the order-@xmath39 results for the same model , at @xmath124 , has been addressed in ref . @xcite and the outcome of that investigation allows us to believe that our present results , at @xmath262 , should also be stable against the inclusion higher - order corrections . it is tempting to compare our results with recent lattice results for the qcd chiral crossover temperature as function of @xmath9 . the first lattice studies @xcite considered two quark flavors , with high values of pion masses ( @xmath263 ) , and have shown that the critical temperature should increase with @xmath9 . however , an improved lattice simulation @xcite , which considered 2 + 1 quark flavors at physical pion mass values ( @xmath264 ) , together with an extrapolation to the continuum , predicted that the critical temperature should decrease with @xmath9 . since then , most models have tried to reproduce these lattice results , showing , however , that the critical temperature only increases with the magnetic field . since most model results were obtained within the ln / mfa , one may wonder if the discrepancy could not be resolved by going beyond this approximation . our results indicate that this may not be sufficient and that other effects besides going beyond the ln approximation are required . we hope that our findings will give further insights in applications that employ four - fermion models in the description of planar condensed matter systems , which we intend to further explore in the future . let us derive here the momentum integrals appearing in the expression for the effective potential ( [ vlde1ddim ] ) and then give the @xmath102 , @xmath72 , integrals eqs . ( [ int1 ] ) , ( [ int2 ] ) and ( [ int3 ] ) . using the replacements ( [ intp ] ) , @xmath265 and @xmath67 , with @xmath266 , @xmath63 , are the matsubara frequencies for fermions and @xmath64 labels the lls . the integral @xmath246 is defined as @xmath267 \label{intln}\end{aligned}\ ] ] where @xmath76 and @xmath268 . performing the matsubara sum one gets @xmath269 + t \ln [ 1 + e^{-(e_j-\mu)/t } ] \right \}\;. \label{inti1}\ ] ] in the same way , we have that @xmath270 } - \frac{1}{e_j [ 1 + e^{(e_j-\mu)/t } ] } \right \}\ ; , \label{inti2}\end{aligned}\ ] ] while the last momentum integral remaining that we need is @xmath271 \nonumber \\ & = & \frac{eb}{8\pi}\sum_{j=0}^{\infty}\alpha_j\left [ \frac{1}{e^{\left ( e_j-\mu \right ) /t}+1}-\frac{1}{e^{\left ( e_j+\mu \right ) /t}+1}\right]\ , . \label{inti3}\end{aligned}\ ] ] there is a very convenient trick to perform the sum over landau levels for the @xmath272 independent terms which can be expressed in a closed form by means of riemann hurwitz zeta functions @xcite . for example , consider the @xmath7 @xmath8 independent term on the right - hand side of eq . ( [ inti1 ] ) . by adding and subtracting a lowest landau energy level term , @xmath273 , to it one can write @xmath274^{1/2 } \,\,.\ ] ] the infinite sum can be related to the riemann hurwitz zeta function eq . ( [ zetafunc ] ) , yielding @xmath275 the same technique , when applied to the @xmath7 and @xmath8 independent term of eq . ( [ inti2 ] ) , gives @xmath276 the same types of manipulations can be applied to divergent terms , as discussed in refs . @xcite . it is also useful to have the limiting cases for the functions @xmath102 when @xmath15 and/or @xmath18 . taking the @xmath15 limit in eqs . ( [ inti1 ] ) , ( [ inti2 ] ) and ( [ inti3 ] ) , we obtain @xmath277 where in the above equations @xmath278 and @xmath279 means the integer part of @xmath280 . taking the @xmath18 limit in eqs . ( [ inti1 ] ) , ( [ inti2 ] ) and ( [ inti3 ] ) , we obtain @xmath281 while for @xmath15 and @xmath18 , we obtain @xmath282 it is instructive to analyze the origin of the oscillating behavior at low @xmath9 found in the case of @xmath15 and @xmath3 in sec . [ t0section ] . since the terms @xmath158 and @xmath153 are directly derived from @xmath246 , it is enough for our purposes here to analyze the low-@xmath9 behavior of only @xmath246 . the @xmath101 dependent part of @xmath246 at @xmath15 , from eq . ( [ i1t0 ] ) , is given by @xmath286 \theta(\mu-\|\eta\| ) \nonumber \\ & + & \frac{eb}{4\pi^{2 } } \mu \sum_{n=1}^{\infty}\frac{1}{n } \left\{p_{2}\left(\frac{n\pi}{eb}\mu^{2}\right)\ , \cos\left[\frac{n\pi}{eb}(\eta^{2}- \mu^{2})\right ] \right . \nonumber \\ & + & \left . q_{2}\left(\frac{n\pi}{eb}\mu^{2}\right)\ , \sin\left[\frac{n\pi}{eb}(\eta^{2}-\mu^{2})\right]\right\ } \theta(\mu-\|\eta\| ) \ , , \label{i1osc2}\end{aligned}\ ] ] @xmath289 \cos(x)+\left[s_{2}(x)-\frac{1}{2}\right]\sin(x)\right\}\ , , \nonumber \\ & & p_{2}(x)=\sqrt{\frac{2\pi}{x}}\left\{\left[c_{2}(x)-\frac{1}{2}\right ] \sin(x)-\left[s_{2}(x)-\frac{1}{2}\right]\cos(x)\right\}\,,\end{aligned}\ ] ] in eq . ( [ i1osc2 ] ) the oscillatory character of the effective potential at low-@xmath9 becomes explicit , as the origin of the de haas van alphen oscillations in magnetized systems . for @xmath21 in the symmetry broken phase , as far @xmath293 and since @xmath46 is determined from the pms solution ( [ baretat0mu ] ) , @xmath294 , we have that for @xmath295 all quantities depending on @xmath246 will exhibit an oscillatory behavior , with the different basic functions in eq . ( [ i1osc2 ] ) of periods @xmath296 characteristic of de haas van alphen oscillations . this determines the low-@xmath9 behavior seen in figs . [ figmuc ] and [ figmuclowh ] , as well the multiple minima structure exhibited by the effective potential for values of magnetic field @xmath259 and shown in the example given by fig . [ multiplept ] . when @xmath4 , we know that in the absence of a magnetic field there is no chiral symmetry breaking . however , when @xmath297 , magnetic catalysis induces chiral symmetry breaking and , as first shown in ref . @xcite , in the ln case , the vacuum expectation value for @xmath43 is @xmath298 , but we have that @xmath299 and then , for @xmath300 , the magnetic oscillations become highly suppressed when @xmath4 . this is why we do not see ( or hardly can see any ) oscillations in this case . m.b.p . and r.o.r . are partially supported by conselho nacional de desenvolvimento cientfico e tecnolgico ( cnpq ) . is also partially supported by a research grant from fundao carlos chagas filho de amparo pesquisa do estado do rio de janeiro ( faperj ) and m.b.p . is also partially supported by fundao de amparo pesquisa e inovao do estado de santa catarina ( fapesc ) . y. nambu and g. jona - lasinio , phys . 122 * , 345 ( 1961 ) ; phys . 124 * , 246 ( 1961 ) . d. j. gross and a. neveu , phys . d * 10 * , 3235 ( 1974 ) . u. wolff , phys . b * 157 * , 303 ( 1985 ) ; k. g. klimenko , z. phys . * c37 * , 457 ( 1988 ) ; b. rosenstein , b. j. warr and s. h. park , phys . d * 39 * , 3088 ( 1989 ) ; a. barducci , r. casalbuoni , m. modugno , g. pettini and r. gatto , phys . d * 51 * , 3042 ( 1995 ) ; f. c. khanna , a. p. c. malbouisson , j. m. c. malbouisson and a. e. santana , europhys . lett . * 92 * , 1101 ( 2010 ) . d. v. khveshchenko , phys . * 87 * , 206401 ( 2001 ) ; e. v. gorbar , v. p. gusynin , v. a. miransky and i. a. shovkovy , phys . b * 66 * , 045108 ( 2002 ) ; h. leal and d. v. khveshchenko , nucl . phys . b * 687 * , 323 ( 2004 ) . g. w. semenoff , i. a. shovkovy and l. c. r. wijewardhana , mod . a * 13 * , 1143 ( 1998 ) ; k. krishana , n. p. ong , y. zhang , et al . 1999 , lett . * 82 * , 5108 ( 1999 ) ; d. v. khveshchenko , w. f. shively , phys . b * 73 * , 115104 ( 2006 ) . k. fukushima , d. e. kharzeev and h. j. warringa , phys . d * 78 * , 074033 ( 2008 ) ; d. e. kharzeev , l.d . mclerran and h. j. warringa , nucl . a * 803 * , 227 ( 2008 ) ; d. e. kharzeev and h. j. warringa , phys . d * 80 * , 0304028 ( 2009 ) ; d. e. kharzeev , nucl . a * 830 * , 543c ( 2009 ) . k. g. klimenko , theor . phys . * 90 * , 1 ( 1992 ) [ teor . mat . fiz . * 90 * , 3 ( 1992 ) ] ; z. phys . c * 54 * , 323 ( 1992 ) ; theor . * 89 * , 1161 ( 1992 ) [ teor . mat . * 89 * , 211 ( 1991 ) ] . v. p. gusynin , v. a. miransky and i. a. shovkovy , phys . lett . * 73 * , 3499 ( 1994 ) [ erratum - ibid . * 76 * , 1005 ( 1996 ) ] ; phys . d * 52 * , 4718 ( 1995 ) . d. c. duarte , r. l. s. farias and r. o. ramos , phys . d * 84 * , 083525 ( 2011 ) ; a. ayala , m. loewe , j. c. rojas and c. villavicencio , phys . d * 86 * , 076006 ( 2012 ) . a. okopinska , phys . d * 35 * , 1835 ( 1987 ) ; m. moshe and a. duncan , phys . b215 * , 352 ( 1988 ) ; s. chiku and t. hatsuda , phys . d * 58 * , 076001 ( 1998 ) . r. seznec and j. zinn - justin , j. math . * 20 * , 1398 ( 1979 ) ; j. c. le guillou and j. zinn - justin , ann . phys . * 147 * , 57 ( 1983 ) ; v. i. yukalov , teor . * 28 * , 92 ( 1976 ) ; w. e. caswell , ann . phys . ( n.y ) * 123 * , 153 ( 1979 ) ; i. g. halliday and p. suranyi , phys . lett . * b85 * , 421 ( 1979 ) ; j. killinbeck , j. phys . * a14 * , 1005 ( 1981 ) ; r. p. feynman and h. kleinert , phys . a * 34 * , 5080 ( 1986 ) ; h. f. jones and m. moshe , phys . b234 * , 492 ( 1990 ) ; a. neveu , nucl . phys . ( proc . suppl . ) * b18 * , 242 ( 1990 ) ; v. yukalov , j. math . phys * 32 * , 1235 ( 1991 ) ; c. m. bender et al . , phys . d * 45 * , 1248 ( 1992 ) ; k. g. klimenko , z. phys . * c60 * , 677 ( 1993 ) ; a.n . sissakian , i. l. solovtsov and o. p. solovtsova , phys . lett . * b321 * , 381 ( 1994 ) ; h. kleinert , phys . rev . d * 57 * , 2264 ( 1998 ) ; phys . b434 * , 74 ( 1998 ) ; for a review , see h. kleinert and v. schulte - frohlinde , _ critical properties of @xmath301-theories _ , chap . 19 ( world scientific , singapure 2001 ) . f. f. de souza cruz , m. b. pinto and r. o. ramos , phys . b * 64 * , 014515 ( 2001 ) ; j. -l . kneur , m. b. pinto and r. o. ramos , phys . a * 68 * , 043615 ( 2003 ) ; phys . lett . * 89 * , 210403 ( 2002 ) ; h. kleinert , mod . b17 * , 1011 ( 2003 ) ; b. kastening , phys . rev . a * 68 * , 061601 ( 2003 ) ; phys.rev . a * 69 * ( 2004 ) 043613 ; j. -l . kneur , a. neveu and m. b. pinto , phys . a * 69 * , 053624 ( 2004 ) ; j. -l . kneur and m. b. pinto , phys . rev . a * 71 * , 033613 ( 2005 ) . h. caldas , j. -l . kneur , m. b. pinto and r. o. ramos , phys . b * 77 * , 205109 ( 2008 ) . j. o. andersen , l. e. leganger , m. strickland and n. su , jhep * 1108 * , 053 ( 2011 ) ; j. o. andersen , m. strickland and n. su , jhep * 1008 * , 113 ( 2010 ) . kneur and a. neveu , phys . d * 81 * , 125012 ( 2010 ) . kneur and a. neveu , phys . d * 85 * , 014005 ( 2012 ) . kneur and a. neveu , arxiv:1305.6910 . j. -l . kneur , m. b. pinto , r. o. ramos and e. staudt , phys . d * 76 * , 045020 ( 2007 ) . j. b. kogut and c. g. strouthos , phys . d * 63 * , 054502 ( 2001 ) . f. preis , a. rebhan and a. schmitt , lect . notes phys . * 871 * , 51 ( 2013 ) ; f. preis , a. rebhan and a. schmitt , jhep * 1103 * , 033 ( 2011 ) . t. w. appelquist , m. j. bowick , d. karabali and l. c. r. wijewardhana , phys . d * 33 * , 3704 ( 1986 ) . s. coleman , _ aspects of symmetry _ ( cambridge university press , cambridge , 1985 ) . j. -l . kneur , m. b. pinto and r. o. ramos , phys . d * 74 * , 125020 ( 2006 ) . m. b. pinto and r. o. ramos , phys . d * 60 * , 105005 ( 1999 ) ; phys . rev . d * 61 * , 125016 ( 2000 ) ; r. l. s. farias , g. krein and r. o. ramos , phys . d * 78 * , 065046 ( 2008 ) . p. m. stevenson , phys . d * 23 * , 2916 ( 1981 ) . r. g. root , phys . d * 11 * , 831 ( 1975 ) . e. elizalde , a. d. odintsov and a. romeo , _ zeta regularization techniques with applications _ ( world scientific , singapore , 1994 ) . a. s. vshivtsev , k. g. klimenko and b. v. magnitsky , theor . * 106 * , 319 ( 1996 ) [ teor . mat . fiz . * 106 * , 390 ( 1996 ) ] . a. chodos , k. everding and d. a. owen , phys . d * 42 * , 2881 ( 1990 ) . s. s. avancini , d. p. menezes , m. b. pinto and c. providencia , phys . d * 85 * , 091901 ( 2012 ) . g. n. ferrari , a. f. garcia and m. b. pinto , phys . d * 86 * , 096005 ( 2012 ) . d. ebert , k.g . klimenko , m.a . vdovichenko and a.s . vshivtsev , phys . d * 61 * , 025005 ( 2000 ) . r. o. ramos and p. h. a. manso , phys . d * 87 * , 125014 ( 2013 ) . k. fukushima , phys . d * 78 * , 114019 ( 2008 ) . h. caldas and r. o. ramos , phys . b * 80 * , 115428 ( 2009 ) . m. delia , s. mukherjee and f. sanfilippo , phys . d * 82 * , 051501 ( 2010 ) . g. s. bali , f. bruckmann , g. endrodi , z. fodor , s. d. katz , s. krieg , a. schafer and k. k. szabo , jhep * 1202 * , 044 ( 2012 ) ; g. s. bali , f. bruckmann , g. endrodi , z. fodor , s. d. katz and a. schafer , phys . rev . d * 86 * , 071502 ( 2012 ) . e. w. weisstein , _ poisson sum formula _ from http://mathworld.wolfram.com/poissonsumformula.html ( mathworld a wolfram web resource . )	the phase diagram and thermodynamic properties of the ( 2 + 1)-dimensional gross neveu model are studied in the presence of a constant magnetic field . the optimized perturbation theory ( opt ) is used to obtain results going beyond the large-@xmath0 approximation . the free energy and the complete phase diagram of the model , in terms of temperature , chemical potential and magnetic field are obtained and studied in details . we find that some of the main qualitative changes induced by the opt finite @xmath0 corrections concern the region of intermediate to high chemical potentials where this approximation adds a term proportional to @xmath1 to the free energy . then , depending on the sign of @xmath2 ( relative to the critical coupling ) , and magnitude of the magnetic field , we observe a weakening ( when @xmath3 ) or enhancement ( when @xmath4 ) of the chiral broken region in the magnetized fermionic system . by comparing the results from the opt and the large-@xmath0 approximation , we conclude that finite @xmath0 effects favor the phenomenon of inverse magnetic catalysis when the coupling constant is negative . we show that with the opt the value of the coexistence chemical potential at vanishing temperature tends to decrease for large values of the magnetic field . this is opposite to what is seen in the large-@xmath0 approximation , where for large magnetic fields the coexistence chemical potential starts again to increase . likewise , at finite temperature , the value of the chemical potential at the tricritical point also decreases with the magnetic field in the opt case . consequently , the shape of the phase diagrams predicted by the opt and by the large-@xmath0 approximation look very different in the presence of high magnetic fields . finally , for small values of magnetic field and temperature , we identify the presence of possible intermediate nonchiral phase transitions when varying the chemical potential . we show that these phenomena are not an artifact of the large-@xmath0 approximation and that they also occur within the opt framework . these intermediate transitions are interpreted to be a consequence of the de haas van alphen oscillations . we also explain why this type of phenomenon can happen in general for negative couplings but not for positive couplings .
the first challenging requirement to develop semiconductor ( sc ) spintronics@xcite _ i.e. _ using both carrier charge and spin in electronic devices consists in injecting spin polarized electrons in the conduction band of a sc at room temperature . scs should be further compatible with silicon mainstream technology for implementation in microelectronics making silicon , germanium and their alloys among the best candidates.@xcite in si , due to low spin - orbit coupling , very long spin diffusion lengths were predicted and measured experimentally.@xcite germanium exhibits the same crystal inversion symmetry as si , a low concentration of nuclear spins but higher carrier mobility and larger spin - orbit coupling which should allow in principle spin manipulation by electric fields such as the rashba field.@xcite so far in order to perform spin injection from a ferromagnetic metal ( fm ) into si or ge , one needs to overcome at least three major obstacles : _ ( i ) _ the conductivity mismatch which requires the use of a highly - resistive spin - conserving interface between the fm and the sc,@xcite _ ( ii ) _ the fermi level pinning at the sc surface due to the presence of a high density of interface states and the interface spin flips which are generally associated@xcite and finally _ ( iii ) _ the presence of random magnetic stray fields created by surface magnetic charges at rough interface@xcite around which the electrically injected spins are precessing and partly lost by decoherence . in this work , we have inserted a thin mgo tunnel barrier between ge and the cofeb ferromagnetic electrode in order to : _ ( i ) _ circumvent the conductivity mismatch and _ ( ii ) _ partly alleviate fermi level pinning by strongly reducing the interface states density@xcite which leads to a modest schottky barrier height at the mgo/_n_-ge interface . we have then investigated the spin injection mechanisms using the so called three - terminal device.@xcite in this geometry , the same ferromagnetic electrode is used for spin injection and detection . this three - terminal device used in non - local geometry represents a simple and unique tool to probe spin accumulation both into interface states and in the sc channel.@xcite in particular , we could measure spin injection in the silicon and germanium conduction bands at room temperature.@xcite + the spin hall angle ( @xmath3 , ratio between the transverse spin current density and the longitudinal charge current density ) @xcite is a key material parameter to develop new kinds of devices based on the spin hall effect ( she ) . the she is the conversion of a charge current into a spin current via the spin - orbit interaction . conversely the inverse spin hall effect ( ishe ) is the conversion of a spin current into a charge current . several methods have been developed to determine quantitatively @xmath3 : pure magnetotransport measurements on lateral spin valves ( lsv),@xcite ferromagnetic resonance ( fmr ) along with spin pumping ( sp - fmr ) @xcite and spin torque fmr ( st - fmr ) @xcite on ferromagnetic / non - magnetic bilayers ( fm / n ) . recently @xmath3 could be estimated in @xmath1 and @xmath4-gaas @xcite as well as in @xmath4-si @xcite by spin pumping and inverse spin hall effect . the precession of the fm layer in direct contact with the sc has been excited by microwaves which pumps a spin current into the sc . the spin current was then detected by inverse spin hall effect . in that case , the interface resistance to overcome the conductivity mismatch issue was given by the reminiscent schottky barrier at the fm / sc interface . here we have similarly used a combined sp - fmr method to study inverse spin hall effect in germanium . in the first section , we describe the sample preparation and the experimental techniques . in the second section , the phenomenological models for the ferromagnetic resonance and electromotive force are presented . finally the experimental results are shown and discussed in sections 3 and 4 respectively . in particular , we propose a microscopic model based on the presence of localized states at the mgo / ge interface to explain the spin pumping mechanism in our system . we finally discuss about the influence of the mgo tunnel barrier and sample annealing on the ishe signal . the multi - terminal device we initially used for electrical spin injection , detection and manipulation@xcite is made of a full stack ta(5 nm)/cofeb(5 nm)/mgo(3 nm ) grown by sputtering on a 40 nm - thick germanium film on insulator ( goi).@xcite goi wafers are made of a si @xmath4 + degenerate substrate and a 100 nm - thick sio@xmath5 layer ( box ) . they were fabricated using the smart cut@xmath6 process and ge epitaxial wafers.@xcite the transferred 40 nm - thick ge film was @xmath1-type doped in two steps : a first step ( phosphorus , @xmath7 @xmath8 , 40 kev , annealed for 1h at 550@xmath9c ) that provided uniform doping in the range of 10@xmath10 @xmath11 , and a second step ( phosphorus , @xmath12 @xmath8 , 3 kev , annealed for 10 s at 550@xmath9c ) that increased surface @xmath1 + doping to the vicinity of 10@xmath13 @xmath11 . the thickness of the _ _ n__+-doped layer is estimated to be 10 nm . the goi surface was finally capped with amorphous sio@xmath5 to prevent ge from surface oxidation . the tunnel barrier and ferromagnetic electrode were then fabricated from magnesium ( mg , 1.1 nm ) and cobalt - iron - boron ( co@xmath14fe@xmath15b@xmath15 , 5 nm ) layers deposited by conventional dc magnetron sputtering onto germanium ( ge ) after removing the sio@xmath5 capping layer using hydrofluoric acid and de - ionized water . the deposition rates were respectively of 0.02 nm.s@xmath16 and 0.03 nm.s@xmath16 at an argon pressure of @xmath17 mbar . the base pressure was @xmath18 mbar . all the depositions were performed at room temperature . the oxidation of the insulating barrier was performed by plasma oxidation , exposing the mg metallic layer to a 30 seconds radio - frequency oxygen plasma at a pressure of @xmath19 mbar and a radio - frequency power of 100 w. three successive mg deposition plus oxidation steps were achieved ( [ mg 1.1 / oxidation]@xmath20 ) to grow a 3.3 nm thick mgo layer . the sample annealings were performed under 10@xmath21 mbar at 300@xmath9c for 90 minutes . the ferromagnetic layer is then capped with 5 nm of ta to prevent oxidation . after depositing the spin injector , samples have been processed using standard optical lithography . in a first step , we define the ferromagnetic electrode ( @xmath22 @xmath23m@xmath24 ) and ohmic contacts ( @xmath25 @xmath23m@xmath24 on ge and @xmath26 @xmath23m@xmath24 on top of the ferromagnetic electrode ) . in a second step , the germanium channel is etched down to the box to form a mesa of 1070 @xmath23 m long and 420 @xmath23 m large . finally soft argon etching is used to remove the top 10 nm - thick _ _ n__+-doped germanium layer . the whole device is shown in fig . [ fig01ab](a ) . in order to test the influence of the mgo tunnel barrier ( resp . sample annealing ) , similar devices without mgo ( resp . without annealing ) were processed and studied . ferromagnetic resonance and inverse spin hall effect measurements were performed in a brker esp300e @xmath27-band cw spectrometer with a cylindrical brker er 4118x - ms5 cavity . the measurement geometry is depicted in fig . [ fig01ab](b ) . complementary measurements of fmr lines at different frequencies between 2 and 24 ghz were performed in a stripe - line vector network analyzer system . and @xmath28 . @xmath29 is the radiofrequency magnetic field . the thickness of each layer is given in nanometers between parenthesis . ( b ) drawing of the device inserted into the cylindrical _ x_-band electron paramagnetic resonance ( epr ) cavity.,width=15 ] fig . [ fig1](a ) shows a schematic drawing of the reference sample and the definition of the magnetization and external magnetic field polar angles , @xmath30 and @xmath28 , respectively . the reference sample is made of a single cofeb layer grown on sio@xmath5 as follows : ta(5nm)/cofeb(5nm)/mgo(3.3nm)//sio@xmath5 . we have inserted a thin mgo oxide layer between cofeb and sio@xmath5 in order to make a comparison with the ta(5nm)/cofeb(5nm)/mgo(3.3nm)/ge system studied in the next sections . . electrical contacts are directly made on top of the metallic layer . the fmr spectrum ( b ) and the electromotive force ( c ) have been measured simultaneously by placing the sample in the epr cavity . the red line in ( c ) is the fit according to eq . ( [ eqvamr ] ) . symmetric and asymmetric contributions are also shown separately in blue dotted curves.,width=6 ] from fmr measurements ( fig . [ fig1](b ) ) , we can determine the peak - to - peak linewidth and the resonance field . by sweeping the external magnetic field * h * under a microwave excitation of frequency @xmath31 , the resonance condition is achieved in a ferromagnetic film when:@xcite @xmath32}\ ] ] where @xmath33 is the precession angular frequency , @xmath34 is the gyromagnetic ratio with the land factor @xmath35 , @xmath36 is the reduced planck constant , @xmath37 is the bohr magnetron and @xmath38 is the saturation magnetization of the ferromagnetic film . the second order partial derivatives of the free energy density are evaluated at the equilibrium angles @xmath39 and @xmath40 of the magnetization * m * for which : @xmath41 and @xmath42 . the shape anisotropy in a ferromagnetic polycristalline film ( @xmath43 ) is usually much larger than in - plane crystalline anisotropy . by recording fmr spectra with the dc magnetic field in the film plane at different azimuthal angles ( not shown ) , we could demonstrate that in - plane anisotropy in the cofeb electrodes used in this work is indeed negligible with respect to shape anisotropy . thus we can consider that the free energy density is given by : @xmath44 where the first term is the zeeman energy and the second one accounts for shape anisotropy and any other perpendicular uniaxial anisotropy @xmath45 . the effective saturation magnetization @xmath46 is thus defined as : @xmath47 . by minimizing numerically @xmath48 we can obtain the magnetization equilibrium angles : @xmath40 and @xmath39 . the resonance field is then given by combining eq . ( [ eq1 ] ) and ( [ eqf ] ) , it can be plotted as a function of the external static field orientation ( @xmath30 ) and the excitation frequency as shown in fig . [ fig2dispersion ] . @xmath46 and @xmath35 are extracted from the out - of - plane ( oop ) angular dependence of the resonance field using a least square fit ( as shown for instance in fig . [ fig4oop](a , d ) ) . ) and ( [ eqf ] ) by applying the external magnetic field at different polar angles @xmath30 . curves are shown every 2.5@xmath49 between * h * parallel ( @xmath50 ) and perpendicular ( @xmath51 ) to the film plane . the intercept of each curve with the dotted line ( @xmath52 ghz ) yields the resonance field @xmath53.,width=7 ] analytical solutions of @xmath53 can be obtained for the parallel , @xmath50 , and perpendicular , @xmath51 , cases . in the parallel case , we find : @xmath54 the frequency dependence of the fmr linewidth also allows calculating the gilbert damping constant @xmath55 using the following expression:@xcite @xmath56 where the peak - to - peak linewidth , @xmath57,is measured when @xmath58 is applied parallel to the film plane ( @xmath59 ) . the @xmath60 term accounts for the frequency - independent contributions due to inhomogeneities in the ferromagnetic layer and @xmath61 is the fmr linewidth due to the gilbert damping . as shown experimentally in section [ sec : results ] , @xmath62 at high frequency and we systematically neglect this contribution to the fmr linewidth . moreover , the oop angular dependence of the peak - to - peak linewidth @xmath57 at a given frequency can be written : @xmath63 where the gilbert contribution can be calculated from:@xcite @xmath64 , and @xmath65 is the angular dispersion of the perpendicular anisotropy and demagnetizing field ( @xmath66 ) due to inhomogeneities in the fm layer . we show in the following that @xmath55 and @xmath67 can be extracted from the frequency and oop angular dependences of @xmath57 . the electromotive force generated in the ferromagnetic layer and shown in fig . [ fig1](c ) is simultaneously recorded with the fmr spectrum . the origins of this electromotive force are the anisotropic magnetoresistance ( amr ) @xcite and anomalous hall effect which manifest at the resonance field . at the resonance field , the precessing magnetization induces a time varying resistivity of the ferromagnetic layer which combines with the radiofrequency induced currents ( along @xmath68 in fig . [ fig01ab](b ) ) to produce a dc voltage . the radiofrequency currents are likely produced within the metallic layer by the non - vanishing radiofrequency electric field in the cavity at the sample level . in our set - up geometry , the electromotive force is measured as a voltage along @xmath69 ( see fig . [ fig01ab](b ) ) and thus in the transverse hall geometry . therefore we measure the planar hall effect ( phe ) and the anomalous hall effect ( ahe ) in cofeb . it was proposed and shown by azevedo _ _ @xcite and harder _ et al . _ @xcite that the resulting voltage is well described by both a symmetric and an asymmetric contributions . in the cofeb reference film , the asymmetric component is dominant . the electromotive force can be written as ( see appendix a ) : @xmath70 where @xmath71 ( resp . @xmath72 ) is the amplitude of the symmetric ( resp . asymmetric ) contribution to the electromotive force . we have taken into account a non - resonant offset voltage @xmath73 , @xmath53 is the resonance field and @xmath74 . @xmath75 where @xmath76 is the phase shift between the radiofrequency current and the magnetization.@xcite the symmetric and asymmetric contributions as well as the offset voltage are proportional to the microwave power . it means that @xmath73 , @xmath71 , and @xmath72 are proportional to @xmath77 , where @xmath29 is the microwave magnetic field strength . here we consider the device shown in fig . [ fig01ab ] . under radiofrequency excitation the magnetization precession of the ferromagnetic layer pumps spins to the non - magnetic germanium layer ( n ) and the corresponding spin current generates an electric field in ge due to ishe : @xmath78 . @xmath79 is the spin - current density along @xmath80 and @xmath81 its spin polarization vector . this electric field @xmath82 is converted into a voltage @xmath83 between both ends of the ge channel.@xcite in the case of germanium we overcome the conductivity mismatch issue by inserting a thin mgo tunnel barrier ( i ) between ge and cofeb . this additional interface resistance allows for spin accumulation in the germanium conduction band . as a consequence of spin pumping , the damping constant , @xmath84 , is enhanced with respect to the one of the reference sample , @xmath85 . the real part of the tunnel spin mixing conductance , @xmath86 , is given by:@xcite @xmath87 where @xmath88 is the cofeb thickness . when the static magnetic field is applied parallel to the interface ( @xmath89=90@xmath90 ) , the spin - current density at the interface between cofeb / mgo and ge , @xmath91 is given by:@xcite @xmath92\ ] ] where @xmath29 is the strength of the microwave magnetic field into the resonance cavity . @xmath29 is calculated by measuring the @xmath93 factor of the resonance cavity @xmath94 , where @xmath95 is the width at half maximum of the frequency distribution when the sample is placed into the cavity . to measure @xmath96 we use a second frequencemeter in series with the first one . the voltage @xmath83 due to the inverse spin hall effect is always symmetric with respect to the resonance field and its amplitude is discussed in ref . we then modify the equivalent circuit used in ref . @xcite and refined the model used in ref . @xcite to account for electron transport through the tunnel and schottky barriers back to the fm ( see appendix b ) . then the ishe voltage in our system is given by : @xmath97\theta_{she}l_{sf}^{cb}\tanh\left(\frac{t_{n}}{2l_{sf}^{cb}}\right)\left(\frac{2e}{\hbar}\right)j_{s}^{0}\ ] ] where @xmath98 is the width of the ferromagnetic electrode ( 150 @xmath23 m ) , @xmath99 ( resp . @xmath100 ) is the ge ( resp . ta / cofeb ) thickness , @xmath101 ( resp . @xmath102 ) is the ge ( resp . ta / cofeb ) conductivity . @xmath103 where @xmath88 and @xmath104 ( resp . @xmath105 and @xmath106 ) are the thickness and conductivity of the cofeb ( resp . ta ) layer . @xmath107 depends on the resistance - area product @xmath108 of the interface between cofeb / mgo and ge as : @xmath109 in order to estimate the @xmath83 magnitude , the electromotive force and the ferromagnetic spectrum are measured simultaneously . the measured voltage might have one symmetric , @xmath110 , one asymmetric , @xmath72 , and one offset contributions . the raw data will be fitted with : @xmath111 note that eq . ( [ eqv2 ] ) is similar to eq . ( [ eqvamr ] ) but in the presence of spin pumping the symmetric voltage is : @xmath112 . fig . [ fig4oop ] shows the oop dependence of the resonance field , peak - to - peak linewidth and electromotive force on the as - grown and annealed ta(5nm)/cofeb(5nm)/mgo(3.3nm)//sio@xmath5 reference samples . from the angular dependence of @xmath53 , we obtain the effective saturation magnetization ( @xmath46 ) and the @xmath35 factor and from @xmath57 we obtain the damping constant ( @xmath55 ) and the angular dispersion ( @xmath113 ) . the angular dependence of the peak - to - peak linewidth can be calculated using the following method : after fitting numerically the oop dependence of the resonance field ( fig . [ fig4oop](a , d ) ) we use @xmath46 and @xmath35 to calculate the theoretical dispersion relationship between @xmath31 and the external magnetic field for different @xmath30 angles . this is shown in fig . [ fig2dispersion ] where the dotted line corresponds to the frequency at which the measurements are performed . the intercept of each curve with the dotted line gives the value of the resonance field @xmath53 along with the equilibrium polar angle @xmath39 of @xmath114 at different @xmath30 values . the oop linewidth angular dependence is shown in fig . [ fig4oop](e ) and fitted using eq . ( [ eqdhppline ] ) . in addition , in fig . [ fig4oop](c ) , we have used the @xmath115 formula of appendix a to fit the oop angular dependence of the symmetric voltage contribution to the electromotive force in the as - grown reference sample . the oop angular dependence of the symmetric voltage in the ge - based device ( @xmath83 ) clearly shows a different behavior ( see fig . [ fig8geoop](c ) and [ fig8geoop](f ) ) . ( a , d ) , @xmath57 ( b , e ) , and amplitude of the electromotive force ( c , f ) according to eq . ( [ eqvamr ] ) . samples are multilayers - see fig . [ fig1](a)- of ta(5nm)/cofeb(5nm)/mgo(3.3nm)//sio2 as - grown ( a - c ) and annealed ( d - f ) . red curves in ( a , d ) are numerical fits as explained above . the curves in ( b , e ) are numerical calculations to the @xmath57 contributions according to eq . ( [ eqdhppline ] ) . the solid curve in ( c ) is a fit using the @xmath115 formula of appendix a , eq . ( [ eqvphe ] ) . , width=12 ] we have also recorded the power dependence of both the fmr signal and the electromotive force when @xmath58 is applied parallel to the film plane . the results are shown in fig . [ fig5pow ] . the electromotive force was fitted according to eq . ( [ eqvamr ] ) . @xmath73 and @xmath72 depend linearly on the applied power in the whole power range whereas @xmath71 slightly deviates from the linear behavior for high powers . nevertheless we note that @xmath116 in both reference samples . the sample sizes are @xmath117 mm@xmath24 for the as - grown sample and @xmath118 mm@xmath24 for the annealed one . for the same rf power of 200 mw and the field applied parallel to the film plane ( @xmath30=90@xmath90 ) , we found @xmath119 - 159 @xmath120 for the as - grown sample and @xmath119 - 64.2 @xmath120 for the annealed one . since @xmath72 depends linearly on the ferromagnetic electrode width , we can estimate the expected @xmath72 value for the cofeb bar of fig . [ fig01ab](a ) : @xmath1196.8 @xmath120 for the as - grown sample and @xmath26.4 @xmath120 for the annealed one . the expected symmetric contribution to the electromotive force @xmath71 will then be almost one order of magnitude less . furthermore , in the device of fig . [ fig01ab ] , the electrical contacts are no more made on the metallic multilayer , as show in fig . [ fig1](a ) , but on au / ti ohmic contacts on top of ge which would reduce the phe and ahe contributions . ) and asymmetric ( @xmath72 ) contributions to the electromotive force according to eq . ( [ eqvamr ] ) . samples are multilayers of ta(5nm)/cofeb(5nm)/mgo(1.1nm)//sio2 as grown ( squares ) and annealed ( circles ) . solid lines are guides for the eyes and dashed lines in ( a ) show the non linear behavior of @xmath71 in all the experimental frequency range . the insets show the electromotive force measured under different power excitations.,width=6 ] fig . [ figdampingrefer ] shows the frequency dependence of the resonance field ( a ) and peak - to - peak linewidth ( b ) of the as - grown and annealed reference samples . in both figures , we have used the @xmath35 factors deduced from the oop angular dependence of @xmath53 ( see fig . [ fig4oop](a ) and [ fig4oop](d ) ) and adjusted the @xmath46 and @xmath55 values according to eq . ( [ eqkittel ] ) and ( [ eqalpha ] ) respectively to fit the curves . the frequency independent part of the peak - to - peak linewidth @xmath60 which is due to inhomogeneities in the magnetic layer is very weak in both samples . we find 1.1 oe for the as - grown sample and 2.3 oe for the annealed one which confirms that @xmath121 at a frequency close to 9 ghz . moreover the effective perpendicular anisotropy ( 4@xmath122 ) increases from 1.175 t up to 1.545 t upon annealing as recently reported.@xcite both samples exhibit very low damping constants comparable to the ones found in ref . interestingly the peak - to - peak linewidth and damping constants decrease upon annealing in contrast with other results.@xcite it means that we have effectively reduced the intrinsic inhomogeneities of the cofeb electrode by annealing . in particular , the annealing process did not promote chemical inter - diffusion at the interfaces with cofeb as found in thinner cofeb films in magnetic tunnel junctions.@xcite ) in ( a ) , and eq.([eqalpha ] ) in ( b).,width=6 ] in this section , we consider the device shown in fig . [ fig01ab](a ) where the cofeb electrode has been directly grown on the ge film without tunnel barrier . the fmr line and the corresponding electromotive force are shown in fig . [ figge_cfb](a ) . a clear absorption is observed in the fmr spectrum whereas the electromotive force at the resonance field is negligible . hence , in the measuring geometry of fig . [ fig01ab](a ) where the voltage is directly probed on the germanium layer , we do not detect the phe and the ahe in the cofeb ferromagnetic layer at the resonance . the angular dependence of the resonance field and peak - to - peak linewidth are displayed in fig . [ figge_cfb](b ) . the frequency dependence of @xmath57 and @xmath53 are shown in fig . [ figdampingcfbge ] . first , the effective cofeb saturation magnetization @xmath46 is lower than in the reference sample : this is probably due to the intermixing between cofeb and ge at the interface . in the same way , the larger damping constant @xmath55 may be due to interface inhomogeneities as a consequence of intermixing and not to spin pumping since no electromotive force is observed . we have then performed the same measurements on the annealed sample . in that case , both the ferromagnetic resonance signal and the electromotive force vanish and the cofeb film has completely diffused into the ge layer . these results show that the mgo tunnel barrier is not only necessary to overcome the conductivity mismatch issue but also to prevent the intermixing between cofeb and ge at the interface . . similar devices without mgo barrier and after annealing process do not exhibit ferromagnetic resonance.,width=12 ] ) in ( a ) , and eq.([eqalpha ] ) in ( b ) , width=6 ] we now consider the same device as in the previous section but with a thin mgo tunnel barrier inserted between cofeb and ge as shown in fig . [ fig01ab](a ) . the fmr spectrum and the corresponding electromotive force are shown in fig . [ figge_mgo_cfb](a ) for the as - grown sample and fig . [ figge_mgo_cfb](b ) for the annealed one . here a clear electromotive force is detected at the resonance field in both cases . the red line is the fit according to eq . ( [ eqv2 ] ) considering a single symmetric contribution . moreover by annealing we observe an enhancement of the electromotive force signal . in fig . [ fig8geoop ] , the oop angular dependence of the resonance field , peak - to - peak linewidth and the amplitude of the electromotive force of both samples are displayed . like in the previous section , the complete analysis of these data yields @xmath46 , @xmath35 , @xmath55 , and @xmath123 . the solid lines in fig . [ fig8geoop](c ) and fig . [ fig8geoop](f ) are fits according to the formula:@xcite @xmath124/[(2h_{res}/(4\pi m_{eff}))cos(\theta_{m0}-\theta_h)-cos(2\theta_{m0})-cos^2(\theta_{m0})]^2\}$ ] . .,width=12 ] we also measured the fmr spectrum in the parallel case at different frequencies on both devices ( not shown ) . the frequency dependence of @xmath57 always shows a linear behavior with a very low @xmath125 value showing that the gilbert - type effect is the dominating contribution to the damping in all the samples studied . in table [ tablealpha ] , we can clearly see that the annealing process increases the perpendicular magnetic anisotropy of the system ( enhancement of @xmath46 ) and reduces the intrinsic damping constant . spin pumping in ge leads to an increase of the damping constant ( @xmath126 ) with respect to that of the reference system ( @xmath127 ) . lccccc & & + & cofeb / mgo ref . & n - ge device & & cofeb / mgo ref . & n - ge device + as - grown & @xmath128 & @xmath129 & & @xmath130 & @xmath131 + annealed & @xmath132 & @xmath133 & & @xmath134 & @xmath135 + the power dependence of the @xmath83 amplitude when the external dc magnetic field is applied parallel to the fm layer is shown in fig . [ fig9powdepge ] where the solid line is a linear fit . such linear behavior accounts well for the @xmath136 dependence of the v@xmath137 since the microwave power is proportional to the square of the rf magnetic field ( @xmath138 ) . all these results support the fact that the measured electromotive force is due to spin pumping from the cofeb electrode and inverse spin hall effect in germanium . in order to estimate the spin hall angle @xmath3 in @xmath1-ge , we have calculated the tunnel spin mixing conductance according to eq . ( [ eqgmix ] ) . we found : @xmath139 @xmath140 for the device with the as - grown cofeb layer and @xmath141 @xmath140 for the device with the annealed cofeb layer . the spin - current density at the interface @xmath142 , when @xmath89=90@xmath90 , is calculated using eq . ( [ eqjs0 ] ) where the cofeb effective saturation magnetization @xmath46 , the gyromagnetic ratio @xmath143 , and the damping factor @xmath55 were deduced from fmr measurements . the results are reported in table [ tableall ] . for a power of 200 mw , the microwave magnetic field ( @xmath29 ) was measured with the sample inside the resonator cavity . the @xmath83 amplitude is calculated according to eq . ( [ eqvishe ] ) with the width of the ferromagnetic electrode @xmath98=150 @xmath23 m , the fm thickness @xmath144 nm , the ge channel thickness @xmath145=40 nm and the conductivities and the resistance - area product of the interfase @xmath146 given in table [ tableall ] . the spin diffusion length in the semiconductor channel is @xmath1471.3 @xmath23 m ( ref . the conductivities ( including the interface @xmath108 value ) and the spin diffusion length were measured independently on the same device ( ref . @xcite ) . lcc & & + @xmath86 @xmath148 @xmath140 ) & 0.6 & 1.0 + @xmath29 ( g ) & 0.73 & 0.77 + @xmath149 ( @xmath150 ) & 0.047 & 0.11 + @xmath151 & 9440 & 1250 + @xmath152 & 270 & 270 + @xmath153 & 1.047 & 1.085 + @xmath154 & 0.0011 & 0.00044 + we then estimate the spin hall angle in @xmath1-ge from the annealed sample at room temperature : @xmath155 , which is of the same order of magnitude as in @xmath1-gaas ( 0.007 in ref . @xcite ) and one order of magnitude larger than in @xmath4-si ( 0.0001 in ref . @xcite ) . in a similar way we could estimate the spin hall angle in @xmath1-ge using the data from the as - grown sample and found : @xmath1560.00044 . such a difference might come either from the error bars and/or from the phenomenological model we have used here . we have measured several annealed devices from the same batch and found @xmath3 between 0.0010 and 0.0012 which gives an estimation of the error bar . we thus conclude that the phenomenological model we use to estimate @xmath3 is not adapted to our system . in particular , this model does not account for the presence of interface states between mgo and ge . we have shown in a previous work that interface states play a crucial role in the spin injection mechanism.@xcite electrical spin injection into ge proceeds by two - step tunneling : the electrons tunnel from the fm to the localized interface states ( is ) through the mgo barrier and from the is to the ge conduction band through the schottky barrier . because spin flips occur into interface states , the spin accumulation ( hence the spin current ) is drastically reduced in the ge conduction band . by annealing , the density of interface states is reduced and direct spin injection into the ge conduction band is favored . as a consequence , the spin current in the as - grown sample @xmath157 is reduced as compared to the spin current in the annealed sample which leads to the underestimation of @xmath3 as found experimentally . we thus give in the next section a microscopic model accounting for the presence of the tunnel barrier and interface states to accurately describe spin pumping and ishe in germanium . in the as - grown and annealed cofeb / mgo / ge samples , we could clearly measure an electromotive force due to ishe at the ferromagnetic resonance of cofeb . this photovoltage has a symmetric lorentzian shape . furthermore we have shown that all our findings are in good agreement with the observation of ishe : symmetrical behaviour of @xmath83 around the resonance field @xmath53 , @xmath83=0 when the external magnetic field is applied perpendicular to the film ( @xmath89=0 ) , @xmath83 changes its sign when crossing @xmath89=0 ( fig . [ fig8geoop ] ) , and finally the linear dependence of its amplitude with the microwave power excitation ( fig . [ fig9powdepge ] ) . this result clearly demonstrates the presence of both spin accumulation and related spin current in the ge conduction band at room temperature . it was also supported by temperature dependent measurements in a previous work.@xcite in order to confirm that the photovoltage we measure is really due to ishe and rule out any spurious effects , we carried out complementary measurements . first we studied the photovoltage in millimeter - sized reference samples ( both as - grown and annealed ) made of cofeb / mgo / sio@xmath5 with the voltage probes directly connected to the cofeb film . in that case , we found a dominant asymmetric voltage contribution with respect to the resonance field . it corresponds to the planar hall effect in the ferromagnet as a combination of anisotropic magnetoresistance and the rf current induced in the ferromagnet by the non - vanishing electric field from the cavity ( see appendix a ) . this asymmetric voltage contribution due to phe could not be detected on the device of fig . [ fig01ab ] with ge . moreover the out - of - plane angular dependence of this weak symmetric voltage on the reference sample is different from that of the symmetric voltage we detected in the device of fig . [ fig01ab ] with ge . to summarize this study on the reference sample , we can claim that the symmetric photovoltage observed in cofeb / mgo / ge samples is due to ishe and not to phe in the ferromagnet . we also carried out the same measurements on cofeb / ge samples to study the effect of the mgo tunnel barrier . without mgo tunnel barrier , we never detected a photovoltage in ge . it first proves that the photovoltage due to the phe in cofeb is undetectable in ge . it also shows that the mgo tunnel barrier is necessary to perform spin injection in ge by spin pumping . furthermore , as shown in fig . [ figseebeck ] , we have recorded several voltages on the same cofeb / mgo / ge device at the ferromagnetic resonance in order to estimate the tunneling spin seebeck effect.@xcite indeed at the ferromagnetic resonance , the cofeb electrode absorbs part of the incident microwave power which increases its temperature . as a consequence , a vertical temperature difference may appear between the cofeb electrode and the ge layer . this temperature gradient may create a tunneling seebeck voltage and a tunneling spin seebeck voltage which is only a few percents of the seebeck voltage.@xcite the resulting spin current gives rise to ishe in germanium just like spin pumping . in order to discriminate between spin pumping and the tunneling spin seebeck effect , we measured the following voltages at the fmr : @xmath158 , @xmath159 and @xmath160 shown in fig . [ figseebeck ] . the sum of the tunneling seebeck and tunneling spin seebeck voltages ( @xmath161 ) is given by : @xmath162 or @xmath163 . as shown in fig . [ figseebeck ] , @xmath161 is negligible ( below the noise level ) which rules out the presence of tunneling spin seebeck at the ferromagnetic resonance in our system . therefore spin injection in ge proceeds by spin pumping and not by tunneling spin seebeck effect . ) and the voltage between the fm layer and one of the ohmic contacts ( @xmath160 ) . on the left are shown the contacts geometries . finally , the expected tunneling spin seebeck voltage is shown at the bottom right of the figure.,width=10 ] we now address the important issue of the microscopic origin of spin pumping effects in ge through a mgo tunnel barrier from a theoretical point of view . as demonstrated below , the origin of spin pumping into scs through a tunnel barrier lies in the _ evanescent _ but however _ non - zero _ exchange coupling between a band of localized states ( ls ) and the ferromagnet through the tunnel barrier , nonetheless sufficiently transparent . indeed , spin - pumping @xcite in metallic tunnel junctions is expected to fall - off in the absence of any exchange field experienced from the ferromagnet ( fm ) by the delocalized carriers injected in the non - magnetic metal ( n ) . on the other hand , spin injection into a sc by electrical means , as well as by spin - pumping , requires a tunnel barrier at the interface between both types of materials @xcite in order to overcome the impedance mismatch issue @xcite describing a total diffusive spin current backflow towards the fm . as shown in our experiments , spin - pumping in a semiconductor with a tunnel barrier can be recovered with some conditions . first , the carriers injected by tunneling from a fm contact have to remain localized at the interface between the tunnel barrier and the semiconductor in the timescale of a single magnetization precession . second , the effective _ tunnel exchange _ field experienced by the carriers , that we call hereafter @xmath164 , has to be large enough for the spin to rotate in a timescale of a magnetization precession . these two necessary conditions may be fulfilled within a two - step tunneling picture of spin injection into evanescent ( or localized states ) @xcite and in the limit of an effective exchange field larger than a certain lower bound . this will be demonstrated below . the third condition to observe spin - pumping in fm / tunnel barrier / sc systems is a minimum value for the conductance of the schottky barrier delimiting the two regions _ i.e _ the evanescent states and the sc channel . a thermal activation may be needed to fulfill this third condition . we thus give an analytical expression for the source term taking into account a two - step tunneling process . let us consider the standard theory of spin - pumping at the fm / n interface . the source term is known to be equal to:@xcite @xmath165 in the case of a fm / n ohmic contact where @xmath166 is the unit magnetization vector and @xmath167 the complex spin mixing conductance . the spins pumped into n then create a _ diffusive _ spin current backflow to the fm according to the three - dimensional spin - dependent transmission matching at the fm / n interface:@xcite @xmath168~{\bf{m}}\nonumber \\ -\frac{\mathcal{r}e~g^{\uparrow \downarrow}}{4\pi}{\bf{m}}\times ( \bf{\mu_s^n}\times { \bf{m}})+\frac{\mathcal{i}m~g^{\uparrow \downarrow}}{4 \pi}({\bf{m}}\times \bf{\mu_s^n})\end{aligned}\ ] ] where @xmath169 , @xmath170 in n and @xmath171 , @xmath172 in fm are respectively the charge and spin accumulations at the interface . @xmath35 is the sum of spin - up and spin - down conductances and @xmath4 is the interfacial spin asymmetry coefficient . this backflow of spin current results in a down - renormalization of the spin current pumped in the non magnetic material as shown in a recent couple of papers.@xcite the exact form of the corresponding _ down renormalization _ has to be considered case by case . the new source term describing spin - pumping in a broad band of evanescent states at the tunnel barrier / sc interface has to involve a small but however non - zero exchange interaction @xmath173 between localized states and the magnetization @xmath114 ( of unit vector @xmath174 ) of the fm through the tunnel barrier ; this exchange interaction couples evanescent wavefunctions inside the barrier . in the following , we will define @xmath173 in the form : @xmath175 where @xmath176 is the bare _ on - site _ exchange interaction of the order of the exchange interaction in fm or even larger ( about 1 ev ) , @xmath177 is the imaginary electronic wavevector in the barrier and @xmath178 the barrier thickness . @xmath179 is the tunnel transmission coefficient . we note @xmath180 the mean energy broadening of the localized states due to the finite carrier lifetime ( @xmath181 ) through their escape towards the ferromagnetic reservoir fm . this energy broadening can be expressed _ vs. _ the localization energy @xmath182 within the centers and @xmath179 according to @xmath183 ( ref . @xcite ) . note that the escape towards the semiconductor channel , moderately doped , is generally prohibited in an energy band located downward the fermi energy . if one defines three different components of the carrier spin vector @xmath184 injected in the evanescent states by @xmath185 , @xmath186 and @xmath187 , the equation of motion for the injected spin , along the @xmath68 direction at time @xmath188 , in a localized state in the exchange field of the magnetization @xmath166 rotating in the ( @xmath68,@xmath69 ) plane follows the heinsenberg evolution for the spin - operator @xmath189_{-}$ ] ; thus giving _ in fine _ : @xmath190\end{aligned}\ ] ] with @xmath191 the rf pulsation frequency , @xmath192 the _ exchange _ pulsation and @xmath193 the effective pulsation of the spin during its rotation . in an homogeneous fm layer , the precession frequency due to the exchange interaction @xmath194 rad.s@xmath16 is very large compared to the rf frequency . this results in a small average component of the spin vector pumped along @xmath80 : @xmath195 , of the order of @xmath196 . however , this small spin rotation is counterbalanced by a large number of uncompensated spins due to the strong exchange and whose number equals @xmath197 ( @xmath198 is the density of states ) . the total spin along the @xmath80 direction then writes @xmath199 . one recovers the standard formula for the spin - current pumped at the ohmic fm / n interface if the interfacial spin - mixing conductance @xmath200 is introduced hereafter . in the case of spin - pumping into evanescent states , the _ exchange _ pulsation @xmath201 can be of the order of magnitude of the rf pulsation or even smaller . to derive the average @xmath185 component pumped in a localized center , one has to perform a time average of @xmath185 on the carrier lifetime @xmath181 to give : @xmath202\end{aligned}\ ] ] by analogy with the previous calculations relative to the bulk fm layer , and taking into account that the total number of uncompensated spins introduced by the _ tunneling _ exchange interactions , @xmath203 , one can generalize the total spin accumulation ( @xmath204 ) pumped along the @xmath80 direction as : @xmath205\times \left[\hbar { \bf{m}}\times\frac{d { \bf{m}}}{dt}\right]\end{aligned}\ ] ] for any rotation @xmath206 vector . it comes two important conditions on the effective exchange @xmath173 to generate significant spin - pumping at the fm / tunnel barrier / sc interface : \1 ) @xmath164 must be larger than the intrinsic energy broadening @xmath207 ( or equivalently @xmath208 ) corresponding to a time of interaction larger than the time of the spin precession . \2 ) @xmath164 must be larger than the rf frequency energy @xmath209 of the order of @xmath210 in the present case . this condition corresponds to a characteristic spin precession time due to exchange , and necessary for any spin rotation , smaller than the magnetization @xmath166 precession time itself . once these two conditions are satisfied , the spin - pumping effect at the fm / tunnel barrier / sc interface becomes efficient , a large rotation angle of the spin @xmath211 compensating the small number of uncompensated spins @xmath212 . the total spin - current pumped ( @xmath213 ) at the ls / sc channel interface , that is the _ source term _ , equals @xmath214 where @xmath215 is the schottky conductance playing the role of the mixing conductance @xmath167 for fm / n interfaces . we now proceed to the _ down - renormalization _ of the spin - current pumped in the sc as described previously . in the light of the recent published works,@xcite this total spin - current has then to be decomposed into the _ real _ spin - current injected in the ge channel added to a backflow of spin - current relaxing either into the localized states or into the fm reservoir by back - absorption . we have : @xmath216 where @xmath217 is the spin conductance of the ge layer of thickness @xmath218 and spin diffusion length @xmath219 ( ref . @xcite ) and where @xmath220 is the corresponding bulk spin resistance . @xmath221 ( _ resp . _ @xmath222 ) is the spin accumulation generated in the ge layer ( _ resp . _ in the ls ) and @xmath223 is the spin resistance of the ls ( @xmath224 is the corresponding spin lifetime ) . ( [ gsh ] ) describes the continuity of the spin - current backflow through the schottky barrier . it results from these calculations that the effective spin - current @xmath225 injected in the ge channel writes : @xmath226 with @xmath204 the spin accumulation generated in the localized states by spin - pumping like calculated previously . a zero schottky conductance leads to zero spin - current . on the opposite case of a large schottky conductance _ e. g. _ on increasing the temperature , the spin current pumped in the ge channel writes @xmath227 _ i.e. _ it corresponds to the maximum spin - current pumped weighted by the ratio of spin - flips occurring in the channel itself over the total number of spin - flips also possible in the band of ls and parameterized by @xmath228 . consequently , the _ real _ spin current pumped into the ge layer depends on the schottky conductance and on the different spin - resistances involved in the spin - relaxation process . the main question that has to be addressed in the future is the fraction of the spin - current pumped and relaxing in the ls by spin - flip . indeed , this part of the spin - current would contribute to the broadening of the fmr spectra but not to the ishe voltage . finally these parameters have to be determined experimentally in order to relate this microscopic model to our data . we have demonstrated that at the fm / i / n interface where n is a non - magnetic semiconductor channel we could inject a spin current by spin pumping from the fm layer into the n channel at the ferromagnetic resonance . we have also shown that the mgo tunnel barrier is useful not only to overcome the conductivity mismatch between cofeb and ge but also to keep the magnetic properties of the fm after annealing the samples . there is no spin pumping nor inverse spin hall effect voltage signal on devices without barrier while it clearly appears on the devices with the mgo barrier . moreover there is an enhancement of the ishe signal and consequently of the spin hall angle when the device is annealed . a microscopic model involving interface states and _ evansecent _ tunnel exchange coupling has been developed in order to explain spin pumping into ge from a fm electrode through a tunnel barrier . we could finally find and discuss the spin hall angle in @xmath1-ge : @xmath229 from the annealed sample and @xmath230 from the as - grown one . rojas - sanchez would like to acknowledge the eurotalent cea program for its financial support . the sample is rotated in the electromagnet with the static magnetic field @xmath231 applied perpendicular ( @xmath30=0@xmath90 ) to parallel to the film plane ( @xmath30=90@xmath90 ) . we define the sample magnetization : @xmath232 and the current in the ferromagnetic electrode : @xmath233 where @xmath234 is the equilibrium magnetization in static conditions . it makes an angle @xmath235 with the normal to the sample @xmath236 . @xmath237 is the time - varying part of the magnetization . we do not apply any bias current to the system : @xmath238 and consider the rf current created in the ferromagnetic layer by the non - vanishing rf electric field : @xmath239 where @xmath76 is the constant phase difference between the current and the magnetization precession at resonance . the generalized ohm s law then writes:@xcite @xmath240 @xmath241 , @xmath242 and @xmath243 are the resistivity , the anisotropic magnetoresistance and the anomalous hall constant of the ferromagnetic electrode . since we experimentally measure a dc voltage , we calculate the time average of @xmath244 : @xmath245-r_h\left\langle \vec{j}\times\vec{m}\right\rangle\ ] ] where : @xmath246 , @xmath247 and @xmath248 . we then obtain the voltage : @xmath249 where @xmath250 . the first term corresponds to the planar hall effect and the second one to the anomalous hall effect . @xmath251 and @xmath252 are then determined by solving the landau - lifschitz - gilbert ( llg ) equation : @xmath253 where @xmath143 is the gyromagnetic ratio , @xmath55 the damping factor and @xmath254 the effective magnetic field given by : @xmath255 here we only consider the shape anisotropy field . the radiofrequency magnetic field can be written as : @xmath256 , @xmath257 where @xmath31=9.4 ghz is the x - band cavity frequency . in static conditions , the magnetization equilibrium angle @xmath28 is found by solving : @xmath258 _ i.e. _ @xmath259 . the resolution of the llg equation yields : @xmath260 where : @xmath261 , @xmath262 and @xmath263 . by using : @xmath264 , @xmath265 and @xmath266 , we finally find : @xmath267\\ m_y & = & \frac{m_s h_{rf}cos(\theta_m -\theta_h)}{a^2+b^2}\left[(\alpha a+\frac{\gamma}{\omega}bb)cos(\omega t)-(\frac{\gamma}{\omega}ba-\alpha b)sin(\omega t)\right]\end{aligned}\ ] ] where : @xmath268 $ ] and @xmath269 $ ] . the fmr spectrum is defined by @xmath252 . then , after time averaging , we obtain : @xmath270 \\ v_{ahe}&=&\frac{w_fr_h j_0 h_{rf}cos(\theta_m -\theta_h)sin\theta_m}{2(a^2+b^2)}\times \left[bsin\psi -acos\psi\right]\end{aligned}\ ] ] here we point out that the rf electric field @xmath271 induces an additional angular dependence because @xmath272 change with the dc magnetic field angle as @xmath273 where @xmath274 is the direction of the rf electric field in the cavity ( see fig . [ fig_annex1 ] ) . note that @xmath272 is proportional to the strength of the rf electric field , _ i.e. _ to the strength of the rf magnetic field . as a consequence the magnitude of the electromotive forces @xmath275 and @xmath276 are proportional to @xmath277 . hence they exhibit a linear dependence with the microwave power as shown in the main text . such linear dependence might allow to deduce either the ratio @xmath278 or the phase shift @xmath76 . since the hall coefficient is of the order of @xmath279 @xmath280 and @xmath281 is of the order of @xmath282 g , then @xmath283 is much smaller than the anisotropic magnetoresistance , @xmath242 . + the oop angular dependence of the symmetric component of either @xmath275 or @xmath276 clearly shows a behavior different from that of the ishe out - of - plane angular dependence ( @xmath284 ) as shown in fig . [ fig4oop](c ) where we considered @xmath285 and @xmath286 . at the ferromagnetic resonance of the cofeb electrode , the combination of spin pumping and ishe creates a charge current ( in a / m ) @xmath287 in the ge layer . part of this charge current flows back to the ferromagnetic and tantalum capping layers which affects the estimation of @xmath83 and @xmath3 in germanium . in the following , we make an estimation of this backflow current . in fig . [ figbackflow ] , the current density crossing the interface at @xmath68 corresponds to the variation of the current in the layers : where @xmath100 , @xmath289 and @xmath290 are the thickness , current density and potential in the ta / cofeb bilayer ; @xmath108 is the resistance - area product of the interface between cofeb / mgo and ge . the current densities in each layer with conductivities @xmath102 and @xmath101 can be written : a. jain , l. louahadj , j. peiro , j. c. le breton , c. vergnaud , a. barski , c. beign , l. notin , a. marty , v. baltz , s. auffret , e. augendre , h. jaffrs , j. m. george , and m. jamet , appl . lett . * 99 * , 162102 ( 2011 ) . a. jain , j .- c . rojas - sanchez , m. cubukcu , j. peiro , j. c. le breton , e. prestat , c. vergnaud , l. louahadj , c. portemont , c. ducruet , v. baltz , a. barski , p. bayle - guillemaud , l. vila , j .- attan , e. augendre , g. desfonds , s. gambarelli , h. jaffrs , j .- george , and m. jamet , phys . lett . * 109 * , 106603 ( 2012 ) . c. deguet , j. dechamp , c. morales , a .- m . charvet , l. clavelier , v. loup , j .- hart- mann , n. kernevez , y. campidelli , f. allibert , c. richtarch , t. akatesu and f. letertre , electrochem . . proc . * 2005 - 06 * , 78 ( 2005 ) . a. jain , j .- c . rojas - sanchez , m. cubukcu , j. peiro , j .- c . le breton , c. vergnaud , a. augendre , l. vila , j .- attan , s. gambarelli , h. jaffrs , j .- george and m. jamet , eur . j. b * 86 * , 140 ( 2013 ) . a. jain , c. vergnaud , j. peiro , j. c. le breton , e. prestat , l. louahadj , c. portemont , c. ducruet , v. baltz , a. marty , a. barski , p. bayle - guillemaud , l. vila , j .- attan , e. augendre , h. jaffrs , j .- george , and m. jamet , appl . * 101 * , 022402 ( 2012 ) .	we have measured the inverse spin hall effect ( ishe ) in _ n_-ge at room temperature . the spin current in germanium was generated by spin pumping from a cofeb / mgo magnetic tunnel junction in order to prevent the impedance mismatch issue . a clear electromotive force was measured in ge at the ferromagnetic resonance of cofeb . the same study was then carried out on several test samples , in particular we have investigated the influence of the mgo tunnel barrier and sample annealing on the ishe signal . first , the reference cofeb / mgo bilayer grown on sio@xmath0 exhibits a clear electromotive force due to anisotropic magnetoresistance and anomalous hall effect which is dominated by an asymmetric contribution with respect to the resonance field . we also found that the mgo tunnel barrier is essential to observe ishe in ge and that sample annealing systematically lead to an increase of the signal . we propose a theoretical model based on the presence of localized states at the interface between the mgo tunnel barrier and ge to account for these observations . finally , all of our results are fully consistent with the observation of ishe in heavily doped @xmath1-ge and we could estimate the spin hall angle at room temperature to be @xmath20.001 .
one can conceive of information theory in the broad sence as covering the theory of gaining , transferring , and storing information , where the first is usually called statistics."@xcite . shannon information theory and mathematical statistics interaction revealed to be effective . this interplay is mutually fruitful , in some works results of probability theory and statistics were obtained with application of information - theoretical methods and there are studies where statistical results provide ground for new findings in information theory @xcite , @xcite , @xcite@xcite , @xcite , @xcite , @xcite , @xcite , @xcite@xcite . this paper can serve an illustration of application of information - theoretical methods in statistics : on one hand this is analogy in problem formulation and on the other hand this is employment of technical tools of proof , specifically of the method of types @xcite , @xcite . it is often necessary in statistical research to make decisions regarding the nature and parameters of stochastic model , in particular , the probability distribution of the object . decisions can be made on the base of results of observations of the object . the vector of results is called a sample . the correspondence between samples and hypotheses can be designed based on some selected criterion . the procedure of statistical hypotheses detection is called test . the classical problem of statistical hypothesis testing refers to two hypotheses . based on data samples a statistician makes decision on which of the two proposed hypotheses must be accepted . many mathematical investigations , some of which have also applied significance , were implemented in this direction @xcite . the need of testing of more than two hypotheses in many scientific and applied fields has essentially increased recently . as an instance microarray analysis could be mentioned @xcite . the decisions can be erroneous due to randomness of the sample . the test is considered as good if the probabilities of the errors in given conditions are as small as possible . frequently the problem was solved for the case of a tests sequence , where the probabilities of error decrease exponentially as @xmath0 , when the number of observations @xmath1 tends to the infinity . we call the exponent of error probability @xmath2 the _ reliability_. in case of two hypotheses both reliabilities corresponding to two possible error probabilities could not be increased simultaneously , it is an accepted way to fix the value of one of the reliabilities and try to make the tests sequence get the greatest value of the remaining reliability . such a test is called _ logarithmically asymptotically optimal _ such optimal tests were considered first by hoeffding @xcite , examined later by csiszr and longo @xcite , tusnady @xcite , @xcite ( he called such test series _ exponentially rate optimal _ ( ero ) ) , longo and sgarro @xcite . the term lao for testing of two hypotheses was proposed by birge @xcite . amongst papers on testing , associated with information theory , we can also note works of natarajan @xcite , gutman @xcite , anantharam @xcite , han @xcite and of many others . some objectives in this direction were first suggested in original introductory article by dobrushin , pinsker and shiryaev @xcite . the achievable region of error probability exponents was examined by tuncel @xcite . the problem has common features with the issue studied in the information theory on interrelation between the rate @xmath3 of the code and the exponent @xmath2 of the error probability . in information theory the relation @xmath4 is called according to shannon the _ reliability function _ , while @xmath5 is named the @xmath2-_capacity _ , or the _ reliability - rate function _ , as it was introduced by haroutunian @xcite , @xcite , @xcite . simple but actual concept of not only separate but also simultaneous investigation of some number of objects of the same type , evidently , was first formulated by ahlswede and haroutunian @xcite for reliable testing of distributions of multiple items . but simultaneous examination of properties of many similar objects may be attractive and effective in plenty of other statistical situations . the organization of this paper is as follows . we start with the definitions and notations in the next section . in section 3 we introduce the problem of multihypotheses testing concerning one object . in section 4 we consider the reliability approach to multihypotheses testing for many independent and dependent objects . section 5 is dedicated to the problem of statistical identification under condition of optimality . section 6 is devoted to description of characteristics of lao hypotheses testing with permission of rejection of decision for the model consisting of one and of more independent objects . we denote finite sets by script capitals . the cardinality of a set @xmath6 is denoted as @xmath7 . random variables ( rvs ) , which take values in finite sets @xmath6 , @xmath8 are denoted by @xmath9 , @xmath10 . probability distributions ( pds ) are denoted by @xmath11 , @xmath12 , @xmath13 , @xmath14 , @xmath15 , @xmath16 . let pd of rv @xmath9 , characterizing an object , be @xmath17 , and conditional pd of rv @xmath9 for given value of state @xmath18 of the object be @xmath19 . the shannon entropy @xmath20 of rv @xmath9 with pd @xmath11 is : @xmath21 the conditional entropy @xmath22 of rv @xmath9 for given rv @xmath10 with corresponding pds is : @xmath23 the divergence ( kullback - leibler information , or distance " ) of pds @xmath11 and @xmath13 on @xmath6 is : @xmath24 and conditional divergence of the pd @xmath25 and pd @xmath26 is : @xmath27 for our investigations we use the method of types , one of the important technical tools in shannon theory @xcite . the type @xmath28 of a vector @xmath29 is a pd ( the empirical distribution ) @xmath30 where @xmath31 is the number of repetitions of symbol @xmath32 in vector @xmath33 . the joint type of vectors @xmath34 and @xmath35 is the pd @xmath36 where @xmath37 is the number of occurrences of symbols pair @xmath38 in the pair of vectors @xmath39 . the conditional type of @xmath40 for given @xmath41 is conditional pd @xmath42 defined by relation @xmath43 for all @xmath44 , @xmath45 we denote by @xmath46 the set of all types of vectors in @xmath47 for given @xmath1 , by @xmath48 the set of all types of vectors @xmath49 in @xmath50 and by @xmath51 the set of all possible conditional types of vectors @xmath33 in @xmath52 for given @xmath53 . the set of vectors @xmath40 of type @xmath11 is denoted by @xmath54 and the family of vectors @xmath33 of conditional type @xmath15 for given @xmath55 of type @xmath12 by @xmath56 . the set of all possible pds @xmath11 on @xmath6 and pds @xmath12 on @xmath8 is denoted , correspondingly , by @xmath57 and @xmath58 . we need the following frequently used inequalities @xcite : @xmath59 @xmath60 for any type @xmath61 @xmath62 and for any type @xmath63 and @xmath64 @xmath65 the problem of optimal testing of _ multiple _ hypotheses was proposed by dobrushin @xcite , and was investigated in @xcite@xcite . fu and shen @xcite explored the case of two hypotheses when side information is absent . the problem concerning arbitrarily varying sources solved in @xcite was induced by the ideas of the paper of ahlswede @xcite . the case of two hypotheses with side information about states was considered in @xcite . in the same way as in @xcite from result on lao testing , the rate - reliability and the reliability - rate functions for arbitrarily varying source with side information were obtained in @xcite . the problem of multiple hypotheses lao testing for discrete stationary markov source of observations was solved by haroutunian @xcite@xcite . in @xcite haroutunian and grigoryan generalized results from @xcite , @xcite@xcite for multihypotheses lao testing by a non - informed statistician for arbitrarily varying markov source . here for clearness we expose the results on multiple hypotheses lao testing for the case of the most simple invariant object . let @xmath6 be a finite set of values of random variable ( rv ) @xmath9 . @xmath66 possible pds @xmath67 , @xmath68 , of rv @xmath9 characterizing the object are known . the statistician must detect one among @xmath66 alternative hypotheses @xmath69 , using sample @xmath70 of results of @xmath1 independent observations of the object . the procedure of decision making is a non - randomized test @xmath71 , it can be defined by division of the sample space @xmath47 on @xmath66 disjoint subsets @xmath72 , @xmath73 . the set @xmath74 consists of all samples @xmath40 for which the hypothesis @xmath69 must be adopted . we study the probabilities @xmath75 of the erroneous acceptance of hypothesis @xmath76 provided that @xmath69 is true @xmath77 the probability to reject hypothesis @xmath69 , when it is true , is also considered @xmath78 a quadratic matrix of @xmath79 error probabilities @xmath80 is the _ power _ of the tests . _ error probability exponents _ of the infinite sequence @xmath81 of tests , which we call _ reliabilities _ , are defined as follows : @xmath82 we see from ( [ 5 ] ) and ( [ 6 ] ) that @xmath83 the matrix @xmath84 called the _ reliabilities matrix _ of the tests sequence @xmath81 is the object of our investigation . we recognize that a sequence @xmath85 of tests is lao if for given positive values of @xmath86 diagonal , elements of matrix @xmath87 the procedure provides maximal values for all other elements of it . now we form the lao test by constructing decision sets noted @xmath88 . given strictly positive numbers @xmath89 , @xmath90 , we define the following regions : @xmath91 @xmath92 @xmath93 and corresponding values : @xmath94 @xmath95 @xmath96 @xmath97 * theorem 3.1 @xcite : * _ if for described model all conditional pds @xmath69 , @xmath73 , are different in the sense that , @xmath98 , @xmath99 , and the positive numbers @xmath100 , @xmath101 are such that the following @xmath86 inequalities , called compatibility conditions , hold @xmath102 @xmath103 @xmath104 , \,\ , m=\overline { 2,m-1},\ ] ] then there exists a @xmath105 sequence @xmath85 of tests , the reliabilities matrix of which @xmath106 is defined in @xmath107@xmath108 and all elements of it are positive . _ when one of inequalities @xmath109 is violated , then at least one element of matrix @xmath110 is equal to @xmath111 . the proof of theorem 3.1 is postponed to the appendix . it is worth to formulate the following useful property of reliabilities matrix of the lao test . * remark 3.1 @xcite : * _ the diagonal elements of the reliabilities matrix of the lao test in each row are equal only to the element of the last column : @xmath112 that is the elements of the last column are equal to the diagonal elements of the same row and due to @xmath113 are minimal in this row . consequently the first @xmath86 elements of the last column also can play a part as given parameters for construction of a lao test_. in @xcite ahlswede and haroutunian proposed a new aspect of the statistical theory investigation of models with many objects . this work developed the ideas of papers on information theory @xcite , @xcite , of papers on many hypotheses testing @xcite-@xcite and of book , devoted to research of sequential procedures solving decision problems such as ranking and identification . the problem of hypotheses testing for the model consisting of two independent and of two strictly dependent objects ( when they can not admit the same distribution ) with two possible hypothetical distributions were solved in @xcite . in @xcite the specific characteristics of the model consisting of @xmath114 objects each independently of others following one of given @xmath115 probability distributions were explored . in @xcite the model composed by stochastically related objects was investigated . the result concerning two independent markov chains is presented in @xcite . in this section we expose these results . let us now consider the model with three independent similar objects . for brevity we solve the problem for three objects , the generalization of the problem for @xmath116 independent objects will be discussed hereafter along the text . let @xmath117 , @xmath118 and @xmath119 be independent rvs taking values in the same finite set @xmath6 , each of them with one of @xmath66 hypothetical pds @xmath120 . these rvs are the characteristics of the objects . the random vector @xmath121 assumes values @xmath122 . let @xmath123 @xmath124 , @xmath125 @xmath126 , @xmath127 be a vector of results of @xmath1 independent observations of the family @xmath121 . the test has to determine unknown pds of the objects on the base of observed data . the detection for each object should be made from the same set of hypotheses : @xmath69 , @xmath73 . we call this procedure the _ compound test _ for three objects and denote it by @xmath128 , it can be composed of three individual tests @xmath129 , @xmath130 , @xmath131 for each of three objects . the test @xmath132 , @xmath133 , is a division of the space @xmath47 into @xmath134 disjoint subsets @xmath135 , @xmath136 . the set @xmath135 , @xmath73 , contains all vectors @xmath137 for which the hypothesis @xmath69 is adopted . hence test @xmath128 is realised by division of the space @xmath138 into @xmath139 subsets @xmath140 , @xmath141 , @xmath133 . we denote the infinite sequence of compound tests by @xmath142 . when we have @xmath116 independent objects the test @xmath142 is composed of tests @xmath143 , @xmath144 , ... , @xmath145 . the probability of the falsity of acceptance of hypotheses triple @xmath146 by the test @xmath128 provided that the triple of hypotheses @xmath147 is true , where @xmath148 , @xmath149 , @xmath133 , is : @xmath150 the probability to reject a true triple of hypotheses @xmath147 by analogy with ( [ 5 ] ) is defined as follows : @xmath151 we study corresponding reliabilities @xmath152 of the sequence of tests @xmath153 , @xmath154 @xmath155 definitions ( [ 17 ] ) and ( [ 18 ] ) imply ( compare with ( [ 7 ] ) ) that @xmath156 our aim is to analyze the reliabilities matrix @xmath157 of lao test sequence @xmath158 for three objects . we call the test sequence lao for the model with many objects if for given positive values of certain part of elements of reliabilities matrix the procedure provides maximal values for all other elements of it . let us denote by @xmath159 the reliabilities matrices of the sequences of tests @xmath160 , @xmath133 . the following lemma is a generalization of lemma from @xcite . * lemma 4.1 : * _ if elements @xmath161 , @xmath162 , @xmath163 are strictly positive , then the following equalities hold for @xmath164 , @xmath165 , @xmath166 : @xmath167-k]}e_{l_i\|m_i}{(\varphi^i)},\nonumber\\ \label{21 } & m_{k}=l_{k},\ , m_i\not = l_i , \,\,k=\overline { 1,3},\\ e _ { l_1,l_2,l_3\|m_1,m_2,m_3}(\phi ) & = e_{l_i\|m_i}{(\varphi^i)},\nonumber\\ \label{22 } & i=\overline { 1,3 } , \,m_{k}=l_{k},\ , m_i\not = l_i , \,\,k\in [ [ 1,2,3]-i].\\ \nonumber\end{aligned}\ ] ] _ equalities @xmath168 are valid also if @xmath169 for several pairs @xmath170 and several @xmath171 . the proof of lemma 4.1 is exposed in appendix . now we shall show how we can erect the lao test from the set of compound tests when @xmath172 strictly positive elements of the reliabilities matrix @xmath173 , @xmath174 and @xmath175 , @xmath90 , are preliminarily given . the following subset of tests : @xmath176 is distinguished by the property that when @xmath177 the elements + @xmath178 , @xmath179 and @xmath180 , @xmath90 , of the reliabilities matrix are strictly positive . really , because @xmath181 , @xmath73 , @xmath133 , then in view of ( [ 7 ] ) @xmath182 are also strictly positive . from equalities ( [ 20])([22 ] ) we obtain that the noted elements are strictly positive for @xmath183 and @xmath90 @xmath184 @xmath185 @xmath186 for given positive elements @xmath187 define the following family of decision sets of pds : @xmath188 @xmath189 @xmath190 @xmath191 define also the elements of the reliability matrix of the compound lao test for three objects : @xmath192-i],\\ e _ { m_1,m_2,m_3\|l_1,l_2,m_3}^ * & { \stackrel{\bigtriangleup}{=}}\sum\limits_{i\not= k}\inf\limits_{q\in r_{l_i}^{(i)}}d(q\|\|g_{m_i}),\nonumber\\ \label{26 } & m_{k}=l_{k},\,m_i\not = l_i,\,k=\overline { 1,3},\,i\in [ [ 1,2,3]-k],\\ \label{27 } e_{l_1,l_2,l_3\|m_1,m_2,m_3}^ * & { \stackrel{\bigtriangleup}{=}}\sum\limits_{i= 1}^3\inf\limits_{q\in r_{l_i}^{(i)}}d(q\|\|g_{m_i}),\ , m_i\not = l_i , i=\overline { 1,3}.\\ \nonumber \end{aligned}\ ] ] the following theorem is a generalization and improvement of the corresponding theorem proved in @xcite for the case @xmath193 . * theorem 4.1 @xcite : * _ for considered model with three objects , if all distributions @xmath69 , @xmath73 , are different , ( and equivalently @xmath98 , @xmath99 , @xmath194 ) , then the following statements are valid : _ \a ) when given strictly positive elements @xmath173 , @xmath174 and @xmath175 , @xmath90 , meet the following conditions @xmath195 @xmath196 , \\ \label{30 } e_{m , m , m\|m , m , m } & < \min\left [ \min\limits_{l= \overline { 1,m-1 } } e_{m , l , m\|m , m , m}^,\,\min\limits_{l= \overline { m+1,m } } d(g_l\|\|g_m)\right ] , \\ \label{31 } e_{m , m , m\|m , m , m } & < \min\left [ \min\limits_{l= \overline { 1,m-1 } } e_{m , m , l\|m , m , m}^,\,\,\min\limits_{l= \overline { m+1,m } } d(g_l\|\|g_m)\right ] , \\ \nonumber\end{aligned}\ ] ] then there exists a lao test sequence @xmath197 , the reliability matrix of which @xmath198 is defined in @xmath199@xmath200 and all elements of it are positive , \b ) if even one of the inequalities @xmath201@xmath202 is violated , then there exists at least one element of the matrix @xmath198 equal to @xmath111 . for the proof of theorem 4.1 see appendix . when we consider the model with @xmath116 independent objects the generalization of lemma 4.1 will take the following form . * lemma 4.2 : * _ if elements @xmath203 , @xmath204 , @xmath205 are strictly positive , then the following equalities hold for _ @xmath206 : @xmath207 for given @xmath208 strictly positive elements @xmath209 , + @xmath210 , .... , @xmath211 , @xmath212 , for @xmath116 independent objects we can find the lao test @xmath213 in a way similar to case of three independent objects . * comment 4.1 : * idea to renumber @xmath116-distributions from @xmath214 to @xmath215 and consider them as pds of one complex object offers an alternative way of testing for models with @xmath116 objects . we can give @xmath216 diagonal elements of corresponding large matrix @xmath164 and apply theorem 3.2 concerning one composite object . in this direct algorithm the number of the preliminarily given elements of the matrix @xmath217 would be greater ( because @xmath218 , @xmath219 ) but the procedure of calculations would be longer than in our algorithm presented above in this section . our approach to the problem gives also the possibility to define the lao tests for each of the separate objects , but the approach with renumbering of @xmath116-vectors of hypotheses does not have this opportunity . in the same time in the case of direct algorithm there is opportunity for the investigator to define preliminarily the greater number of elements of the matrix @xmath164 . in applications one of two approaches may be used in conformity with preferences of the investigator . some illustrations of exposed results are in an example concerning two objects . the set @xmath220 contains two elements and the following pds are given on @xmath6 : @xmath221 @xmath222 @xmath223 . as it follows from relations ( [ 24])([27 ] ) , several elements of the reliability matrix are functions of one of given elements , there are also elements which are functions of two , or three given elements . for example , in fig . 1 and fig . 2 the results of calculations of functions @xmath224 and @xmath225 are presented . for these distributions we have @xmath226 and @xmath227 . we see that , when the inequalities ( [ 28 ] ) or ( [ 31 ] ) are violated , @xmath228 and @xmath229 . we consider characteristics of procedures of lao testing of probability distributions of two related ( _ stochastically _ , _ statistically _ and _ strictly dependent _ ) objects . we use these terms for different kinds of dependence of two objects . let @xmath117 and @xmath118 be rvs taking values in the same finite set @xmath6 and @xmath230 be the space of all possible distributions on @xmath6 . let @xmath231 be a sequence of results of @xmath1 independent observations of the pair of objects . first we consider the model , which consists of two _ stochastically _ related objects . we name so the following more general dependence . there are given @xmath232 pds @xmath233 the first object is characterized by rv @xmath117 which has one of these @xmath232 possible pds and the second object is dependent on the first and is characterized by rv @xmath118 which can have one of @xmath234 conditional pds @xmath235 joint pd of the pair of objects is @xmath236 where @xmath237 the probability @xmath238 of @xmath1-vector @xmath239 is the following product : @xmath240 with @xmath241 and @xmath242 in somewhat particular case , when @xmath117 and @xmath118 are related _ statistically _ @xcite , @xcite , the second object depends on index of pd of the first object but not depends on value @xmath243 taken by the first object . the second object is characterized by rv @xmath118 which can have one of @xmath234 conditional pds @xmath244 . in the third case of _ strict _ dependence , the objects @xmath117 and @xmath118 can have only different distributions from the same given family of @xmath66 pds @xmath245 , @xmath246 , ... , @xmath247 . discussed in comment 4.1 the _ direct approach _ for lao testing of pds for two related objects , consisting in consideration of the pair of objects as one composite object and then use of theorem 3.1 , is applicable for first two cases @xcite . but now we consider also another approach . let us remark that test @xmath248 can be composed of a pair of tests @xmath249 and @xmath250 for the separate objects : @xmath251 . denote by @xmath252 , @xmath253 and @xmath153 the infinite sequences of tests for the first , for the second and for the pair of objects , respectively . let @xmath117 and @xmath118 be related _ stochastically_. for the object characterized by @xmath117 the non - randomized test @xmath254 can be determined by partition of the sample space @xmath52 on @xmath232 disjoint subsets @xmath255 , i.e. the set @xmath256 consists of vectors @xmath257 for which the pd @xmath258 is adopted . the probability @xmath259 of the erroneous acceptance of pd @xmath258 provided that @xmath260 is true , @xmath261 is defined by the set @xmath262 @xmath263 we define the probability to reject @xmath260 , when it is true , as follows @xmath264 corresponding error probability exponents are : @xmath265 it follows from ( [ 32 ] ) and ( [ 33 ] ) that @xmath266 for construction of the lao test we assume given strictly positive numbers @xmath267 , @xmath268 and we define regions @xmath269 , @xmath270 as in ( [ 8])([9 ] ) . for the second object characterized by rv @xmath118 depending on @xmath117 the non - randomized test @xmath271 , based on vectors @xmath272 and on the index of the hypothesis @xmath273 adopted for @xmath117 , can be given for each @xmath273 and @xmath274 by division of the sample space @xmath47 on @xmath275 disjoint subsets @xmath276 let @xmath277 the probabilities of the erroneous acceptance for @xmath278 are @xmath279 corresponding reliabilites are denoted @xmath280 and are defined as in ( [ 18 ] ) . we can upper estimate the probabilities of the erroneous acceptance for @xmath278 @xmath281 these upper estimates of @xmath282 for each @xmath283 we denote by @xmath284 consequently we can deduce that for @xmath285 new parameters @xmath286 are lower estimates for reliabilities @xmath287 + we can introduce for @xmath288 @xmath289 and also consider @xmath290 @xmath291 the corresponding estimates of the reliabilities of test @xmath292 , are the following @xmath293 @xmath294 it is clear from ( [ 34 ] ) and ( [ 35 ] ) that for every @xmath295 @xmath296 for given positive numbers @xmath297 , for @xmath298 and for each pair @xmath299 let us define the following regions and values : @xmath300 @xmath301 we denote by @xmath302 the matrix of lower estimates for elements of matrix @xmath303 * theorem 4.2 @xcite : * _ if for given @xmath304 all conditional pds @xmath305 , @xmath306 , are different in the sense that @xmath307 , for all @xmath298 , @xmath308 , when the strictly positive numbers @xmath309 , @xmath310, ... ,@xmath311 are such that the following compatibility conditions hold @xmath312 then there exists a sequence of tests @xmath313 , such that the lower estimates are defined in ( 4.29) ( 4.32 ) and are strictly positive . _ inequalities @xmath314 , @xmath315 are necessary for existence of test sequence with matrix @xmath316 of positive lower estimates having given elements @xmath317 in diagonal . let us define the following subsets of @xmath230 for given strictly positive elements @xmath318 , @xmath319 , @xmath320 : @xmath321 assume also that @xmath322 * theorem 4.3 @xcite : * _ if all pds @xmath260 , @xmath323 , are different , that is + @xmath324 , @xmath325 , @xmath326 , and all conditional pds @xmath305 , + @xmath327 , are also different for all @xmath328 in the sense that + @xmath307 , @xmath329 , then the following statements are valid . _ when given strictly positive elements @xmath330 and @xmath331 , @xmath332 , meet the following conditions @xmath333,\nonumber\\ \label { 32.c } & m_1=\overline { 2,m_1 - 1},\\ f_{l_1,m_2\|l_1,m_2 } & < \min[{\min\limits_{l_2= \overline { 1,m_2 - 1 } } f_{l_1,l_2\|l_1,m_2}^,\,\,\min\limits_{l_2= \overline { m_2 + 1,m_2 } } \inf\limits_{q\in { \cal r}_{l_1}}d(g_{l_2\|l_1}\|\|g_{m_2\|m_1}\|q)}],\nonumber\\ \label { 32.d } & m_2=\overline { 2,m_2 - 1},\\ \nonumber\end{aligned}\ ] ] then there exists a @xmath105 test sequence @xmath158 , the matrix of lower estimates of which @xmath334 is defined in ( [ 31.a])-([31.e ] ) and all elements of it are positive . when even one of the inequalities ( [ 32.a])-([32.d ] ) is violated , then at least one element of the lower estimate matrix @xmath335 is equal to @xmath111 . when @xmath117 and @xmath118 are related statistically @xcite , @xcite we will have instead of ( [ 34 ] ) , ( [ 35 ] ) @xmath336 , and @xmath337 . in that case we have error probabilities @xmath338 for the second object the conditional probabilities of the erroneous acceptance of pd @xmath305 provided that @xmath339 is true , for @xmath340 , @xmath341 , are the following @xmath342 the probability to reject @xmath339 , when it is true is denoted as follows @xmath343 thus in the conditions and in the results of theorems 4.2 and theorems 4.3 instead of conditional divergences @xmath344 , + @xmath345 we will have just divergences @xmath346 , + @xmath347 and in place of @xmath348 , @xmath349 , @xmath340 , @xmath350 , will be @xmath351 , @xmath352 , @xmath340 , @xmath341 . and in that case regions defined in @xmath353 , @xmath354 will be changed as follows : @xmath355 in case of two statistically dependent objects the corresponding regions will be @xmath356 so in this case the matrix of reliabilities @xmath357 , will have the following elements : @xmath358 theorem 4.4 : * @xcite _ if all pds @xmath260 , @xmath323 , are different , that is + @xmath324 , @xmath325 , @xmath326 , and all conditional pds @xmath305 , @xmath327 , are also different for all @xmath328 in the sense that @xmath359 , @xmath329 , then the following statements are valid . _ when given strictly positive elements @xmath318 and @xmath360 , @xmath361 , @xmath362 , meet the following compatibility conditions @xmath363,\nonumber\\ & m_1=\overline { 2,m_1 - 1 } , \nonumber\\ e_{l_1,m_2\|l_1,m_2 } & < \min\left [ { \min\limits_{l_2= \overline { 1,m_2 - 1 } } e_{l_1,l_2\|l_1,m_2}^,\,\,\min\limits_{l_2= \overline { m_2 + 1,m_2 } } d(g_{l_2\|l_1}\|\|g_{m_2\|m_1})}\right ] , \nonumber\\ & m_2=\overline { 2,m_2 - 1},\nonumber\end{aligned}\ ] ] then there exists a @xmath105 test sequence @xmath158 , the matrix of which @xmath198 is stated above and all elements of it are positive . when even one of the compatibility conditions is violated , then at least one element of the matrix @xmath198 is equal to @xmath111 . in @xcite bechhofer , kiefer , and sobel presented investigations on sequential multiple - decision procedures . this book concerns principally with a particular class of problems referred to as ranking problems . chapter 10 of the book by ahlswede and wegener @xcite is devoted to statistical identification and ranking . problems of distribution identification and distributions ranking for one object applying the concept of optimality developed in @xcite , @xcite , @xcite@xcite were solved in @xcite . in papers @xcite , @xcite , @xcite and @xcite identification problems for models composed with two independent , or strictly dependent objects were investigated . in @xcite , @xcite , @xcite and @xcite models considered in @xcite and @xcite and variations of these models inspired by the pioneering paper by ahlswede and dueck @xcite , applying the concept of optimality developed in @xcite , @xcite@xcite , @xcite , were studied . first we formulate the concept of the identification for one object , which was considered in @xcite . there are known @xmath364 possible pds , related with the object in consideration . identification gives the answer to the question whether @xmath365-th pd occured , or not . this answer can be given on the base of a sample @xmath40 and by a test @xmath366 . more precisely , identification can be considered as an answer to the question : is result @xmath367 of testing algorithm equal to @xmath365 ( that is @xmath368 ) , or not equal @xmath367 ( that is @xmath369 ) . there are two types of error probabilities of identification for each @xmath370 : the probability @xmath371 to accept @xmath367 different from @xmath365 , when @xmath365 is in reality , and the probability @xmath372 that @xmath365 is accepted by test @xmath373 , when @xmath365 is not correct . the probability @xmath371 coincides with the error probability of testing @xmath374 ( see ( 6 ) ) which is equal to @xmath375 . the corresponding reliability @xmath376 is equal to @xmath377 which satisfies the equality ( [ 7 ] ) . and what is the reliability approach to identification it is necessary to determine the dependence of optimal reliability @xmath378 upon given @xmath379 , which can be assigned a value satisfying conditions analogical to ( [ 15 ] ) . the result from paper @xcite is : theorem 5.1 : * _ in the case of distinct hypothetical pds @xmath380 , for a given sample @xmath40 we define its type @xmath11 , and when @xmath381 ( see ( [ 8])([10 ] ) ) we accept the hypothesis @xmath367 . under condition that the a priori probabilities of all @xmath66 hypotheses are positive the reliability of such identification @xmath382 for given @xmath383 is the following : _ @xmath384 we can accept the supposition of positivity of a priory probabilities of all hypotheses with loss of generality , because the pd which is known to have probability @xmath111 , that is being impossible , must not be included in the studied family . now let us consider the model consisting of two independent objects . let hypothetical characteristics of objects @xmath117 and @xmath118 be independent rvs taking values in the same finite set @xmath385 with one of @xmath66 pds . identification means that the statistician has to answer the question whether the pair of distributions @xmath386 occurred or not . now the procedure of testing for two objects can be used . let us study two types of error probabilities for each pair @xmath386 , @xmath387 . we denote by @xmath388 the probability , that pair @xmath386 is true , but it is rejected . note that this probability is equal to @xmath389 . let @xmath390 be the probability that @xmath386 is identified , when it is not correct . the corresponding reliabilities are @xmath391 and @xmath392 . our aim is to determine the dependence of @xmath393 on given @xmath394 . let us define for each @xmath365 , @xmath395 , the following expression : @xmath396.\ ] ] * theorem 5.2 @xcite:_for the model consisting of two independent objects if the distributions @xmath69 , @xmath68 , are different and the given strictly positive number @xmath397 satisfy condition @xmath398,\ ] ] then the reliability @xmath399 is defined as follows : @xmath400 @xmath401,\ ] ] where @xmath402 and @xmath403 are determined by @xmath404_. now we will present the lower estimates of the reliabilities for lao identification for the _ dependent _ object which can be then applied for deducing the lower estimates of the reliabilities for lao identification of two _ related _ objects . there exist two error probabilities for each @xmath405 : the probability @xmath406 to accept @xmath407 different from @xmath408 , when @xmath408 is in reality , and the probability + @xmath409 that @xmath408 is accepted , when it is not correct . the upper estimate @xmath410 for @xmath411 is already known , it coincides with @xmath412 which is equal to + @xmath413 . the corresponding @xmath414 is equal to + @xmath415 , which satisfies the equality ( [ 19 ] ) . we determine the optimal dependence of @xmath416 upon given + @xmath417 . theorem 5.3 @xcite : * _ in case of distinct pds @xmath418 , under condition that a priori probabilities of all @xmath275 hypotheses are strictly positive , for each @xmath419 the estimate of @xmath420 for given @xmath421 is the following : _ @xmath422 @xmath423 the result of the reliability approach to the problem of identification of the probability distributions for two _ related _ objects is the following . * theorem 5.4 : * _ if the distributions @xmath260 and @xmath339 , @xmath323 , @xmath424 , are different and the given strictly positive number @xmath425 satisfies the condition @xmath426,\end{aligned}\ ] ] + @xmath427,\end{aligned}\ ] ] then the lower estimate @xmath428 of the reliability + @xmath399 can be calculated as follows @xmath429 @xmath430,\ ] ] where @xmath431 and @xmath432 are determined respectively by ( [ 12 ] ) and ( [ 40 ] ) . _ the particular case , when @xmath117 and @xmath118 are related _ statistically _ , was studied in @xcite , @xcite . this section is devoted to description of characteristics of lao hypotheses testing with permission of decision rejection for the model consisting of one or more objects . the multiple hypotheses testing problem with possibility of rejection of decision for arbitrarily varying object with side information and for the model of two or more independent objects was examined by haroutunian , hakobyan and yessayan @xcite , @xcite . these works ware induced by the paper of nikulin @xcite concerning two hypotheses testing with refusal to take decision . an asymptotically optimal classification , in particular hypotheses testing problem with rejection of decision ware considered by gutman @xcite . in this section we consider multiple statistical hypotheses testing with possibility of rejecting to make choice between hypotheses concerning distribution of a discrete arbitrarily varying object . the arbitrarily varying object is a generalized model of the discrete memoryless one . let @xmath6 be a finite set of values of rv @xmath9 , and @xmath8 is an alphabet of states of the object . @xmath66 possible conditional pds of the characteristic @xmath9 of the object depending on values @xmath18 of states , are given : @xmath433 but it is not known which of these alternative hypotheses @xmath434 , @xmath73 , is real pd of the object . the statistician must select one among @xmath66 hypotheses , or he can withdraw any judgement . it is possible for instance when it is supposed that real pd is not in the family of @xmath66 given pds . an answer must be given using the vector of results of @xmath1 independent experiments @xmath435 and the vector of states of the object @xmath436 , @xmath437 , @xmath438 . the procedure of decision making is a non - randomized test @xmath439 , it can be defined by division of the sample space @xmath47 for each @xmath440 on @xmath441 disjoint subsets @xmath442 , @xmath443 . the set @xmath444 , @xmath445 , consists of vectors @xmath40 for which the hypothesis @xmath446 is adopted , and @xmath447 includes vectors for which the statistician refuses to take a certain answer . we study the probabilities of the erroneous acceptance of hypothesis @xmath446 provided that @xmath434 is true @xmath448 when decision is declined , but hypothesis @xmath434 is true , we consider the following probability of error : @xmath449 if the hypothesis @xmath434 is true , but it is not accepted , or equivalently while the statistician accepted one of hypotheses @xmath446 , @xmath450 , @xmath451 , or refused to make decision , then the probability of error is the following : @xmath452 corresponding reliabilities are defined similarly by to ( [ 5 ] ) : @xmath453 it also follows that for every test @xmath81 @xmath454 the matrix @xmath455 is the reliabilities matrix of the tests sequence @xmath81 for the described model . we call the test lao for this model if for given positive values of certain @xmath66 elements of the matrix @xmath456 the procedure provides maximal values for other elements of it . for construction of lao test positive elements @xmath457 , @xmath458 , @xmath459 are supposed to be given preliminarily . the optimal dependence of error exponents was determined in @xcite . this result can be easily generalized for the case of an arbitrarily varying markov source . for brevity we consider the problem for two objects , the generalization of the problem for @xmath116 independent objects will be discussed along the text . let @xmath117 and @xmath118 be independent rvs taking values in the same finite set @xmath6 with one of @xmath66 pds @xmath460 , @xmath68 . these rvs are the characteristics of the corresponding independent objects . the random vector @xmath461 assumes values @xmath462 . let @xmath463 @xmath464 , @xmath125 @xmath465 , be a vector of results of @xmath1 independent observations of the pair of rvs @xmath461 . on the base of observed data the test has to determine unknown pds of the objects or withdraw any judgement . the selection for each object should be made from the same set of hypotheses : @xmath69 , @xmath73 . we call this procedure the compound test for two objects and denote it by @xmath128 , it can be composed of two individual tests @xmath129 , @xmath130 for corresponding objects . the test @xmath132 , @xmath466 , can be defined by division of the space @xmath47 into @xmath441 disjoint subsets @xmath135 , @xmath467 . the set @xmath135 , @xmath73 contains all vectors @xmath468 for which the hypothesis @xmath69 is adopted and @xmath469 includes all vectors for which the test refuses to take a certain answer . hence @xmath128 is division of the space @xmath470 into @xmath471 subsets @xmath472 , @xmath473 . we again denote the infinite sequences of tests by @xmath474 , @xmath475 , @xmath144 . let @xmath476 be the probability of the erroneous acceptance of the pair of hypotheses @xmath477 by the test @xmath128 provided that the pair of hypotheses @xmath478 is true , where @xmath479 , @xmath480 , @xmath481 , @xmath466 : @xmath482 when the pair of hypotheses @xmath478 , @xmath483 is true , but we decline the decision the corresponding probabilities of errors are : @xmath484 or @xmath485 or @xmath486 if the pair of hypotheses @xmath478 is true , but it is not accepted , or equivalently while the statistician accepted one of hypotheses @xmath477 , or refused to make decision , then the probability of error is the following : @xmath487 we study reliabilities @xmath352 of the sequence of tests @xmath153 , @xmath488 definitions ( [ 60 ] ) and ( [ 61 ] ) imply that @xmath489 we can erect the lao test from the set of compound tests when @xmath490 strictly positive elements of the reliability matrix @xmath491 and @xmath492 , @xmath73 , are preliminarily given ( see @xcite ) . * remark 6.1 : * it is necessary to note that the problem of reliabilities investigation for lao testing of many hypotheses with possibility of rejection of decision for the model consisting of two or more independent objects can not be solved by the direct method of renumbering . a broad class of statistical problems arises in the framework of hypothesis testing in the spirit of identification for different kinds of sources , with complete or partial side information or without it . paper @xcite is a start . " @xcite . in this paper , we exposed solutions of a part of possible problems concerning algorithms of distributions optimal testing for certain classes of one , or multiple objects . for the same models pd optimal identification is discussed again in the spirit of error probability exponents optimal dependence . but these investigations can be continued in plenty directions . some problems formulated in @xcite and @xcite , particularly , concerning the remote statistical inference formulated by berger @xcite , examined in part by ahlwede and csiszr @xcite and han and amari @xcite still rest open . all our results concern with discrete distribution , it is necessary to study many objects with general distributions as in @xcite . for multiple objects multistage and sequential testing @xcite can be also considered . problems for many objects are present in statistics with fuzzy data @xcite , bayessian detection of multiple hypotheses testing @xcite and geometric interpretations of tests @xcite . _ proof of theorem 3.1 : _ probability @xmath493 for @xmath494 can be presented as follows : @xmath495\right \}}.\label { 70}\end{aligned}\ ] ] let us consider the sequence of tests @xmath496 defined by the sets @xmath497 each @xmath40 is in one and only in one of @xmath498 , that is @xmath499 really , for @xmath500 , @xmath501 , for each @xmath502 let us consider arbitrary @xmath503 . it follows from ( [ 8 ] ) and ( [ 10 ] ) that there exists type @xmath61 such that @xmath504 and @xmath505 . from ( [ 15 ] ) and ( [ 12 ] ) we have @xmath506 . from definition of @xmath498 we see that @xmath507 . definitions ( [ 13 ] ) , ( [ 15 ] ) and ( [ 10 ] ) show also that @xmath508 now , let us remark that for @xmath90 , using ( [ 1 ] ) , ( [ 2 ] ) , ( [ 4])([6 ] ) and ( [ 70 ] ) we can estimate @xmath509 as follows : for @xmath513 , @xmath73 , @xmath99 , using ( [ 1 ] ) , ( [ 2 ] ) , ( [ 4])([6 ] ) and ( [ 70 ] ) , we can obtain similar estimates : @xmath514 now let us prove the inverse inequality : @xmath515 according to the definition ( [ 6 ] ) @xmath516 , taking into account ( [ 72 ] ) , ( [ 73 ] ) and the continuity of the functional @xmath517 we obtain that + @xmath518 exists and in correspondence with ( [ 12 ] ) equals to @xmath519 . thus @xmath520 , @xmath73 , @xmath521 , @xmath451 . similarly we can obtain upper and lower bounds for @xmath522 , @xmath73 . applying the same resonnement we get that the reliability @xmath523 . by the definition ( [ 7 ] ) @xmath524 . the proof of the first part of the theorem will be accomplished if we demonstrate that the sequence of tests @xmath85 is lao , that is for given @xmath525 and every sequence of tests @xmath526 for all @xmath527 , @xmath528 these conditions are equivalent for @xmath1 large enough to the inequalities @xmath532 let us examine the sets @xmath533 , @xmath136 . this intersection can not be empty , because in that case @xmath534 let us show that @xmath535 , @xmath536 , @xmath537 . if there exists @xmath11 such that @xmath538 and @xmath539 , then @xmath540 when @xmath541 , we also obtain that @xmath542 thus it follows that @xmath543 , which in turn according to ( [ 7 ] ) provides that @xmath544 . from condition ( [ 15 ] ) it follows that @xmath545 , for all @xmath546 , hence @xmath547 for all @xmath546 , which contradicts to ( [ m ] ) . hence we obtain that @xmath548 for @xmath212 . the intersection @xmath549 is empty too , because otherwise @xmath550 which contradicts to ( [ 74 ] ) , hence @xmath551 , @xmath73 . the proof of the second part of the theorem 3.1 is simple . if one of the conditions ( [ 15 ] ) is violated , then from ( [ 12])([14 ] ) it follows that at least one of the elements @xmath552 is equal to 0 . for example , let in ( [ 15 ] ) the @xmath553-th condition be violated . it means that @xmath554 , then there exists @xmath555 such that @xmath556 from latter and ( [ 12 ] ) we obtain that @xmath557 . _ proof of lemma 4.1 : _ it follows from the independence of the objects that @xmath558 @xmath559-k ] } \alpha_{l_i\|m_i}(\varphi_{n}^i),\ ] ] @xmath560 @xmath561-i ] } \left ( 1-\alpha_{l_k\|m_k}(\varphi_{n}^i)\right ) , \ ] ] @xmath562 remark that here we consider also the probabilities of right ( not erroneous ) decisions . because @xmath563 are strictly positive then the error probability + @xmath564 tends to zero , when @xmath565 . according this fact we have @xmath566 from definitions ( [ 18 ] ) , equalities ( [ 13.a])([13.c ] ) , applying ( [ 14.a ] ) we obtain relations ( [ 20])([22 ] ) . _ proof of theorem 4.1 : _ the test @xmath567 , where @xmath568 , @xmath569 are lao tests of objects @xmath570 , belongs to the set @xmath571 . our aim is to prove that such @xmath158 is a compound lao test . conditions ( [ 28])([31 ] ) imply that inequalities analogous to ( [ 15 ] ) hold simultaneously for tests for three separate objects . according to remark 3.1 for lao test @xmath577 , @xmath569 , we obtain that ( [ 123 ] ) meets conditions ( [ 15 ] ) of theorem 3.1 for each test @xmath177 , @xmath578 , @xmath133 , hence it follows from ( [ 6 ] ) that @xmath579 are also strictly positive . thus for a test @xmath177 conditions of lemma 4.1 are fulfilled and the elements of the reliability matrix @xmath164 coincide with elements of matrix @xmath580 , @xmath569 , or sums of them . then from definition of lao test it follows that @xmath581 , then @xmath582 . consequently @xmath158 is a lao test and @xmath583 verify @xmath199@xmath200 . \b ) when even one of the inequalities @xmath201@xmath202 is violated , then at least one of inequalities ( [ 123 ] ) is violated . then from theorem 3.2 one of elements @xmath584 is equal to zero . suppose @xmath585 , then the elements + @xmath586 . ahlswede , r. ( 1979,1980 ) . coloring hypergraphs : a new approach to multi - user source coding . _ journal of combinatorics , information and system sciences_. i , vol . 4 , no . 76 - 115 , ii , vol . 5 , no . 3 , pp . 220268 . ahlswede , r. , aloyan , e. and haroutunian , e. ( 2006 ) . on logarithmically asymptotically optimal hypothesis testing for arbitrarily varying source with side information . _ lecture notes in computer science _ , vol . 4123 , general theory of information transfer and combinatorics " , springer verlage , pp . 457 461 . ahlswede , r. and haroutunian , e. ( 2006 ) . on logarithmically asymptotically optimal testing of hypotheses and identification . _ lecture notes in computer science _ , vol . 4123 , general theory of information transfer and combinatorics " , springer verlage , pp . 462 478 . csiszr , i. ( 1998 ) . information thoretic methods in probability and statistics . _ shannon lecture at isit in ulm germany 1997 , ieee information theory society newsletter_. vol . 1 , pp . 1 and 2130 . dobrushin , r. , pinsker , m. and shiryaev , a. ( 1963 ) . application of the notion of entropy in the problems of detecting a signal in noise . ( in russian ) , _ lithuanian mathematical transactions_. vol . 1 , pp . 107122 . haroutunian , e. ( 1967 ) . upper estimate of transmission rate for memoryless channel with countable number of output signals under given error probability exponent . ( in russian ) , _ 3rd all union conf . on theory of information transmission and coding , uzhgorod , publishing hous of the uzbek ac . , tashkent_. pp . 83 - 86 . haroutunian , e. ( 1988 ) . many statistical hypotheses : interdependence of optimal test s error probabilities exponents . ( in russian ) , abstract of the report on the 3rd all - union school - seminar , _ program - algorithmical software for applied multi - variate statistical analysis " _ , tsakhkadzor , part 2 , pp . 177178 . haroutunian , e. ( 1989 ) . asymptotically optimal testing of many statistical hypotheses concerning markov chain . ( in russian ) , _ 5-th inter . vilnius conferance on probability theory and mathem . statistics _ , vol . 1 , ( a - l ) , pp . 202203 , 1989 . haroutunian , e. and grigoryan , n. ( 2007 ) . on reliability approach for testing of many distributions for pair of markov chains . _ transactions of iiap of nas of ra , mathematical problems of computer science_. vol . 99 - 96 . haroutunian , e. and grigoryan , n. ( 2009 ) . on arbitrarily varying markov source coding and hypothesis lao testing by non - informed statistician . _ proc . of ieee int . inform . theory _ , seoul , south korea , pp . 981 - 985 . haroutunian , e. and hakobyan , p. ( 2004 ) . on multiple hypotheses testing by informed statistician for arbitrarly varying object and application to source coding . _ transactions of iiap of nas of ra and of ysu , mathematical problems of computer science_. vol . 36 - 46 . haroutunian , e. and hakobyan , p. ( 2010 ) . remarks about reliable identification of probability distributions of two independent objects . _ transactions of iiap of nas of ra , mathematical problems of computer science_. vol . 33 , pp . 9194 . haroutunian , e. , hakobyan , p. and yessayan , a. ( 2011 ) . many hypotheses lao testing with rejection of decision for arbitrarily varying object . _ transactions of iiap of nas of ra , mathematical problems of computer science _ , vol . 77 - 85 . haroutunian , e. , hakobyan , p. and yessayan , a. ( 2011 ) . on multiple hypotheses lao testing with rejection of decision for many independent objects . _ proceedings of international conference csit 2011_. pp . 141 - 144 . haroutunian , e. , haroutunian , m. and harutyunyan , a. ( 2008 ) . reliability criteria in information theory and in statistical hypothesis testing . _ foundations and trends in communications and information theory_. vol . 4 , no . 2 - 3 . haroutunian , e. and yessayan , a. ( 2007 ) . on logarithmically asymptotically optimal hypothesis testing for pair of statistically dependent objects . _ transactions of iiap of nas of ra , mathematical problems of computer science _ , vol . 97 103 . haroutunian , e. and yessayan , a. ( 2008 ) . on optimal hypotheses testing for pair of stochastically coupled objects . _ transactions of iiap of nas of ra , mathematical problems of computer science _ , vol . 49 59 . haroutunian , e. , yessayan , a. and hakobyan , p. ( 2010 ) . on reliability approach to multiple hypotheses testing and identification of probability distributions of two stochastically coupled objects . _ international journal informations theories and applications"_. vol . 259288 . haroutunian , e. and yessayan , a. ( 2011 ) . on reliability approach to multiple hypotheses testing and to identification of probability distributions of two stochastically related objects . _ ieee international symposium on information theory , proceedings _ . 2671 - 2675 . longo , g. and sgarro , a. ( 1980 ) . the error exponent for the testing of simple statistical hypotheses : a combinatorial approach . _ journal of combinatorics , information and system sciences_. vol . 5 , no . 1 , pp . 58 - 67 , 1980 . nikulin , m. ( 1986 ) . on one result of l. n. bolshev from theory of hypotheses statistical testing . ( in russian ) , _ studies on mathematical statistics . notes of scientific seminars of saint - petersburg branch of the mathematical institute _ 129137 . yessayan , a. ( 2009 ) . on reliability approach to identification of probability distributions of two statistically dependent objects . _ transactions of iiap of nas of ra , mathematical problems of computer science_. vol . 32 , pp . 65 69 .	we survey a series of investigations of optimal testing of multiple hypotheses concerning various multiobject models . these studies are a bright instance of application of methods and technics developed in shannon information theory to solution of typical statistical problems .

Subsets and Splits